Just as the buzz around DeepSeek was beginning to fade, Chinese AI has made waves again with the AI agent “Manus,” launched on March 6th, 2025. Today, we’re here to unpack the Manus launch, explore the business model of Manus’ parent company, and offer a glimpse into the mind of Xiao Hong 肖弘, the founder behind China’s latest viral AI product.

What is Manus?

Manus claims to be the world’s first general-purpose AI agent. It ostensibly outperforms OpenAI’s ChatGPT Deep Research on the General AI Assistants (GAIA) benchmark. Currently in beta testing, access is restricted to those with invitation codes, which are reportedly being listed second-hand for 50,000-100,000 RMB (whether anyone actually paid that much is another question). Users report impressive performance in basic tasks, for instance rebooking airline flights, beyond what Anthropic’s Computer Use and OpenAI’s Operator have thus far provided to users. The product is also experiencing slowing response times, hinting that Monica.ai may be struggling to scale up compute to meet skyrocketing demand.

Manus started in 2022 as an AI-powered browser plugin, backed by ZhenFund (真格基金). In 2023, the company secured Series A funding led by Tencent (腾讯) and Sequoia Capital China (红杉资本中国). What began as a simple “ChatGPT for Google” browser plugin has since evolved into a full-fledged AI agent.

Monica, the company that developed Manus, operates from Wuhan, rather than from China's major tech hubs like Beijing or Shanghai. In early 2024, ByteDance attempted to acquire Monica for $30 million, but founder Xiao Hong (肖弘) turned down the offer. ByteDance’s plan was to absorb Monica’s team and technology into its Doubao AI ecosystem, a move that would have diluted Monica’s distinct market position. Instead, Monica closed a new funding round at the end of 2024, reaching an estimated valuation of nearly $100 million.

The exact AI models powering Manus remain unclear. The company claims to use multiple models for different tasks. Notably, when prompted to reveal its own system files, Manus reveals it may be powered by Anthropic’s Claude models — which would make operating in China illegal. This probably explains why Monica’s website appears to be blocked in China.

Edit: confirmed by co-founder.

Anyway, the fact that Manus appears to disclose more than it should hints at broader potential security vulnerabilities.

Who is behind Manus?

Founder & CEO, Xiao Hong (肖弘), is a serial entrepreneur and a graduate of Wuhan’s Huazhong University of Science and Technology (华中科技大学). He first made his mark by building WeChat-related tools as a student, admitting that while his “academic performance was quite poor,” he partnered with more technical classmates to build tools. In 2015, he launched Nightingale Technology (夜莺科技) and created Yiban Assistant (壹伴助手), a WeChat management tool that secured early backing from ZhenFund (真格基金).

By 2019, Xiao saw a bigger opportunity in enterprise WeChat tools and developed Weiban Assistant (微伴助手). His timing was perfect—when rival WeTool (微商工具WeTool) was shut down in 2020, Weiban became the go-to alternative, attracting investment offers from Sequoia Capital China (红杉资本中国) and Youzan (有赞). Eventually, Minglue Technology (明略科技) acquired Weiban, marking Xiao’s first major financial success.

Sensing the potential of large AI models, Xiao left Minglue in 2022 to create Monica.ai, originally designed as a “ChatGPT for Google” browser plugin.

Co-founder & Chief Scientist, Ji Yichao (季逸超) dropped out of high school at 17 to develop Mammoth Browser (猛犸浏览器). His talent caught the eye of Sequoia Capital China’s Zhou Kui (周逵), who introduced him to investor Xu Xiaoping (徐小平). Xu invested 1.5 million yuan, giving Ji complete creative freedom. Recognizing the large potential of LLMs, Ji joined Xiao Hong to start Monica in late 2022.

Interview Quotes

Unlike DeepSeek’s media-shy Liang Wenfeng, Xiao Hong has done a ton of press. Below are selected translations from several in-depth interviews with Monica’s founder and CEO, Xiao Hong, offering insights into his vision, strategy, and the future of AI agents.

The vibe of Xiao Hong’s interviews is distinct from the AGI-driven idealism blended with national pride we’ve seen from the founders of DeepSeek and Unitree. Xiao is pragmatic and focused on profitability rather than research. A newly published three-hour podcast with Xiao opens with offering this piece of advice:

“I remember there was a Northeastern Chinese restaurant near my university. I made enough money to treat my tech club friends to dinner there every day. Here’s a tip for the audience: if you’re in college, take your most talented classmates out for meals as often as you can. If you wait until after graduation to recruit them for your startup, you’ll have to treat them to Michelin-starred restaurants instead.”

In another interview from January 2024, Xiao openly admits that he didn’t initially believe in AI’s potential, and “remained cautious” despite the hype surrounding GPT-3.5 in the fall of 2022. He describes coming to two conclusions about AI investment, which eventually led him to focus on AI products as opposed to chasing AGI with foundational model research:

"First, I wouldn’t consider working on big models without sufficient business scale. Second, I believe that in China, big model services will eventually integrate fully with cloud computing. I’ve discussed this with our CTO and believe that cloud computing companies will provide customized deployment services, so we don’t need to dive into that ourselves."

…

"I focused more on what big models could do, and what kind of applications I could build with them. In the beginning, many people were financing based on concepts, but by the second half of the year, both domestic and international, there was much less of that. Everyone was returning to business rationality, focusing on finding PMF (Product-Market Fit). By February of 2023, I had a conversation with an investor focused on big models, and no matter how I asked, they refused to talk about products. They weren’t discussing technology or plans. By March, the product’s valuation plummeted. People realized that simply building a single application based on big models might not work, and that’s when the consensus started forming: either focus on technological breakthroughs or work on relatively closed-loop application scenarios."

…

"In March and April of 2023, the fastest-growing product outside of ChatGPT globally was Poe. It was essentially a shell around a big model, and I told investors that if you can perfect the shell, that’s still a big deal. So we decided to do it too, and instead of resisting the demand, we decided to embrace it. In the first half of 2023, Monica integrated all the major models because that’s what the users wanted, and we started by doing that, figuring out how to find more use cases step by step."

Monica’s business model focuses on catering to the overseas market, which likely explains why their website is devoid of any reference to being based in China. Besides English, Monica’s website has dedicated versions in traditional and simplified Chinese, as well as Russian, Ukrainian, Bahasa Indonesian, Persian, Arabic, Thai, Vietnamese, Hindi, Japanese, Korean, and a slew of European languages.

In Xiao’s words, “We chose to target the overseas ToC market because I felt it was a larger, more commercially viable market. The domestic market seemed a bit more challenging.” Their focus shows: in contrast to DeepSeek’s very low key model launches, Manus’ launch came with a whole sophisticated press push like one you would see out of a YC startup, complete with a very well-produced English-language launch video and early access for select YouTubers and twitter influencers.

International expansion comes with its own difficulties, but Xiao believes those challenges made Monica stronger as a company. He’s recently argued that China would benefit from having more firms look abroad:

Xiao Hong: I think we are still in a great era with many opportunities…. First, it's the AI era. Second, I think we are also in a great era of globalization. I'm not a geopolitical expert, but it seems like every country has its own problems — internally, everyone has their own issues. So overall, the world is becoming more conservative and more isolationist, right? But at the same time, no one wants others to be isolationist; they only want to be isolationist themselves. So, everyone hopes that their own entrepreneurs will think more globally.

I believe China’s entrepreneurs of today should be more aggressive in globalizing. If we see overseas markets as better opportunities, it’s not just about market-driven decisions — we should step into international markets to gain experience. We need to participate in global competition, rather than just competing in the markets we are familiar with.

By the way, this process requires a lot of things. When I started this company, none of our founders had lived abroad for an extended period. Everyone’s English proficiency peaked in high school and declined in college! [

I once joked that if, at the same time, there was another founder who had lived in the U.S. and was placed next to me, I would have chosen to work with that founder myself. But this shouldn’t be the way we compare things — it should be about doing our own thing. Secondly, I had a simple belief at the time: the global market is much bigger, and the market itself will provide the tuition fees for founders to learn. (Laughter)

Besides the AI era, another crucial topic is that we are now thinking about things with a globalized mindset.

Unsurprisingly, this business model also relies on collecting vast amounts of user data. Monica’s free Chrome extension requests expansive access to browser data, including permission to log keystrokes, and Manus “crawls” devices to make suggestions. Xiao is betting that widespread adoption of these products will unlock a treasure trove of monetizable insights.

“The data we collect through our browser plugin is critical. Even though this might not guarantee success, it’s a step in the right direction. The private data we gather, along with contextual information, will help differentiate us from the competition. This is one of the key assets we need to grow.”

Xiao is explicitly describing an intent to build an incumbent advantage on a foundation of user data, and TikTok demonstrates how effective that strategy can be. Reliance on eventual mass adoption could partially explain the high-publicity invite-only launch strategy for Manus (although limited access to compute is also certainly a factor).

That said, he is aware that the politics exist and could get in the way of a Chinese-owned AI agent gaining widespread adoption abroad. He spoke about it in a recent podcast alluding to NeZha 2.

I've come to understand that many things are beyond your control. You should focus on doing well with the things you can control. There are truly too many things beyond our control, like geopolitics. You simply can't control it—you can only treat it as an input, but you can't control it.

I recently asked DeepSeek to explain three terms 贪 (greed), 嗔 (hatred), and 痴 (ignorance) [the ‘three poisons’ of Buddishm recently spotighted in the truly excellent animated movie NeZha 2]. It explained it very well: greed is attachment to favorable circumstances; anger is dissatisfaction with adverse circumstances; and ignorance is not understanding the truth of the world. The "truth of the world" is very profound, so I won't discuss that. But greed and anger are problems many people encounter, as are attatchment to favorable circumstances and dissatisfaction with unfavorable ones.

This business-minded pragmatism shines through in Xiao’s vision for the future — instead of techno-optimist visions of AI-powered drug discoveries or a moon colony staffed by robots, he imagines a world where humanity can return to a glorious past:

“I think that the white-collar lifestyle may be a detour for mankind. If you look at it in terms of a curve or over a longer period, say thousands of years, or even the ten-thousand-year span of human history — it's actually quite rare for people to sit in one place and engage in intense mental work without much physical activity. This is probably only a phenomenon of the past hundred years.

For a longer time in history, maintaining physical health and developing spiritual civilization have gone hand in hand. In ancient times, people also needed spiritual and cultural development, but that involved physical labor as well, which helped strengthen their bodies.

In the past hundred years, however, issues like diabetes and high blood pressure have become widespread because people work in this sedentary way. If we look at humanity as a whole, sitting and working for eight or more hours a day is an anomaly.

If AI can take over these tasks, then people can work fewer hours and go back to living more like they did in the past — focusing more on spiritual and cultural enrichment while also taking better care of their physical health.”

To close, here’s a quote from Xiao about how it feels to live through history:

Xiao Hong: From the time I was born in the 90s until now… there have been significant shifts, from PCs to mobile, then the semiconductor industry, which has been booming behind the scenes, the rise of the internet, and now artificial intelligence. I feel like these opportunities are emerging very intensively. When I watched The Godfather, I realized that if I had lived in that era — it was also a time of change — but if you lived in certain periods, you might not have witnessed such rapid technological progress. Sometimes, when we read history books or ancient texts, it feels like things barely changed, which I think would be a little frustrating!

From an Anon contributor who would know:

TLDR: U.S. export controls targeting China’s AI capabilities focus primarily on limiting training hardware but overlook the growing importance of inference compute as a key driver of AI innovation. Current restrictions don’t effectively limit China’s access to inference-capable hardware (such as NVIDIA’s H20) and don’t account for China’s strong inference efficiency. While China’s fragmented computing infrastructure has historically been a disadvantage, the shift towards inference-heavy AI paradigms positions their compute ecosystem to be more utilized and valuable. As reasoning models, agentic AI, and automated AI research elevate the role of inference to advancing AI capabilities, the US should urgently strengthen export controls to hinder China’s inference capacity and develop a coherent open-source AI strategy to maintain competitive advantage.

The Export Control Status Quo is Broken

The global AI competition is unfolding along two critical axes: innovation — the development of advanced AI capabilities — and diffusion — deploying and scaling those capabilities. The United States has prioritized outpacing China in AI innovation by focusing on pre-training as the main driver of progress. However, a new paradigm is emerging where inference, not just training, is becoming central to advancing AI capabilities.

This shift has significant implications for U.S. AI policy. Current export controls aim to limit China’s ability to train frontier AI models by restricting access to advanced chips, based on the belief that scaling pre-training is the primary driver of AI progress. By limiting China’s access to compute resources, these controls aimed to slow its AI development.

Yet, these same controls are far less effective at restricting China’s inference capabilities — exposing a critical gap in U.S. strategy. As inference becomes more central to AI innovation, current policies are increasingly misaligned with the realities of AI development. To effectively counter China’s growing inference capabilities, the U.S. must strengthen its export controls.

Inference Compute is a Key Driver of AI Innovation

The AI landscape is evolving beyond the scaling pre-training paradigm that dominated recent years. Emerging solutions are shifting innovation toward a paradigm where inference compute — not just training compute — has become a critical driver of AI progress.

Three interconnected trends are driving the link between inference and AI capabilities:

Reasoning Models

These models require significantly more inference than traditional LLMs, leveraging test-time scaling laws that suggest a link between amount of inference compute and model performance. Inference demand is further driven by a feedback loop that accelerates AI capabilities: reasoning models generate high-quality synthetic data, which enhances base models via supervised fine-tuning (SFT). These stronger models can be adapted into stronger reasoning systems, creating even better synthetic data and fueling continuous capability gains.

Agentic AI

AI agents — systems capable of taking autonomous actions in complex environments to pursue goals — are often powered by reasoning models, which drives up inference demand. Many agents have access to external tools and environments such as code execution environments, databases, and web search, which enhance their capabilities by enabling them to retrieve information, plan, and interact with digital and physical environments.

Some agents continuously learn by interacting with their environment via reinforcement learning. Unlike standard language models that handle one-off queries, agentic AI systems require persistent inference as they continuously interact with external environments, adapt to new information, and make complex, multi-step decisions in real time — significantly increasing overall inference requirements.

Automated AI Research

Automated AI researchers can design new architectures, improve training methods, run experiments, and iterate on findings. Scaling in this paradigm requires both inference compute to power research agents and training compute to execute their proposed experiments. Greater inference capacity allows more of these systems to operate in parallel, expanding both the breadth and depth of AI exploration. This, in turn, enlarges the search space they can navigate and directly increases the rate of AI innovation.

Greater inference also enhances research agents through iterative reasoning, self-play debates, and automated evaluation — capabilities already demonstrated in AI-driven scientific discovery. As these automated systems achieve early breakthroughs, they become better at identifying promising research directions and architectural improvements, potentially setting off a compounding cycle of progress. Thus, even small initial advantages in inference capacity can compound, leading to a significant, potentially decisive, lead in AI capabilities.

In an era of reasoning models, agents, and automated AI research, inference capacity is not just an enabler — it is a primary determinant of the speed and trajectory of AI innovation. This shift has significant implications for the U.S.-China AI competition and underscores the need for stronger U.S. export controls.

China’s Inference Capacity is Key

Current U.S. export controls aim to restrict China’s ability to train frontier AI models but overlook the growing importance of inference and China’s capacity to scale it. As AI development shifts towards inference, China’s position strengthens considerably due to three key factors:

1. Steady access to inference-viable GPUs

2. Leading inference efficiency

3. Compute ecosystem being better suited for inference rather than pre-training

Access to Inference Hardware: The H20 Loophole

Despite U.S. export controls restricting access to cutting-edge AI chips like the H100 and H800, China maintains strong access to inference-capable hardware through several avenues — most notably through Nvidia's H20 GPU.

The H20 represents a significant gap in current export restrictions. Specifically designed to comply with export controls and serve the Chinese market, the H20 is actually superior to the H100 for particular inference workloads. The H20 outperforms the H100 for inference workloads due to its superior memory capacity and bandwidth. It delivers 20% higher peak tokens per second and 25% lower token-to-token latency at low batch sizes—key advantages given that inference performance is driven more by memory bandwidth and batch efficiency than by raw computational power. With 96GB of HBM3 memory and 4.0TB/s memory bandwidth, compared to the H100’s 80GB and 3.4TB/s, the H20 is highly viable for inference, making it a significant gap in current export restrictions.

China has been importing large sums of the H20. SemiAnalysis estimates that in 2024 alone, NVIDIA produced over 1 million H20s, most of which likely went to China. Additionally, orders by Chinese companies, including ByteDance and Tencent, for the H20 have spiked following DeepSeek’s model releases.

Access to Inference Hardware: Trailing-Edge GPUs

Trailing-edge GPUs remain surprisingly effective for inference workloads. China retains strong access to trailing-edge GPUs due to large stockpiles of the A100, A800, and H800 in 2022 and 2023. Additionally, Chinese firms, including Huawei, Alibaba and Biren, have also developed indigenous chips. The viability of trailing-edge GPUs for inference suggests that China’s inference capacity is stronger than their volumes of cutting-edge GPUs may suggest.

The effectiveness of older GPUs for inference stems from fundamental differences between inference and training workloads:

Long-Context Inference is Memory-Bound, Not Compute-Bound

Unlike training, inference only runs forward passes, avoiding computationally intensive processes like backpropagation and gradient updates. As a result, inference is significantly less compute-intensive than training.

The real constraint for inference is memory. Inference, particularly long context inference, is currently memory-bound rather than compute-bound due to several factors:

1. Model Weights & Key-Value (KV) Cache: For transformer-based models, inference requires storing both the model parameters and a key-value (KV) cache. The KV cache stores the past tokens' key-value pairs, allowing the model to retain context and coherence, and grows linearly with the context length. While compute resources are only required to process each newly generated token, memory usage continuously increases as new key-value pairs for each transformer layer are stored in the cache with every additional token generated. Consequently, total memory consumption rises steadily as the context expands, in contrast to compute needs, which remain stable and do not accumulate in the same manner. As a result, inference often becomes memory-constrained before it becomes compute-constrained, particularly for long-context tasks, where the KV cache can exceed the model weights in size.

2. Autoregressive Bottleneck: Input tokens can be processed in parallel, leveraging the full sequence since it’s known upfront. However, output tokens are generated sequentially, with each new token depending on all the previously generated tokens. This creates a bottleneck during output generation:

   1. Full KV Cache Access: Each generated output token requires accessing the entire KV cache.

   2. Memory Bandwidth Limitation: On long sequences, this repeated full KV cache access for every output token creates a memory bandwidth bottleneck (data transfer rate between memory and processor), which becomes the primary limiting factor.

   3. Constrained Batch Sizes: The size of the KV cache directly limits batch size during output generation. Longer sequences consume more GPU memory, reducing space for batching multiple sequences. This forces smaller batch sizes–the amount of independent user queries that can be processed in parallel–which reduces GPU utilization and restricts inference throughput.

This memory constraint becomes evident when examining FLOP utilization rates. During inference operations, GPUs typically achieve only about 10% FLOP utilization when generating tokens, compared to 30-50% during training. This underutilization occurs because GPUs spend much of their time retrieving and managing the KV cache rather than performing actual computations. The inefficiency grows even more pronounced with newer, more compute-dense chips, where increasingly powerful processing cores sit idle waiting for data to arrive from memory.

The Memory Wall

This inference bottleneck reflects a broader structural limitation in computing hardware. While GPU compute performance has grown exponentially (approximately 3.0x every 2 years), memory bandwidth and capacity have improved at a much slower rate (around 1.6x every 2 years). This growing gap creates a “memory wall” where performance is constrained not by processing speed but by how quickly and how much data the GPU can store and access.

This memory-bound nature of inference has significant implications for hardware viability. While newer GPUs offer exponential improvements in raw computational power (measured in FLOPs), they provide more limited gains in memory capacity and bandwidth — the true bottlenecks for inference workloads.

As a result, inference workloads often cannot fully utilize the computational resources available in cutting-edge GPUs. When memory bandwidth is the primary bottleneck rather than raw compute power, older GPUs remain surprisingly effective for inference tasks. The performance gap between newer and older GPU generations becomes much less significant than their computational performance might suggest.

These technical characteristics create a unique hardware dynamic that changes the calculus around AI chips. Trailing-edge GPUs retain viability in an inference-dominated landscape — a generation-old GPU might deliver 60-70% of current-generation inference performance, making it highly viable for most applications. This shifts the cost-effectiveness equation; dollar-for-dollar, older GPUs often provide better inference performance per unit cost than cutting-edge hardware optimized for training workloads. While trailing-edge GPUs quickly become obsolete for training, they remain viable for inference much longer.

Architectural Innovations and Shifting GPU Viability

A single architectural innovation can reshape which GPUs are viable for inference tasks. DeepSeek's Multi-Head Latent Attention (MLA) highlights this dynamic, reducing KV cache requirements by over 90% and fundamentally changing inference bottlenecks.

By shrinking KV cache memory demands, MLA shifts short and medium-context inference tasks from being memory-bound to increasingly compute-bound. Lower memory demands mean GPUs spend less time waiting for data retrieval and more time on actual computation, significantly increasing GPU utilization rates. For China's AI ecosystem, this unlocks substantially more inference throughput from trailing-edge GPUs.

Custom optimizations further amplify these benefits. DeepSeek has demonstrated that Huawei's domestically-produced Ascend 910C can achieve 60% of Nvidia's H100 inference performance through targeted optimizations. This showcases how software and architectural innovations continually reshape the viability and relative strengths of different GPUs for AI workloads.

MLA renders short- and medium-context inference tasks far more efficient by reducing memory bottlenecks, allowing cutting-edge GPUs to fully leverage their computational power. While this widens the performance gap between cutting-edge and trailing-edge GPUs, it also increases China’s overall inference capacity by making older hardware more efficient. Leading-edge GPUs like the H100 will continue to dominate compute-bound workloads, but MLA significantly boosts the total inference power that can be extracted from China’s existing GPU stockpile.

For long-context inference, the hardware calculus shifts again. When context length becomes sufficiently large, tasks remain memory-bound even with MLA, reducing the performance advantage of cutting-edge hardware over trailing-edge hardware for these specific workloads. Long-context inference tasks are particularly important for reasoning, agentic AI, and automated research applications. The capacity of trailing-edge hardware to support these AI capability-enhancing tasks strengthens China’s ability to advance AI progress despite hardware constraints on cutting-edge GPUs..

Implications for Export Controls

The implications for export controls are significant: inference capacity is growing across the board, and restrictions on cutting-edge hardware won’t prevent China's inference capacity from expanding. Cutting-edge GPUs will retain significant performance advantages for short and medium-context workloads, but trailing-edge hardware remains surprisingly effective for long-context inference where memory constraints persist.

The prolonged viability of trailing-edge GPUs for inference extends the lifespan of China's existing hardware stockpile. Even as export controls limit China’s access to cutting-edge AI accelerators, China’s large stock of A100, A800, and H800 GPUs remains useful for inference applications far longer than they would for training. This sustains China's AI infrastructure and boosts its inference capacity despite limits on acquiring new chips.

Moreover, China has developed indigenous AI chips capable of inference. Huawei's Ascend 910C has demonstrated competitive performance for inference workloads. Notably, the Ascend 910C’s yield rate has doubled since last year to 40%, and Huawei plans to produce 100,000 units of the 910C and 300,000 units of the 910B in 2025, signaling a significant expansion of domestic chip production. Biren Technology's BR100, a 7nm, 77-billion transistor GPU, rivals the A100 for both training and inference. China’s growing production of inference-viable chips, substantial stockpile of trailing-edge GPUs, and continued access to the H20 reinforce its ability to sustain AI capabilities in an inference-heavy AI paradigm despite restrictions on acquiring cutting-edge hardware.

The Hardware Multiplier: China’s Inference Efficiency

Beyond hardware access, China’s advances in inference efficiency have significant strategic implications for U.S. export controls. DeepSeek’s recent innovations — particularly its v3 and R1 models — demonstrate China’s ability to push the frontier of inference efficiency. By implementing innovative techniques like a sparse Mixture of Experts architecture, multi-head latent attention, and mixed precision weights, DeepSeek’s R1 model achieves approximately 27x lower inference costs than OpenAI’s o1 while maintaining competitive performance.

This efficiency advantage effectively counterbalances U.S. hardware restrictions. Even if export controls limit China to 15x less hardware capacity, a 30x inference efficiency advantage would enable China to run nearly twice as much inference as the U.S. This acts as a multiplier on China’s hardware base, potentially giving China greater total inference capacity despite hardware restrictions.

The efficiency gains extend the utility of trailing-edge GPUs in China’s AI ecosystem, as improved inference efficiency compensates for computational and memory limitations. While DeepSeek’s achievements are a continuation of the observed decline in inference costs, this case demonstrates that Chinese AI labs have already developed the expertise to push the frontier of inference efficiency and could choose to withhold future breakthroughs if strategic considerations change.

The Sleeping Dragon: China’s Compute Overcapacity

Additionally, China’s massive but fragmented compute ecosystem is structurally better aligned with inference requirements than training needs. The aggressive GPU stockpiling during China’s “Hundred Model War” of 2023 created substantial compute capacity that became underutilized as many firms abandoned their foundation model ambitions. As Alibaba Cloud researcher An Lin observed, many of China’s claimed “10,000-GPU clusters” are actually collections of disconnected GPUs distributed across different locations or models. While this fragmentation makes the infrastructure suboptimal for training frontier models, it remains viable for inference workloads that can run effectively on smaller, distributed clusters.

Open-source models are particularly well-positioned to leverage this distributed infrastructure, enabling deployment across China’s fragmented GPU ecosystem and transforming previously idle compute into a strategic asset for widespread inference. This approach allows companies to preserve limited high-quality compute for model development while unlocking latent compute capacity.

China’s once-idle compute resources are increasingly valuable in an inference-heavy AI landscape, improving China’s position along both the innovation and diffusion axes.

How Should the U.S. Respond?

An inference-heavy AI paradigm favors China’s AI innovation potential. Its access to inference-viable hardware, leading inference efficiency, and compute overcapacity function better in an inference-driven context than in a pre-training one. U.S. export controls, designed to constrain training, have been less effective at limiting inference. China’s inference capacity remains underestimated. Despite restrictions, access to trailing-edge GPUs, stockpiles, domestic chips, and H20s enable continued progress.

As inference becomes central to AI competition, China’s relative position strengthens, narrowing the U.S. advantage. This shift demands a strategic recalibration: the U.S. must reinforce export controls and develop a coherent open-source AI strategy.

Restricting Exports of the NVIDIA H20

Export controls on AI hardware operate with a lag — typically one to two years before their full impact materializes. This lag effect is central to understanding both current policy outcomes and future strategic decisions for export controls.

Some cite DeepSeek’s latest models as proof that U.S. export controls have failed. However, this outcome is a shortcoming in how the controls were initially calibrated rather than a failure of the broader strategy. The Biden administration initially set narrow thresholds—based on FLOPs and interconnect bandwidth — which NVIDIA circumvented with the H800, designed specifically to remain exportable to China. When controls finally expanded to include the H800 in October 2023, Chinese companies had already stockpiled these GPUs in addition to speculated H100s and H20s, allowing them to maintain frontier development and delaying the policy's actual impact.

This lag highlights how AI hardware and model lifecycles can stretch over many months, so chips purchased immediately before or soon after a policy shift can remain in service for a long time. Consequently, the policy’s full impact may not be evident right away. As older hardware loses its edge for training and frontier development scales, the impact of controls becomes realized through constraints on both the speed of a country’s AI advancement and the extent of its diffusion.

The lag effect of export restrictions is more pronounced for inference hardware. Unlike training, inference workloads can remain viable on older GPU generations for much longer periods, as they depend more on memory capacity and bandwidth than raw compute power. If the U.S. delays restricting inference-oriented chips like the H20 until inference becomes even more central to AI power, the extended lag could substantially weaken the effectiveness of export controls as a defensive measure. By restricting the H20 now, the U.S. can meaningfully limit China’s accumulation of inference hardware before inference becomes the dominant compute paradigm in AI. The sooner these revised controls take effect, the sooner they will impose measurable constraints on China’s ability to compete along both axes of AI competition.

A Strategy for Open-Source AI

Open-source AI is a key vector of competition that requires a strategic U.S. approach. While it fuels innovation, not all models or circumstances warrant taking the same open approach. Open-sourcing an advanced model represents a form of technology transfer to China if that model exceeds the AI capabilities that China has access to. This reduces the U.S. lead on the AI innovation axis, shifting competition toward the diffusion axis — an area where China may be better positioned to compete.

As the compute requirements for pre-training grow, open releases help China overcome its pre-training disadvantage while amplifying the role of inference, where China is stronger. If not managed strategically, open-source AI could accelerate China’s ability to close the gap in both innovation and diffusion. The U.S. must assess whether it retains an edge in leveraging open models for research, application, and deployment. If so, open-source strategies can reinforce leadership; if not, they risk eroding it.

To assess the impact of an open release on U.S. tech competitiveness, we should evaluate how much of an immediate advantage the U.S. is foregoing on the AI innovation axis by open-sourcing a model and compare that to the net effect of how well the U.S. and China can convert open access into gains across both axes. If the U.S. retains a structural advantage in furthering AI research, building applications, fine-tuning, and scaling AI deployment, then open-source strategies can reinforce U.S. leadership. However, if China is more effective at leveraging open models for research, real-world adoption and economic or military applications, then unrestricted open release could benefit China more. This dynamic underscores the need for a structured approach and collaboration between private and public sector regarding deployment decisions.

The Bottom Line

As trends in AI elevate the importance of inference, the U.S. must reassess its strategy to lead along both axes of AI competition. While early export controls are designed to constrain China’s ability to train frontier models, they are less effective in limiting its capacity for large-scale inference. To sustain its competitive edge, the U.S. must expand export controls to address the growing role of inference, particularly by restricting chips like the NVIDIA H20 before their strategic importance escalates further. At the same time, the U.S. must refine its approach to open-source AI, ensuring that its diffusion benefits reinforce, rather than undermine, U.S. national AI leadership. Winning the AI competition requires adapting as fast as the technology evolves, and this is a critical moment for the U.S. to recalibrate its strategy.

It’s been a tough week for the international order. It feels like every TV in every restaurant across Taiwan is blasting nonstop coverage of the Trump-Zelenskyy fallout.

How will Taiwan respond to Trump’s pivot to Putin? Would Taiwan be safer with nuclear weapons? What platforms do Taiwanese people use to debate about politics anyway?

In today’s roundup, we’ll analyze perspectives from Taiwanese legacy newspapers, social media firestorms, and viral political influencers.

Driving Solidarity

We’ll start off by highlighting some reactions on the most popular Taiwanese social media platform, PTT.

PTT is a bit like a Taiwanese version of Reddit. The key difference is that comments are always displayed in chronological order instead of being ranked by popularity. Users can “push” 推, “boo” 噓, or reply to comments to express their opinion. The platform shows whether each comment is being “pushed” or “booed” overall, but doesn’t display the total vote tallies. Like on Reddit, there are sub-forums for topic-specific discussion.

Disclaimer: these forums are hosting open debates with intense back-and-forth between commenters. I’ll be highlighting recurring themes, as well as arguments where both sides are earning push-votes, but I want to be clear that there is no broad consensus on what the Trump-Zelenskyy fallout means for Taiwan at this point.

For example, the following debate emerged below a Mandarin translation of the Foreign Affairs article entitled “Ukraine Will Not Surrender to Russia”:

(Pushed) I support and praise this article, justice will prevail.

(Booed) Then how come there are no soldiers? Conduct an opinion poll or something.

(Pushed) These past couple of days, I've seen quite a few people claim that Ukraine should have originally surrendered to Russia in exchange for peace and prosperity. This kind of argument completely ignores the suffering Ukraine endured under Russian rule in the past.

(Pushed) In the past, we thought that people in democratic countries feared death more than other people — but Ukrainians are not afraid.

(Pushed) The Uyghurs will never surrender, but they will not go to the front line

(Pushed) It is 100000000% reasonable to be suspicious that Trump received personal benefits from Russia or made a blood pact with Russia.

The Taiwanese transliteration of “Zelenskyy” is 澤倫斯基 Zélúnsījī, and in casual writing Taiwanese people refer to him by the nickname 司機 Sījī (literally, “The Driver”) which has the same pronunciation as the last two characters of the transliteration.

From a thread in a military forum about whether Zelenskyy overplayed his hand:

(Pushed) The driver really shouldn’t have talked back to Vance. If he wanted to argue, he could have done it in private.

(Pushed) After apologizing, you still have nothing, so why bother apologizing?

(Pushed) If Little Z doesn’t kneel, America will make explosive corruption accusations against him.

(Reply) East Asian countries are better at licking.

(Pushed) If Ukraine wants to thank someone, it should thank the previous Biden administration. Why thank Trump?

(Pushed) It seems someone is trying to smear and destroy Mr. Z's image. Be careful when responding to this thread.

Indeed, there are signs of disinformation in some discussions of this topic. An FT article entitled “Zelenskyy rejects calls for immediate Ukraine-Russia ceasefire” was posted on PTT with the mistranslated title, “The Driver Rejects Ukrainian and Russian calls for a Ceasefire” (司機拒絕烏克蘭與俄羅斯立即停火的要求), a fact which was quickly pointed out and mocked in the comments.

Marco Rubio is well-known in Taiwan thanks to his long congressional record of support for the island. Here are some comments about him:

(Pushed) Rubio will be replaced soon.

(Pushed) Rubio was once a pioneer in anti-communism, but now he bows down to power.

Underneath an article reporting Trump’s plan to freeze aid to Ukraine in response to the meeting:

(Pushed) Stop it right now immediately!!!!!!!!!!!!!!! I’ve never seen such a cowardly U.S. president!! You truly see everything if you live long enough!!!!!!!!

(Pushed) Will the European big brothers shoulder some of the responsibility? Isn’t this an opportunity for them to show off?

In a financial forum:

(Pushed) Being pro-China is selling out Taiwan, being pro-America is also selling out Taiwan.

(Pushed) In the Budapest Agreement, even China said it would protect Ukraine, but that isn’t happening

(Pushed) Ultimately, [Ukraine] should not have given up its nuclear weapons. Security guarantees are bullshit.

(Booed) Ukraine has no nuclear bombs, so of course it has no bargaining chips.

(Pushed) The driver’s bargaining chip is making the king (Trump) lose face.

(Pushed) Buddha’s mercy 佛祖慈悲 [This phrase is used ironically in situations that are cruel or corrupt to the point of hopelessness.]

Ukraine Today, But Taiwan’s OK?

At the start of the invasion, the DPP popularized the slogan, “Ukraine today, Taiwan tomorrow.” Editor Gu Shu-ren 辜樹仁 of CommonWealth Magazine 天下雜誌 (a Taiwanese publication similar to the Atlantic), addressed fears that Trump will abandon Taiwan after Ukraine in a recent editorial:

Looking back at history, Taiwan's strategic value to the United States has been the key factor in America's decision to either abandon or support Taiwan.

In 1950, when the Korean War broke out, the Republic of China (ROC) government, which had retreated to Taiwan and was on the brink of collapse after being abandoned by the U.S., suddenly became the central hub of the U.S. first island chain strategy in East Asia — a so-called unsinkable aircraft carrier — greatly increasing Taiwan's strategic importance.

In the 1970s, as the U.S. aligned with China to counter the Soviet Union, Taiwan lost its strategic value, leading to the severance of U.S.-Taiwan diplomatic ties and the withdrawal of U.S. troops from Taiwan. …

Today, Taiwan's strategic value to the United States is at its highest since the servering of diplomatic ties, as the primary battleground in the U.S.-China rivalry is now the technology war, with semiconductors at its core. More specifically, TSMC is the most crucial asset for the U.S. in securing a supply of advanced chips and revitalizing its semiconductor manufacturing industry. If the U.S. wants to maintain its technological and military lead over China, it must firmly keep Taiwan within its grasp. …

Ensuring that the U.S. remains dependent on Taiwan’s advanced chip manufacturing — making American national security synonymous with protecting Taiwan — is the most critical factor in maintaining Taiwan’s strategic value to the United States.

Of course, there is another equally important factor. Trump dislikes war, especially costly military interventions where the U.S. cannot be assured of victory. He has repeatedly complained that Ukraine failed to prevent war at the outset. Therefore, avoiding war at all costs is also a key strategy for Taiwan to secure Trump’s support.

Only through this can tomorrow’s Taiwan avoid becoming the Ukraine we saw today.

Reporter Jiang Liangcheng 江良誠 similarly warned that Taiwan would need to become more transactional in its relationship Trump:

“Trump's only vocabulary is actually "money, money, money". All international relations can be measured by money. There is no free lunch in the world. It is impossible to ask Americans to help you defend your country like a plate for free and without any reward. …

However, when it comes to Taiwan's policy toward the United States, Lai Ching-te still sticks to Tsai Ing-wen's international politics, such as the first island chain, geopolitics, and Indo-Pacific security. I'm afraid even Trump doesn't understand these terms.”

The Meihua News Network (梅花新聞網), a Pro-China news outlet owned by a controversial Taiwanese religious leader, argued instead that Taiwan needs to reopen dialogue with Beijing given the reality that the U.S. is an unreliable partner.”

In front of cabinet members and the media, Trump was unwilling to guarantee that the Chinese Communist Party would not invade Taiwan by force during his term, and emphasized that he had a good relationship with Chinese Communist Party leader Xi Jinping. …

“Foreign Affairs” recently published a special article titled “The Taiwan Fixation: American Strategy Shouldn’t Hinge on an Unwinnable War”, co-authored by Professor Kavanagh of the Georgetown University Center for Security Studies and senior scholar Wertheim of the Carnegie Endowment for International Peace. The gist of the article is: Taiwan is certainly valuable to the United States, but if American decision-makers overestimate Taiwan's importance, they will sacrifice the security of maintaining the status quo due to the risk of endless and destructive war; and Taiwan's importance is not enough for the United States to sacrifice tens of thousands of American lives to protect it. Former National Security Council Secretary-General Su Chi 蘇起 described this article as the most powerful article to date advocating the United States to let go of Taiwan. …

Apart from fully relying on the American security umbrella and turning Taiwan into a "porcupine," the DPP also has another option: restoring cross-strait communication and reducing tensions in the Taiwan Strait. If that happens, the so-called "Abandon Taiwan Theory" would naturally dissipate. Rational decision-making should not be obstructed by anti-China or China-hating sentiments.”

By contrast, a popular post from the Taiwanese political influencer James Hsieh argued that Taiwan should be doing whatever it takes to improve relations with the U.S., not criticizing Trump’s Ukraine policy:

“I still see many people online going against the tide, bashing Trump, criticizing the U.S., and supporting all kinds of conspiracy theories.
Here are five reminders:

1. Before the war, Ukraine was extremely pro-China, selling major military technology to China. Just a few days ago, Ukraine even asked China for help.

2. Morally, we must oppose aggression, but in terms of international strategy, we must firmly support the United States.

3. Taiwan is not Ukraine. In terms of historical ties with the U.S., the Taiwan Relations Act, geographical location, type of warfare, and economic strength, Taiwan is completely different. Taiwan is absolutely not a distant European country like Ukraine in America's eyes. Comparing Ukraine to Taiwan is a completely flawed analogy. Saying that the U.S. pulling out of the Russia-Ukraine war implies that it will betray Taiwan is just another favorite conspiracy theory of the dumb lefties (左膠) and the Chinese Communist Party’s propaganda machine.

4. Personally, I hope the Russia-Ukraine war ends quickly so that the U.S. can fully prepare for the Indo-Pacific. This is a practical concern, as China is rapidly advancing its strategic plans. How the U.S. swiftly ends its engagements elsewhere and refocuses on the Indo-Pacific is critical. Just yesterday, Vice President Vance stated that the U.S. military-industrial production can no longer sustain the continuous supply of heavy weaponry to Ukraine.

5. History has shown that during major wars, opportunistic nations take advantage of a great power’s exhaustion to invade smaller neighboring countries. If the Russia-Ukraine war escalates into World War III and the U.S. and Europe are preoccupied with fighting Putin’s alliance, it would be the perfect moment for China to seize Taiwan under the guise of maintaining stability.

If Taiwan's democracy, freedom, and independence from oppression are what you value most, then Taiwan should prioritize its relationships with the U.S. and Japan over everything else — not Ukraine.

Only the U.S. and Japan will help us. Survival comes first before ideals.

Taiwan-U.S. friendship!”

It remains unclear what the Lai administration’s approach will be, but you can be sure that ChinaTalk will keep monitoring the debate as it evolves.

Subscribe now

Will Taiwan Get Nukes?

Zelenskyy’s White House press conference also reignited the old debate about whether Taiwan would benefit from having its own nuclear arsenal. Taiwan abandoned its indigenous nuclear program in response to pressure from the U.S., much like how Ukraine relinquished its nuclear weapons to Russia after the fall of the USSR. Taiwan was estimated to be just two years away from completing a WMD when the U.S. intervened in 1988.

These parallels were drawn explicitly by a CNN profile of Colonel Chang Hsien-yi 張憲義, the Taiwanese nuclear engineer who provided intelligence about Taiwan’s proliferation plans to the CIA. The article was repackaged, translated, and published on the front page of the China Times on Monday.

On PTT, the profile drew comments like:

(Pushed) This person is the reason why Taiwanese independence is impossible.

(Pushed) Nuclear weapons are not something that Taiwan's extremely incompetent politics could handle. If nuclear weapons were in the hands of Chiang Kai-Shek and his family, Taiwan would have ended up like North Korea. The Chiang family would still in power, and there would never have even been a chance for democratization. So many people have no clue what’s going on.

Taiwanese political influencer Mr. Shen 公子沈, who runs a YouTube channel with more than 700k subscribers, posted the following meme on Threads (which is way more popular in Taiwan than the U.S.) with the caption, “With nukes vs without nukes: it’s time for Taiwan to develop nuclear weapons.”

Speaking of bargaining chips…

Reactions to the TSMC Deal

TSMC’s newly announced $100 billion investment in US chip manufacturing led to more online discontent. The following comments from Facebook were curated by Angela Oung:

“Today we are all Ukrainians”

“At least Zelensky has guts”

“ASMC” [American Semiconductor Manufacturing Company]

“So they’re taking our stuff, leaving us with no cards. Think they’ll help in the future? Stop dreaming!”

“Taiwan’s remaining value is becoming a meat grinder like Ukraine.”

“He [TSMC Chairman CC Wei] looks like he has a gun behind his head. Hostage situation.”

“The silicon shield we spent decades building is being handed over by our government without a whimper”

“TSMC: built by the KMT, sold by the DPP”

“Is Lai Ching-te such a pussy that he’s not even gonna say anything?”

“Today Ukraine, tomorrow Taiwan. One step closer to refugee status.”

“Bandits…just like the CCP”

To close, I’ll leave you with another popular post on Threads expressing frustration about Taiwan’s-U.S. relations:

“The U.S. asks us to buy military equipment — we buy it.

The U.S. asks us to extend the length of mandatory military service — we extend it.

The U.S. wants TSMC — we hand it over with both hands.

The U.S. wants us to implement resilient defense — we manage to do it, even if we have to hide and shuffle the budget.

For every single thing the U.S. asks of us, from the issue of eating ractopamine pork in our daily meals to national defense policies involving regional security cooperation, Taiwan follows the U.S.’s demands without question.

But will there come a day, just like today’s Ukraine, where we sign agreements on resource concessions, trading away our country's future rebuilding assets, yet still lack the most basic “security guarantees”?

Ukraine has the support of the entire European continent—but what about Taiwan?

Will today’s Ukraine be a reflection of Taiwan’s future?

Will Taiwan, when that day comes, be even more isolated and helpless?”

To be fair, this commenter is right that Taiwanese pork is way more delicious than the ractopamine pork imported from the U.S. I sincerely hope that every ChinaTalk subscriber has an opportunity to come to Taiwan and eat stewed pork rice (滷肉飯)…before it’s too late!?

A guest piece by Afra, freelance writer and podcaster [Jordan: I highly recommend this show!] with working experience in tech and crypto. Personal site here.

DeepSeek’s winds have already been blowing for some time, but this particular gale seems to have real staying power.

President Trump characterized DeepSeek as “a wake-up call,” Scale AI CEO Alexandr Wang called it “earth-shattering,” and Anthropic CEO Dario Amodei deigned to come on ChinaTalk to discuss fears of powerful AI within “authoritarian systems of government.”

On Chinese social media, the discussions took on a life of their own, with the most popular use case being the calculation of one’s Ba Zi (八字) and astrological chart, using the social media tag “AI玄学” (AI Mysticism). Users weren’t just seeking their personal fortunes — they saw the nation’s destiny itself shifting through DeepSeek’s emergence. These conversations are a swirling mix of collective jubilation, national pride, and gleeful satisfaction over America’s “China envy,”1 often accompanied by playful banter.

Yet amidst this discourse, a deeper and more resonant question emerges: could this be a sign of China’s technological ascension? Is this evidence that Guoyun (国运) — the nation’s long-awaited destiny — has finally arrived?

First, what is Guoyun 国运？

The term 国运 combines two characters: 国 (guó, “nation/state”) and 运 (yùn, “fate/destiny/fortune”). This concept emerged from traditional Chinese cosmological thinking, where the destiny of the state was seen as intertwined with celestial patterns and dynastic cycles.2 This term, once confined to the ornate dialogue of period dramas set in imperial China, has begun to surface with increasing frequency on my social media timeline.

The Guoyun narrative around DeepSeek began when Feng Ji 冯骥, creator of the globally successful game “Black Myth: Wukong,” declared it a “national destiny-level technological achievement.”

The discourse gained momentum when Zhou Hongyi 周鸿祎, Chairperson of Qihoo 360, positioned DeepSeek as a key player in China’s “AI Avengers Team” against U.S. dominance. This sentiment echoed across media, with headlines like “Is DeepSeek a breakthrough of national destiny? The picture could be bigger” and “DeepSeek triggers U.S. stock plunge; can it really change the nation’s destiny?"

For Chinese netizens, discussions about politics on social media are often marked by subtlety and veneration with trepidation (for reasons that require little explanation). However, during the 2025 Chinese New Year, the discourse expanded far beyond politics and DeepSeek into a cacophony of cultural euphoria —a wave of self-congratulatory enthusiasm that evolved into something larger culturally. This included the movie Nezha 2, which shattered box office records and surpassed Inside Out 2 to become the highest-grossing animated film of all time (with patriotism-fueled consumption boosting the box office performance), TikTok refugees flooding Xiaohongshu, and advanced Unitree robotics performing during the Spring Festival Gala. These achievements seemed to occur against a historical backdrop where technological and cultural advances carry deeper significance about China’s rightful place in the cosmic order.3

The Guoyun discourse extends beyond tech leaders, media commentary, and social media posts.

President Xi Jinping has woven the concept of destiny into official rhetoric, though carefully stripped of its more superstitious elements. Speaking at the 19th Academician Conference of the Chinese Academy of Sciences in May 2018, Xi declared, “Innovation determines the future; reform concerns national destiny. The field of science and technology is the area most in need of continuous reform 创新决胜未来，改革关乎国运。科技领域是最需要不断改革的领域.” This statement aligns with his broader techno-nationalist vision, explicitly linking technological advancement to China’s strategic future.

A 2024 People’s Daily article discussing Xi’s thoughts emphasized that “cultural confidence is a major issue concerning national destiny 坚定文化自信，是一个事关国运兴衰...的大问题"。

This rhetorical shift signals a carefully calibrated blend of traditional Chinese concepts with modern governance — a bridge between ancient ideas of dynastic cycles and contemporary aspirations for technological supremacy.

Beyond superstition: is this a collective myth-making or post-pandemic yearning for certainty?

It would be a mistake to dismiss this discourse as mere superstition or propaganda.

The COVID-19 pandemic marked a watershed moment in Chinese society’s relationship with national destiny. To me, Zero COVID became a mirror polished to cruel clarity, reflecting a China I no longer recognized. During the rigid cycles of lockdowns and reopenings, I didn’t see my parents for two years, my grandmother was hospitalized, and my cousin was confined to his university dorm for three whole months culminating in a severe mental breakdown. Friends lost loved ones due to a lack of timely treatment options. Back then, seeing how waves of people wanted to “run (润)” from China, I thought for the first time that I might never return to China, and that I might become part of the Chinese diaspora forever.

COVID created a collective trauma that many Chinese are still processing.

But this experience has paradoxically reinforced a certain earnest faith in China’s future among ordinary citizens. The optimism in the discussion of Guoyun might represent a complex emotional response to the uncertainty and trauma from the COVID era — a blend of traditional fatalism with genuine aspirations. Having weathered the pandemic’s disruption, many ordinary Chinese seek reassurance about the future through familiar cultural frameworks. ‘National Destiny’ provides exactly that — it’s a narrative that contextualizes current struggles within a larger, ultimately triumphant story. It’s therapeutic.

The discourse around 国运论 (guóyùn lùn, or “national destiny theory”) reveals parallels to America’s historical myth-making. Perhaps the most striking similarity between China and the US is their unwavering belief in their own exceptionalism and their destined special place in the world order. While America has Manifest Destiny and the Frontier Thesis, China’s “national rejuvenation” serves as its own foundational myth from which people can derive self-confidence. Through countless repetitions across state and social media, this narrative has become deeply ingrained in China’s national consciousness.

The wounds behind techno-nationalism

Where myths nurture the national consciousness, technology has become the battleground where China’s historical narrative demands its vindication. The roots of China’s techno-nationalism run deep, drawing emotional power from China’s “century of humiliation.” U.S. actions — chip controls, the attempted TikTok ban, tariffs, investigations of Chinese scientists, and suspicions of Chinese espionage — rekindle the historical trauma of humiliation.

For decades, China has been portrayed as a mere copycat or thief of Western innovation. Each technological breakthrough now serves as vindication, a refutation of that dismissive narrative — this shame has never truly been resolved. As Kevin Xu elaborated on DeepSeek’s open-sourced nature, “It’s all for the validation and approval,” — a sharp acknowledgment that when Chinese engineers share their code with the world, they’re not just demonstrating technical prowess but seeking to heal a wound in the national psyche:

In the Chinese open source community, there is this thing that I would call open source “zeal” or “calling” (开源情怀)

Most engineers are thrilled if their open source projects — a database, a container registry, etc-- are used by a foreign company, especially a silicon valley one. They’d tack on free labor on top of already free software, to fix bugs, resolve issues, all day all night. It’s all for the validation and approval.

Implicit in this “zeal” or “calling” is an acute awareness that no one in the West respects what they do because everything in China is stolen or created by cheating. They are also aware that Chinese firms have been taking for free lots of open source tech to advance, but they want to create their own, contribute, and prove that their tech is good enough to be taken for free by foreign firms -- some nationalism, some engineering pride.

So if you want to really understand why DeepSeek does what it does and open source everything, start there. It’s not a political statement, not to troll Stargate or Trump inauguration, or to help their quant fund’s shorts on NVDA (though if that were the case, it’d be quite brilliant and savage)

The drive to prove oneself on behalf of the nation is expressed vividly in Chinese popular culture. I couldn’t stop thinking about Illumine Linga (临高启明), an open-source collaborative novel that has captivated China’s engineering community and become a phenomenon of its own. The story follows modern Chinese engineers who time-travel to the declining Ming dynasty, right before China was conquered by the Manchus, bringing industrial equipment and technical knowledge. They gradually industrialize Hainan and Guangdong provinces before expanding outward with the ultimate goal of establishing global hegemony.4

Though ostensibly just fiction, Illumine Linga pulses with the heartbeat of China’s “Industrial Party” (工业党) — that loose constellation of engineers, programmers, and technically-minded patriots united by an almost religious faith in technology as destiny’s instrument. The novel serves as a sharp allegory for contemporary aspirations: technological mastery as the path to national resurrection and global respect.5

In the Western intellectual tradition, technology and data have undergone phases of detached scrutiny — viewed first as tools of emancipation, and later as vectors of control. Foucault’s panopticon mutated into Zuboff’s surveillance capitalism; Wiener’s Cybernetics birthed both Silicon Valley and Snowden’s disclosures. This academic back-and-forth assumes a fundamental premise: technology can theoretically exist as a neutral substrate awaiting ideological imprint.

However, in my impression, China’s techno-discourse never evinces such “purity.”

From its inception, technology has been semantically encased in the shell of techno-nationalism. In China’s history textbooks, Qian Xuesen’s missiles for the Two Bombs, One Satellite program were never just missiles, but brushstrokes in the narrative of “standing up again.”6 Yuan Longping’s hybrid rice strains didn’t merely feed millions; they were genetic correctives to the “Century of Humiliation,” each harvest a quiet refutation of the colonial-era belief that China couldn’t innovate.

On Chinese New Year’s Eve, a fake response to the “national destiny theory” attributed to Liang Wenfeng circulated widely online, with many believing and sharing it as authentic. This response claimed that DeepSeek’s open-source decision was merely “standing on the shoulders of giants, adding a few more screws to the edifice of China’s large language models,” and that the true national destiny resided in “a group of stubborn fools using code as bricks and algorithms as steel, building bridges to the future.” This fake statement—notably devoid of wolf warrior rhetoric—spread virally, its humility and relentless spirit embodying some values people hoped Chinese technologists would champion. Meanwhile, the real Liang Wenfeng remained silent after DeepSeek’s rise. A month later, he appeared on CCTV sitting beside Tencent’s Ma Huateng at Xi Jinping’s symposium for top business leaders.

The public’s fascination with Liang showed no signs of waning. In Silicon Valley, his previous interviews were swiftly translated into English and meticulously analyzed, while in China, his rise also inspired mystical interpretations—during the Spring Festival holiday, Liang Wenfeng’s ancestral home in Zhanjiang, Guangdong transformed into an impromptu tourist attraction, drawing feng shui masters eager to study the geomantic properties of his family residence.

Humans have always sought ways to calculate the incalculable. Perhaps that’s what makes the conversation around Guoyun so captivating: it’s not just about predicting the future, but about sense-making in China’s present.

Gary Wang spent the past decade developing business and product strategy for Silicon Valley technology companies, with a focus on enterprise software, the industrial internet of things and AI. He has a degree from HKS and worked in China. The views expressed here represent only his own.

About a decade ago, the best forecasts heralded a promising manufacturing future, in the United States and globally, with the advent of the fourth industrial revolution (also called “industry 4.0,” the “industrial internet,” or “industrial internet of things” aka IIoT). The belief was that the falling cost of cloud computing, sensor costs, and machine learning — coupled with new connectivity technologies such as 5G or IPv6 — would lead to a revolution in manufacturing productivity and ultimately higher GDP growth.

Despite these promising forecasts, multiple data points indicate that US manufacturing has largely stagnated. Analysis from the New York Federal Reserve reveals that both total factor productivity and labor productivity have been flat from 2007 to 2022. Meanwhile, US share of global manufacturing value add fell from nearly 25% in 2000 to an estimated 15% today in 2024. The UN Industrial Development Org projects US share of global manufacturing value add will fall to 11% in 2030, while China may account for 45% of global output.

This decline comes after multiple presidential administrations’ efforts to revitalize American manufacturing — from the Obama-era policies such as the Advanced Manufacturing Partnership or the Manufacturing USA initiative, to the Biden administration’s Inflation Reduction Act, and now the Trump administration’s desire to reshore manufacturing via tariffs and other policy tools.

Off-shoring and free-trade agreements go only so far in explaining this decline. And the present debates over US industrial policy — sparked by the advent of emerging technologies (generative AI, quantum computing) as well as intensifying competition with China — perhaps focus on the wrong things.

The real questions US policymakers must grapple with: why did the United States fail to capitalize on technology that was already available to make its manufacturing base more competitive?

Put another way: why have the promises of the IIoT revolution failed to materialize in the United States?

This piece makes a few key arguments:

• The “industrial internet of things” is not an industry. It’s a set of disparate technologies that all need to be adopted together to create value.

• The free market will not always optimize adopting a broad set of technologies for an entire ecosystem of industries. The underwhelming results of today’s industrial internet is a case in point.

• China’s industrial policies to “win” the fourth industrial revolution offer lessons for policymakers in the United States to consider.

• When it comes to revitalizing manufacturing, or ensuring American leadership in AI or quantum computing, policymakers need to craft policies to develop entire value chains and tech ecosystems — not myopically focus on just one strategic technology (eg. advanced semiconductors).

  ChinaTalk is a reader-supported publication. To receive new posts and support my work, consider becoming a free or paid subscriber.

What is IIoT?

IIoT refers to the interconnection of machines, devices, sensors, and systems which are connected on the internet in performing industrial tasks.

Take one of IIoT’s leading “use cases” (ie. applying tech to solve a business problem): predictive maintenance. Sensors connected to a piece of factory equipment, such as a boiler, can measure temperature or vibration. When combined with machine-learning algorithms, manufacturers stand to save millions by predicting when a machine would fail, and then proactively maintaining the machine before a failure occurs — thus reducing factory downtime and increasing productivity.

Another use case: gathering GPS data from truckers could enable machine-learning algorithms to optimize the routes of commercial trucks (saving fuel costs). When paired with data on customer demand (say, Pepsi sales in a city), manufacturers could save billions by optimizing their inventory costs to ensure that the optimal amount of Pepsi reached store shelves at just the right time.

The use cases are endless: deploying robots on the production line, using cameras and AI to automate quality inspection for finished products, creating a “digital twin” of an entire production process for optimization, and much more. All of these use cases required cloud computing, real-world historical data, and connectivity. As a practitioner who has worked with technology companies on their strategy for delivering industrial IoT to manufacturing companies, I can attest to the level of industry enthusiasm for IIoT during this time (as well as the numerous operational challenges).

How off were the IIoT forecasts?

In 2015, McKinsey forecast $1.2 to 3.7 trillion in economic value created per year by 2025 from IoT technologies in factories. Assuming technology vendors alone capture 5% of the value created — a very conservative benchmark — that’s $60 to $185 billion in revenue. The International Data Corp in 2017 forecast that manufacturers would spend $102 billion in the industrial internet, meaning vendors selling IIoT technologies should see comparable revenue figures. Accenture and World Economic Forum joined the hype, intoning that the “Industrial Internet will transform many industries, including manufacturing, oil and gas, agriculture, mining, transportation and healthcare. Collectively, these account for nearly two-thirds of the world economy.” These market forecasts led the Congressional Research Service in 2015 to predict, “The current global IoT market has been valued at about $2 trillion, with estimates of its predicted value over the next 5 to 10 years varying from $4 trillion to $11 trillion.”

These forecasts were off by multiple orders of magnitude. Today, to my knowledge, there is only one publicly listed company in the United States solely focused on IIoT: Samsara, with $1.4 billion in revenue, growing at a healthy ~40% year over year. (Palantir in 2024 reported $700 million in revenue from US private-sector firms, some of which include manufacturing — but the majority of Palantir’s business is with governments.)

General Electric and Siemens both tried to become technology companies by developing their own cloud platforms and AI applications to digitize the manufacturing sector. A series of New York Times headlines, though, tells the saga of GE’s attempt to capture the purported massive opportunity of the industrial internet of things:

• 2011: GE Wants Software for the ‘Industrial Internet’

• 2012: Looking to Industry for the Next Digital Disruption

• 2013: General Electric Adds to Its ‘Industrial Internet’

• 2015: GE Navigates, Carefully, the Industrial Internet of Things

• 2016:  GE, the 124-Year-Old Software Start-Up

• 2018: GE Makes a Sharp ‘Pivot’ on Digital

And finally, later in 2018:

Siemens positioned its industrial internet cloud platform, Mindsphere, as its next growth vector. Today, Mindsphere has been rebranded to “IoT insights hub,” and the last time Siemens company leadership talked about Mindsphere on their earnings call with equity analysts was in 2022, indicating a retrenchment in expectations (unlike when they spoke about Mindsphere on earnings calls with analysts in 2015,  2016, 2019, and 2020; what industry leaders tell Wall Street indicates where they think their companies’ growth will come from).

Why were the predictions so wrong?

IIoT is a cluster of disparate technologies that have to work together to create value. It’s not one technology. Consider the aforementioned predictive maintenance use case. To realize value, a factory owner needs to adopt six or seven different technologies from different vendors.

• There’s the company providing sensors (sometimes with software) for the machines to gather data for analytics.

• Many factories have historically not been connected to the internet, so a company like Verizon needs to get involved to set up an in-plant 5G connectivity network. (Leading analysts have estimated there are only a handful of 5G industrial projects in the United States, compared to likely thousands in China.)

• A company like Cisco has to provide the networking equipment to enable internet connectivity in the factory.

• A cybersecurity company needs to ensure the sensors and machines, now that they’re connected to the internet, are secure from cyberattacks.

• A cloud-computing company, such as Microsoft or Amazon, needs to provide the compute and storage for the customer to develop AI algorithms to analyze the data generated by the sensors. These cloud-computing companies often provide the AI algorithms for customers to customize themselves (assuming they have the in-house data science talent) to analyze the data from factory equipment.

• A company needs to integrate these disparate systems together — usually a system integrator like Accenture or Wipro.

The factory owner has a finite budget, must negotiate with six different vendors (each with their own pricing and profit models, none of whom necessarily coordinate their selling activities) — but still must realize a high enough return on investment (ROI) to justify solving this one use case. Imagine a consumer buying a car — but instead of buying from an OEM like Tesla or General Motors, you have to negotiate individually with the tire company, the engine manufacturer, the seat belt maker, the company making the infotainment display, and every other component manufacturer.

The nature of the physical world makes this coordination problem even more complex:

• Algorithms aren’t immune from false positives. What happens if the algorithms incorrectly predict a machine will break down, but a maintenance technician has already been dispatched to make repairs? That reduces ROI.

• Machine algorithms need to be trained on historical data of when the machine has broken down before — but for many factories, maintenance records aren’t digitized; if available at all, they’re paper logs of when a technician fixed a machine.

• Third, from the perspective of the technology vendor, sales cycles to manufacturers often are usually one to two years — since customers will pilot the technology for one set of machines (one use case) in one factory, measure the cost or productivity savings, and then decide whether they want to scale the technologies to multiple use cases across multiple factories. Factory budgets are managed locally, not globally — meaning a vendor has to sell to a manufacturer’s factory site in, say, the United States, then Brazil, then Germany, and so on.

All of these factors help to explain why venture capitalists — with few exceptions — have not invested in startups tackling industrial IoT, as well as why it’s been hard for existing vendors to scale their business. Even McKinsey admitted in 2021, “To date, value capture across settings has generally been on the low end of the ranges of our estimates from 2015, resulting from slower adoption and impact. For example, in factories, we attribute the slower growth to delayed technological adoption because many companies are stuck in the pilot phase.”

What has China done?

While IIoT hasn’t lived up to its potential in the United States and elsewhere in the West, China has leaped ahead in the fourth industrial revolution: there is no other country in the world that can boast of legions of “dark factories” — ie. factories where entire manufacturing processes are automated.

How has China done it? By focusing on technical challenges and market-coordination problems.

First: Chinese policymakers at the highest level — eg. the State Council — crafted policies to solve known technical challenges which threatened to hold back Chinese manufacturer’s adoption of IIoT technologies.

For example, in the predictive maintenance use case, there is a known problem of “asset mapping” — ensuring all the physical and digital assets in a factory can be identified in a common taxonomy to enable machine-learning analytics and then workflow automation (sending a technician to repair a robot, changing the workload of robots working together if one robot is breaking down, etc.). Specifically, if factory owners want to predict when a robot arm will break down, they need a comprehensive way to uniquely identify the specific robot, the specific arm of that robot, the specific sensor that may be attached to the robot, the specific 3D model of the robot’s arm, and then map all of these physical and digital assets together. Without a common taxonomy, it’s impossible to automate the analysis of sensor readings from the robot arm (eg. its grip strength) and then trigger a workflow to fix the robot arm while enabling the production process to continue seamlessly, that is, in a “lights out” fashion.

China’s State Council, in a 2017 planning document — “Guidance for Deepening the Development of the Industrial Internet ‘internet + promoting manufacturing” 深化“互联网+先进制造业” 发展工业互联网的指导意见 — specifically called for implementing networking connectivity and “identity resolution system” 标识解析体系 to solve this problem, using a combination of known technologies and standards such as IPv6, software-defined networking, 5G connectivity, time-sensitive networking, and passive optical networking. The technologies mentioned in this document were available in China (and the United States) in 2017. An identity resolution system (the English equivalent term would be a digital “tracking system”), when combined with advanced networking technologies, solves this predictive-maintenance problem because then a piece of software — such as a predictive-maintenance application for robots — can automatically locate the robot arm that’s emitting sensor data indicating a breakdown, match that to the 3D model that specifies how the robot arm should function, detect issues with the robot arm, and then trigger a workflow to remediate. Dozens of physical and digital systems are involved in solving this problem.

Of course, the free market can solve this problem as well — but it runs into the same issue mentioned above: coordination of multiple vendors with multiple technologies and standards that all have to work together. No wonder that, in 2024, 5G adoption in the US manufacturing sector was at 2%. After all, a factory doesn’t realize any business value from just deploying 5G by itself, if the rest of the technology stack (sensors, algorithms, applications, cloud computing, security, etc.) isn’t also deployed.

Second: China targeted industrial policy to solve known market-coordination problems that would hold back IIoT adoption.

For example, consider the problem of sub-scale platforms. To better understand what this is, I’ll first lay some foundation on key terms:

A platform is any technology in which an underlying resource, such as computing power (eg. Amazon Web Services), is offered to customers as a software component to build a fully functional piece of software. In the IIoT case, “industrial internet of things platforms” are cloud platforms that allow manufacturers to (1) access compute and data storage, (2) enable data to be sent from physical machines to the cloud, and (3) secure the network and data from machine to cloud. An IIoT application is a packaged piece of software with algorithms and an end-user interface that solves a business problem.

The consumer analogy is how the iPhone is a platform and Google Maps is the application that runs on the platform, using its compute and storage. Manufacturers need the IIoT platform, and they must either (1) build the IIoT application themselves (which is difficult since manufacturers often don’t have the in-house talent), or (2) buy a prepackaged application from a vendor.

The sub-scale platform problem occurs when, in a market, there are too many platform vendors who can’t make enough money to scale their business due to intense competition and operational execution issues (identified above) and when there aren’t enough applications to actually create value for the customer, the manufacturer. The IIoT market in the United States has faced precisely this problem, especially because digital-platform markets tend toward winner-take-all or oligopoly competition dynamics (eg. iPhone vs. Android; the four major cloud-computing platforms: Amazon, Google, Microsoft, and now Oracle), and platforms make money only if application vendors build on the platform.

BCG, in a 2017 report titled “Who Will Win the IoT Platform Wars,” identified over 400 IoT platforms in the market due to the excitement of the industry at that time. But few of these platforms really grew to any significant scale, with some notable failures (see GE’s attempt above) because of the technical and operational issues. As a result, there were few IIoT application vendors building prepackaged software. There too many platforms they could choose to build on, and the lack of platforms at scale meant there were too many technical challenges that were unresolved. The value of the platform is to solve the underlying technical issues so an application developer doesn’t have to. In the IT world, a software developer doesn’t have to worry about which type of server or networking equipment is in the data center to build a cloud application. The same is true for a software developer on mobile: they don’t have to worry about the specific type of camera lens on the phone when building their app.

As a result, there are few if any IIoT applications at scale (Samsara being a notable exception). For example, there is no packaged software application that a factory own can buy to predict when any robot it chooses to deploy will breakdown today, or for any other type of equipment (of which there are literally thousands) in a factory.

Meanwhile, China’s State Council, in the same 2017 policy document, designed policies to solve the sub-scale platform problem in IIoT:

By 2020, form the industrial internet platform system, supporting the construction of approximately 10 cross-industry, cross-domain platforms, and establishing a number of enterprise-grade platforms that support companies’ digital, internet-enabled, and AI-enabled transformations. Incubate 300,000 industry-specific, scenario-specific industrial applications, and encourage 300,000 enterprises to use industrial internet platforms for research and development design, production manufacturing, operations management, and other business activities. The foundational and supportive role of industrial internet platforms in industrial transformation and upgrading will begin to emerge.

到2020年，工业互联网平台体系初步形成，支持建设10个左右跨行业、跨领域平台，建成一批支撑企业数字化、网络化、智能化转型的企业级平台。培育30万个面向特定行业、特定场景的工业APP，推动30万家企业应用工业互联网平台开展研发设计、生产制造、运营管理等业务，工业互联网平台对产业转型升级的基础性、支撑性作用初步显现。

Like most industrial policies in China, the State Council’s high-level policy guidance becomes operationalized in provincial- and city-level policies via funding and other incentives. For example, Jiangsu 江苏 province set a goal of establishing 1,000 “smart” (aka enabled by cloud, AI, advanced connectivity, etc.) factory workshops in 50 provincial-level factories by 2020.

What can the United States learn?

If we’re serious about revitalizing US manufacturing or maintaining leadership in emerging technologies such as AI and quantum computing, here are some things US policymakers should consider:

1. The free market, while efficient for specific markets, may not optimize for transforming entire sets of industries. The technologies for the industrial internet of things were available in the United States — but due to technical and market-coordination challenges, adoption has lagged behind that of China. AI and quantum are foundational technologies that may require an even greater level of market coordination to overcome operational and technical obstacles compared to that of the industrial internet of things.

2. Industrial policy needs to move beyond tax incentives, tariffs, and subsidies to make calculated bets on specific technologies, with deep technical expertise incorporated early on in the policy process. For example, in AI, the policy debate has focused exclusively on semiconductor subsidies and export controls — but there is limited if any discussion on how to make the AI data center itself easier to build and operate. High energy costs and energy availability due to the limits of the utility grid are known technical and business challenges to data center capacity today. Ultimately, the total cost of using AI to make predictions, optimize processes, and create value (eg. cost of inference) is not just the cost and efficiency of the chips, but the entire data center stack, including energy costs.

3. Successful commercialization of a set of technologies creates its own positive feedback loop, which reinforces first-mover advantages. Since China has a significant head start in digitizing its manufacturing base via IIoT technologies, Chinese vendors likely have more real-world data (by deploying more sensors), which enables firms to perfect their machine-learning algorithms, which will further improve manufacturing productivity in China relative to the United States. Robot adoption is a key example: when adjusted for labor costs, China uses 12 times more robots than the United States. This deployment of industrial robots at scale further advantages Chinese manufacturers and the entire technology stack associated with robotics (eg. operating systems for robots, robot supply chain, AI software to control the robots, software integrating robots into production processes, etc.). Recent reports of the Chinese government and enterprises mass-adopting DeepSeek only add urgency for more innovative industrial policies in the United States. Therefore, to achieve policy goals such as restoring US manufacturing or maintaining US leadership in quantum or AI, the United States must support companies to actually buy and use these technologies themselves.

While China may have “won” the initial round of the IoT platform wars, it isn’t too late for the United States, with smart policies and leadership, to win the broader industrial-technical leadership competition with China. While some may object to “picking winners and losers,” without urgent policy action, there may only be losers left to pick from.

Why is Trump appeasing Russia? What lessons can we learn from the battlefield in Ukraine? How will AI change warfare, and what does America need to do to adapt?

To discuss, we interviewed Shashank Joshi, defense editor at the Economist on a generational run with his Ukraine coverage, and Mike Horowitz, professor at Penn who served as Biden’s US Deputy Assistant Secretary of Defense for force development and emerging capabilities in the Pentagon.

We discuss….

• Trump’s pivot to Putin and Ukraine’s chances on the battlefield,

• The drone revolution, including how Ukraine has achieved an 80%+ hit rate with low-cost precision systems,

• How AI could transform warfare, and whether adversaries would preemptively strike if the US was on the verge of unlocking AGI,

• Why Western military bureaucracies are struggling to adapt to innovations in warfare, and what can be done to make the Pentagon dynamic again.

This episode was recorded on Feb. 26, two days before the White House press conference with Zelenskyy, Trump, and JD Vance. Listen now on iTunes, Spotify, YouTube, or your favorite podcast app.

Jordan Schneider: Shashank, it seems you had a lot of fun on Twitter this week?

Shashank Joshi: I was in a swimming pool with my children on holiday in the middle of England and didn’t notice until 18 hours after the fact that the Vice President of the United States had been rage-tweeting at me over my intemperate tweets on the subject of Ukraine. I provoked him into this in much the same way that he believes Ukraine provoked the invasion by Russia.

Jordan Schneider: What does it mean?

Shashank Joshi: It means the Vice President has far too much time on his hands, Jordan.

This is a pretty significant debate. Fundamentally, this was about whether Ukraine is fated to lose. His contention is that Russian advantages in men and weapons or firepower meant that Ukraine’s going to lose no matter what assistance the United States provides.

My argument was that while Ukraine is not doing well — I’m not going to sugarcoat that, I’ve written about this and it’s made me pretty unpopular among many Ukrainians — it’s not true that advantages in manpower and firepower always and everywhere result in decisive wins. Indeed, Russia’s advantage in firepower is much narrower than it was. The artillery advantage has closed. Ukraine’s use of strike drones — which we’ll talk about later — has done fantastic things for their position at the tactical level.

On the manpower side, Russia is still losing somewhere in the region of 1,200-1,300 men killed and wounded every single day. While it can replenish those losses, it can’t do that indefinitely. I’m not saying Vance is completely wrong — I’m just saying he is exaggerating the case that Ukraine has already lost and that nothing can change this.

My great worry is this is driving the Trump administration into a dangerous, lopsided, inadequate deal that is going to be disastrous for Ukraine and disastrous for Europe. I’m worried profoundly about that at this stage.

Michael Horowitz: Quantity generally sets the odds when we think about what the winners and losers are likely to be in a war. Russia has more and will probably always have more. But there are lots of examples in history of smaller armies, especially smaller armies that are better trained or have different concepts of operation or different planning, emerging victorious. Most famously in the 20th century, perhaps Israel’s victory in 1967.

Jordan Schneider: We have three years of data. It’s not like you’re playing this exercise in 2021. You’re doing this exercise in February of 2025. By the way, Mr. Vice President, your government actually has a ton of the cards here to change those odds and change the correlation of forces on the ground, which just makes the argument that this is a tautology so absurd coming from one of the people who is in a position to influence and who has already voted for bills that did influence this conflict.

Shashank Joshi: Wars are also non-linear. You can imagine a war of attrition in which pressures are building up on both sides, but it isn’t simply some mathematical calculation that the side with the greatest attrition fails. It depends on their political cohesion, their underlying economic strength, their defense industrial base, and their social compact.

The argument has been that although Russia feels it has the upper hand — it has been advancing in late 2024 at a pace that is higher than at almost any time since 2022 — there’s no denying that to keep that up, it would have to continue mobilizing men by paying them ever higher salaries and eventually moving to general mobilization in ways that would be politically extremely unpalatable for Vladimir Putin. War is not just a linear process. It’s a really complicated thing that waxes and wanes, and you have to think about it in terms of net assessment.

Michael Horowitz: That’s especially true in protracted wars. I’m teaching about World War I right now to undergraduates at Penn. One of the really striking things about World War I is if you look at the French experience, the German experience, and the Russian experience in particular, given the way that World War I is one of the triggers for the Russian Revolution, how their experience plays out in World War I is in some ways a function of political economy — not just what’s going on on the battlefield, but their economies and the relationship to domestic politics and how it then impacts their ability to stay in and fight.

Jordan Schneider: America has levers on both sides of the political economy of this war. There was a point a few weeks ago when Trump said he was going to tighten the screws on Putin and his economy. The fact that we are throwing up our hands and voting with Putin in the United Nations, saying that they were the aggressor, just retconning this entire past few years is really mind boggling. There was a line in a recent Russia Contingency podcast with Michael Kofman, where he says “The morale in Munich was actually lower than the morale I saw on the front in Ukraine,” which is a sort of absurd concept to grapple with.

Michael Horowitz: If you were to mount a defense here, what I suspect some Trump folks might say is that they believe this strategy will give them more leverage over Russia to cut a better deal. That involves saying things that are very distasteful to the Ukrainians, but they think as a negotiating strategy, that’s more likely to get to a better outcome.

Shashank Joshi: That’s right, Mike. Although they’ve amply shown they are willing to tighten the screws on Zelenskyy. If you were looking at this from the perspective of the Kremlin, would you believe General Keith Kellogg when he says, “If you don’t do a deal, we’re going to ram you with sanctions, batter you with economic weapons"? Or do you listen to Trump’s rhetoric on how we’re going to have a big, beautiful economic relationship with Russia and we’re going to rebuild economic ties, lift sanctions?

You’re going to be led into the belief that the Americans are really unwilling to walk away from the table because the Vice President and others are publicly saying we don’t have any cards, that the Ukrainians are losing, and if we don’t cut a deal now, then Russia has the upper hand. It puts them in a position of desperation.

My big concern is not just that we get a bad deal for Ukraine, it’s that the idea of spheres of influence appeals to Trump, dealing with great men one-on-one, people like Kim Jong Un, Vladimir Putin, Xi Jinping — and that what will be on the table is not just Ukraine, but Europe. Putin will say, “Look, Mr. President, you get your Nobel Peace Prize, we get a ceasefire, we do business together and lift sanctions. And you can make money in Moscow, by the way. Just one tiny little thing, that NATO thing. You don’t like it, I don’t like it. Just roll it back to where it was in 1997, west of Poland. That would be great. You’ll save a ton of money here. I’ve prepared a spreadsheet for you.”

That is the scenario that worries us — a Yalta as much as a Munich.

Jordan Schneider: We have a show coming out with Sergey Radchenko where we dove pretty deep into Churchill’s back-of-a-cocktail-napkin split. At least Churchill was ashamed.

It’s so wild thinking about the historical echoes here. I was trying to come up with comparisons, but the only ones I could do were hypotheticals. Like McClellan winning in 1864, or — I mean, Wendell Willkie was actually an interventionist. There was some Labor candidate that the Nazis were trying to support in the Democratic Party in 1940, but he never made it past first base. Has there ever been a leadership change that shifted a great power conflict this dramatically?

Shashank Joshi: From the Russian perspective, that’s Gorbachev. Putin would look back at glasnost, perestroika, and Gorbachev at the Reykjavik summit as moments where a reformist Soviet leader sold the house to the Americans and threw in the towel.

Michael Horowitz: You also see lots of wars end with leader change, with leadership transitions, when wars are going poorly for countries and you have leaders that are all in and have gambled for resurrection. If you think about the research of someone like Hein Goemans back in the day, then you have to have a leadership transition in some ways to end wars in some cases if leaders are sort of all in on fighting.

Jordan Schneider: The Gorbachev-Trump comparison is a really apt one because it really is like a true conceptual shift in the understanding of your country’s domestic organization as well as role in the world. Gorbachev, for all his faults, at least had this universalist vision of peace, trying to integrate in Europe — he wanted to join NATO at one point. But going from that to whatever this 19th century mercantilism vision is, is really wild to contemplate.

Shashank Joshi: The other thing to remember is Gorbachev’s reforms eventually undid the Soviet empire. They undid its alliances and shattered them. In the American case, the American alliance system is not like the Soviet empire. France and the UK are not the Warsaw Pact. We bring something considerably more to the table. It’s a voluntary alliance. It’s a technological, cultural alliance. These are different things.

Subscribe now

I worry sometimes that this administration or some people within it — certainly not everybody — views allies just as blood-sucking burdens. What they don’t fully grasp is how much America has to lose here. I want to say a word on this because Munich — and I heard this again — the FT reported recently that some Trump administration official is pushing to kick Canada out of the Five Eyes signals intelligence-sharing pact.

Now okay, the Americans provide the bulk of signals intelligence to allies. There’s no surprise about that. But if you lost the 25% provided by non-US allies, it will cost the US a hell of a lot more to get a lot less. It will lose coverage in places like Cyprus, in the South Pacific, all kinds of things in the high north, in the Arctic in the Canadian case. This administration just doesn’t understand that in the slightest.

Michael Horowitz: Traditionally what we’ve seen is regardless of what political hostility looks like, things like intelligence sharing in something like the Five Eyes context continues — in some ways the professionals continue doing their jobs. If you see a disruption in that context, that would obviously be a big deal.

Jordan Schneider: Just staying on the Warsaw Pact versus NATO in 2025 today, America plus its allies accounted for nearly 70% of global GDP during the Cold War. The economic outflows that were needed to sustain Soviet satellites eventually bankrupted the USSR. America isn’t facing anything resembling that situation by stationing 10,000 people in Poland and South Korea.

Michael Horowitz: We are in a competition of coalitions with China, and it is through the coalition that we believe we can sustain technological superiority, economic superiority, military power, et cetera. Look at something like semiconductors and the role that the Netherlands plays in those supply chains, that Japan plays in those supply chains. There are interconnections here. We have thought that we will win because we have the better coalition.

Shashank Joshi: That’s an interesting question to ask more conceptually — does this administration want a rebalancing of its alliances or does it want a decoupling? You could put it in terms of de-risking and decoupling if you want to echo the China debate here. Does it simply want more European burden-sharing? But fundamentally the US will still maintain a presence in Europe, underwrite European security, and provide strategic nuclear weapons as a backstop. That is what many governments are trying to tell themselves.

The more radical prospect is that whilst there are some people who envision that outcome — Marco Rubio, Mike Waltz (the National Security Advisor), and John Ratcliffe (the head of the CIA) — the President and many of the people around him view things in considerably more radical terms. It’s more of a Maoist cultural revolution than a kind of “I’m Eisenhower telling the Europeans to spend more.”

Jordan Schneider: There’s this quote from Marco Rubio that’s really stuck with me from a 2015 Evan Osnos profile where he talks about how he has not only read but is currently rereading The Last Lion, which is this truly epic three-part series. The middle book alone is most famous, which is what Rubio was referring to, where Churchill saw the Nazis coming when no one else did and did everything he could in the 30s to wake the world up and prepare the UK to fight.

Rubio is referring to this moment by comparing it to how he stood up to the Obama administration when they were trying to do the JCPOA nuclear deal with Iran. To go from that to having to sit on TV and blame Ukraine for starting the war, I think is just the level of cravenness. There are different orders and degrees of magnitude.

Shashank Joshi: You have to think about this not in terms of a normal administration in which people do the jobs assigned to them by their bureaucratic standing. You have to think about it like the Kremlin, where you have power verticals, or an Arab dictatorship where you have different people reporting up to the president. Think of this like in Russia, where you have Sergey Naryshkin, the head of the Foreign Intelligence Service, who may say one crazy batshit thing, but actually has no authority to say it. In which Nikolai Patrushev may say another thing, in which Sergey Lavrov may lay down red lines, but they have no real meaning because there’s a sense of detachment from the brain, the power center itself. Ultimately, it’ll still be Putin who makes the call. I think it’s a category error if we try to think about this administration as a normal system of American federal government.

Michael Horowitz: I will say, I can’t believe I’m now going to say this, but let me push back and say that there’s a lot of uncertainty about what the Trump administration wants to accomplish here, given the way they have embraced the notion that Trump is a master negotiator. To be professorial about it, in a Thomas Schelling “threat that leaves something to chance” way, or like madman theory kind of way, they think that there’s a lot of upside here from a bargaining perspective.

Most of Trump’s national security team is not yet in place. We just had a hearing for the Deputy Secretary of Defense yesterday. Elbridge Colby, who’s the nominee for undersecretary, has a hearing coming up, I think either next week or the following week. So a lot of the team is still getting in place.

Jordan Schneider: The thing about Trump 1.0 is there weren’t wars like this. You had two years of sort of normal people who were basically able to stop Trump from doing the craziest stuff. Then the COVID year was kind of a wash. But Trump 2.0 matters a lot more, it’s fair to say, over the coming four years than it did 2016-2020.

Shashank Joshi: It’s much more radical. In the first term, John Ratcliffe had his nomination pulled as DNI because he was viewed as inexperienced and not up to the job. Today, John Ratcliffe looks like Dean Acheson compared to the people being put into place. We have to pause and make sure that we recognize the radicalism of what is being put into place around us.

When you look at the sober-minded people who thought about foreign policy — and I include amongst this people I may disagree with, like Elbridge Colby, who will be probably the Pentagon’s next policy chief — what is the likely bureaucratic institutional political strength they will bring to bear when up against those with a far thinner history of thinking about foreign policy questions?

Jordan Schneider: I haven’t done a Trump-China policy show because I don’t think we have enough data points yet. But what, if anything, from the past few weeks of how he’s thinking and talking about Russia and Ukraine, is it reasonable to extrapolate when thinking about Asia?

Shashank Joshi: Two quick things. One is I see significant levels of concern among Asian allies. The dominant mood is not, “Oh, it’s fine, they’re going to just pull a bunch of stuff from Europe, stick it into Asia and it’ll be a great rebalancing.”

Number two, I think this is important: there is a strong current of opinion that views a potential rapprochement with Russia as being a wedge issue to drive between Russia and China, the so-called reverse Kissinger. Jordan, you know much more about China than I do. I’m not going to comment further on that, but I will say I believe it is an idea that is guiding and shaping and influencing current thinking on the scope of a US-Russia deal.

Michael Horowitz: You certainly have a cast of officials who are pretty hawkish on China, which will be a continuation in some ways of the last administration and the first Trump administration. I think the wild card will be the preferences of the president. There was a New York Times article a few days ago that talked about Trump’s desire for a grand bargain with China — his desire to do personal face-to-face diplomacy with Xi as a potential way to obtain a deal.

Now I think the reality is that every American president that has tried to do that kind of deal, whether in person or not over the last decade, has found that there are essentially irreconcilable differences. There’s a reason why there is US-China strategic competition and why that has been the dominant issue in some ways of the last several years and probably will be over the next generation. But Trump may wish to give it a shot — and it sounds like, at least from that article, that he might.

Jordan Schneider: We’ve also had every administration in the 21st century try to start their term by trying to reset relations with Russia. “Stable and predictable relationship” was Biden’s line. Maybe this stuff is just a blip, but I think Shashank’s right. We’re in really uncharted territory.

Paid subscribers get access to the rest of the conversation, where we discuss…

• AI as a general-purpose technology with both direct and indirect impacts on national power,

• Whether AGI will cause instant or continuous breakthroughs in military innovation,

• The military applications of AI already unfolding in Ukraine, including intelligence, object recognition, and decision support,

• AI’s potential to enable material science breakthroughs for new weapons systems,

• Evolution of drone capabilities in Ukraine and “precise mass” as a new era of warfare,

• How China’s dependence on TSMC impacts the likelihood of a Taiwan invasion,

• Whether AGI development increases the probability of a preemptive strike on the US,

• How defense writers and analysts help shape policy and build bureaucratic coalitions,

• Ukraine as a real-world laboratory for testing theories about warfare, and what that means for Taiwan’s defense.

Jordan Schneider: Let’s talk about the future of war. There is this fascinating tension that is playing out in the newly national security-curious community in Silicon Valley where corporate leaders like Dario Amodei and Alex Wang, both esteemed former ChinaTalk guests, talk about AGI as this Manhattan Project-type moment where war will never be the same after one nation achieves it. What’s your take on that, Mike?

Read more

Off all the wild moves we’ve gotten out of this Administration so far, basic science funding could be the dumbest and hardest to reverse.

does a great job with the basic plot.

Noahpinion

What happens when we gut federal science funding?

Back at the beginning of his first term in 2017, Trump briefly tried to cut federal science spending. I wrote a pair of posts for Bloomberg saying that this was a bad idea. Here are some excerpts from the first of those…

Read more

10 days ago · 252 likes · 68 comments · Noah Smith

I’d like to spotlight the newest NSF directorate, Technology, Innovation and Partnerships (TIP) created by the CHIPS & Science Act, that has been particularly hard-hit by DOGE. The idea was to supplement the world-class basic research that NSF does with more use-inspired and translational research with higher technology readiness levels. I’ve been following this directorate since its creation, recorded a panicked emergency pod when for a hot minute Senate Commerce almost killed it, and have been really impressed with its work so far.

TIP helped stand up NAIRR, has done a fanstastic job helping catalyze regional innovative hubs, and is the only org I’ve seen in government actually be strategic about workforce development. My personal favorite its new APTO program, which is creating the data and intellectual substrate necessary to really do smart S&T and industrial policy. For more of what TIP has been up to, check out their Director’s annual letter here. I’d also encourage DOGE to have a read of the TIP’s roadmap for the next few years and try to spot stuff that America doesn’t need.

The NSF is not perfect. IFP has some excellent proposals on how to incorporate novel funding strategies like lotteries that need faster adoption. But IFP also recently wrote up how the NSF showed its mettle, and was able to move faster than the NIH for COVID-related grants. TIP in particular has collected some of NSF’s most forward-thinking talent and is experimenting with novel programs and funding strategies faster than anyone else in the NSF mothership.

American basic research is our golden goose and the envy of the world, building the basis for scientific innovations that make us richer, live longer, and make us more powerful. Our universities attract the best minds in the world which is an enormous boon to the country, and absent radical intervention will continue to do so. While the NSF could use reform, we are criminally underfunding R&D already, and firing the most forward-thinking junior staff in the directorate singled out by national security heavyweights as critical to competing with China is an error this administration should correct.

Try Picking on Someone Your Own Size

DOGE should really try taking on some government programs that aren’t already running lean, creating the future, preventing pandemics and saving lives. The real discretionary bloat isn’t malaria bednets and fundamental physics research but F-35s and carriers. A real push at a few deadweight DoD programs could deliver way more savings than whatever you can squeeze from NSF and USAID and likely make for a more effective force.

From Jennifer Pahlka:

The only way the DoD was really going to change was through major budget cuts — something that forced people’s hands into new ways of working, into true prioritization, into processes that took less time because they were less burdened by the trappings that come with enormous budgets. I began my comment with an apology to the senior Air Force official sitting next to me, a caveat that I meant no disrespect, and wasn’t arguing for less military might — in fact, what I wanted was a more capable military. To my surprise, he piled on. “She’s right,” he said. “But it has to be much deeper than anything we’ve seen before. We had to cut during the last sequestration, and it was around 15% off the top of everything, which doesn’t force meaningful choices. It needs to be like half.”

To get at wasteful DoD programs and acquisitions regulations this administration would have to do the hard work of wooing Congresspeople into taking votes that would more substantially impact their districts. I hope that Trump 2.0’s staff has the stomach and topcover for this sort of work that could yield real long-term dividends for the country, not just grabbing the lowest hanging political fruit which really even have long term fiscal relevance like cutting probationary employees, foreign aid, and basic R&D.

From a ChinaTalk episode coming out on Monday with Mike Horowitz, former Biden DoD official, and The Economist’s Shashank Joshi:

Jordan Schneider: And I think this is like one of the many shames of the Trump imperial presidency. He has enough control of Congress to do this well and could even get some Dem votes for real defense reform!

Mike Horowitz: Let me muster a point of optimism here. If you look at Hegseth's testimony, his discussion of defense innovation is very coherent. He has takes that are not structurally dissimilar to the ones that we have been making.

There is a potential opportunity here for the Trump administration to push harder and faster on precise mass capabilities, on AI integration, and on acquisition reform in the defense sector. Because the president right now seems to have a strong hand with regard to Congress. Whether the president's willing to use political capital for those purposes is not clear. How the politics of that will play out is unclear. But if the Trump administration does all the things that it says it wants to do from a defense innovation perspective, that may not be a bad thing!

Shashank Joshi: My concern is also that you have people who are good at radicalizing and disrupting many businesses and sectors and fields of life. But the skills that are required to do that are different to the skills in a bureaucracy like this. Because, just because you were able to navigate the car sector and the rocket sector, doesn't mean you know how to cajole, persuade, and massage the ego of a know-nothing congressman who knows nothing about this and who simply cares that you build the attributable mass in his state, however stupid an idea that is, and who wants you to sign off on the 20 million dollars.

I worry that they will either break everything, and I'm afraid what I'm seeing DOGE do right now with a level of recklessness and abandon is worrying to me as an ally of the United States from a country that is an ally, but also that they will just not have the political nous [British for common sense] to navigate these things to make it happen. Just because Trump controls Congress and has sway over Congress doesn't mean that the pork barrel politics of this at the granular level fundamentally change. You need operatives, congressional political operatives. A tech bro may have many virtues and skills, but that isn't necessarily one of them.

Here’s to hoping! Howabout a Washington quote to send us off, from a 1775 letter sent to General Schuyler: “Animated with the Goodness of our Cause, and the best Wishes of your Countrymen, I am sure you will not let Difficulties not insuperable damp your ardour. Perseverance and Spirit have done Wonders in all ages.”

The Death of AI Safety: Moving Past the Pantomime

Tim Hwang is a writer and researcher. Relevant to the NSF topic above, he also hosted a great podcast series on metascience you should check out!

The AI Action Summit, which closed just over two weeks ago in Paris, will be remembered as a historically important gathering — though not how many of its organizers, attendees, and contributors anticipated. Rather than cementing AI safety as a priority for transnational collaboration, it turned into a memorial service for the safety era.

Billed as the successor to the high-profile gatherings of leaders that took place in the United Kingdom in 2023 and Korea in 2024, the Summit was originally intended to build on the frenetic activity that has taken place over the last few years to create international machinery for collaboration on AI safety issues. This has included an agreement on statements of principles, the formation of AI Safety Institutes around the world, and a blue ribbon “IPCC-style” report on safety issues.

This Summit’s lasting moments, however, came not from the success of “open, multi-stakeholder and inclusive approach[es]” on safety championed by the official declaration from the event, but instead dramatic declarations of national primacy unshackled by safety concerns. Vice President JD Vance’s speech made little accommodation for either safety or internationalism, declaring that the United States was “the leader in AI and our administration plans to keep it that way,” and that he was not here to talk about AI safety but instead “AI opportunity.” Macron touted a massive €110 billion fund to back AI projects in France, and the United States and United Kingdom declined to sign the Summit’s declaration language. A wildcat “Paris Declaration on Artificial Intelligence” backed by private industry hit the Summit for failing to back a “strong, clear-eyed, and Western-led international order for AI.”

A sense of stuckness prevailed in the side conversations and events taking place throughout Paris. At the AI Security Forum, a slow carousel of speakers ran through very much the same tropes and ideas that had dominated the discourse for years. Shakeel Hashim captured a feeling widely held — that the Summit was a “pantomime of progress” rather than the genuine article.

The photo ops were taken, and the keynotes were done, but the old gestures and governance rituals — which had seemed so potent just a few years ago in Bletchley Park — are now odd anachronisms in the harsh light of 2025.

This isn’t just a vibe. The “AI safety community” has always nurtured a shared, but often unspoken, agreement that public-minded technical expertise and international cooperation were the most promising pathways to promote good global governance of the technology.

But the safety community made a historically bad bet. The wheels were already coming off multistakeholder, international governance in the world at large even as the safety community began to invest in it seriously in the mid-2010s. Resurgent nationalism, great-power competition, and the fecklessness of international institutions have limited options for global governance across many domains, and AI has been just another one of the casualties. This isn’t just about Trump winning: these changes in the international system are structural, and the domestic shifts in places like France and the UK would have led to a very similar result even if Harris had pulled it out last year.

The safety community was also profligate in the use of its attention and social capital. The political influence of fair-minded technical experts turned out to be a rapidly depleting resource, wasted away as one “high-profile letter from very concerned scientists” and “dramatic demo of hypothetical model threat” followed another to little effect.

Against such a backdrop, it’s no wonder that AI safety in 2025 feels ever more like pantomime. We’re still frantically pulling the same levers, even as the whole constellation of forces that move nations in general and technology policy in particular have rearranged.

We need to be asking hard questions. What are historical models for technological safety and stability in a world of fierce, unrestricted nationalism? What happens when scientific evaluation has lost its ability to persuade the policymaker? How do you slow down or stop a technological race-in-progress?

The real intellectual work is now rebuilding a theory for safety that takes these uncomfortable realities into account and builds as best it can around them.

Mood Music

Can economic warfare really work? What can we learn from the 21st century historical record of American sanctions policy?

To find out, we interviewed Eddie Fishman, a former civil servant at the Department of State and an Adjunct Professor at Columbia. His new book, Chokepoints: American Power in the Age of Economic Warfare, is a gripping history of the past 20 years of American sanctions policy.

In this show, we’ll talk about…

• The evolution of U.S. sanctions policy, from Iraq and Cuba to Iran and Russia,

• How Reagan’s deal with the Saudis turned the dollar into an economic chokepoint,

• The incredible success of sanctions against Iran, and how that playbook could have been used to punish Russia,

• Historical lessons in enforcement that are relevant for export controls on China today,

• The role of great civil servants like Stuart Levey, Daleep Singh, Victoria Nuland, and Matt Pottinger in building state power,

• Institutional challenges for economic warfare and the consequences of failure to reform,

• Strategies for writing groundbreaking books about modern history.

Watch below or listen now on iTunes, Spotify, YouTube, or your favorite podcast app.

Financial Chokepoints

Jordan Schneider: Let’s start with the Bosphorus. How does this little corner of our beautiful planet explain the evolution of sanctions?

Edward Fishman: The Bosphorus is the epitome of a maritime chokepoint. It is a narrow strait between the Black Sea and the Mediterranean Sea. Throughout history, maritime chokepoints like the Bosphorus have been critical for strategic power. Sparta was able to win the Peloponnesian War because they won a battle around the Bosphorus and blockaded it, ultimately starving the Athenians into submission. Athens had relied on the flow of grain through the Bosphorus to feed its population — that was really the whole purpose of ancient Athens’ maritime empire.

Historically, these chokepoints have been geographic features. But now, as a result of globalization, there are chokepoints in the global economy that are not geographic — the most critical of which is the U.S. dollar. This is why the book is called Chokepoints.

For thousands of years throughout history, the only way to block a maritime chokepoint like the Bosphorus was a physical naval blockade. What’s changed is that in the wake of hyperglobalization in the 1990s, the U.S. acquired the ability to block chokepoints like the Bosphorus just by weaponizing its control of the U.S. dollar.

Today, the director of OFAC, the unit at the Treasury Department that oversees sanctions policy, can sign a few documents in her office and blockade a chokepoint like the Bosphorus. This actually happened on December 5, 2022, when the G7 oil price cap went into effect. The Bosphorus was backed up with dozens of oil tankers, because Turkish maritime officials were so nervous about violating the terms of the price cap that they didn’t want the ships to cross. It took OFAC days of very intensive diplomacy with Turkish authorities to persuade them to allow the ships to cross.

Jordan Schneider: You open this book with some wild contrast. Historically, you needed triremes. Now, all you need is a piece of paper from the Treasury Department to clog up the strait in Turkey halfway around the world.

Like you, Eddie, I was a sanctions nerd in college. I wrote my thesis about the origins of the UN and did papers on sanctions policy. I remember very vividly reading this literature arguing that sanctions are useless and don’t have any big impact. There was this great quote from George W. Bush in your book where at some point in the 2000s, he said, “We’ve sanctioned ourselves out of any influence” when it came to Iran’s nuclear program. You put the spotlight on one civil servant who takes that as a challenge and through ingenuity, creativity, and a whole lot of elbow grease, is able to discover and leverage a whole new lens of American power. Let’s briefly tell the story of American sanctions pre-Stuart Levey before we discuss Iran’s nuclear program.

Edward Fishman: When Stuart Levey came in as the Treasury Department’s first undersecretary of terrorism and financial intelligence in 2004, the most recent big case of sanctions that the U.S. had was a 13-year sanctions campaign against Iraq from 1990, when Saddam originally invaded Kuwait, until 2003, when George W. Bush launches the invasion of Iraq. That embargo required full UN backing and was implemented by a 13-year naval blockade. You had literally a multinational naval force parked outside of Iraqi ports inspecting every single oil shipment going in and out of Iraq.

The lesson from this situation was that sanctions didn’t work — Saddam didn’t come to heel. He seemed to be just as aggressive, if not more so. Over time, this embargo wound up leading not only to humanitarian problems in Iraq, which are very well documented, but also significant corruption. Saddam was siphoning away oil money under the nose of the UN.

By the time Levey comes in, sanctions had been seen as something that had been tried and failed against Iraq, and in fact had paved the way for the U.S. invasion of Iraq. In many ways, the 2003 invasion of Iraq was a direct result of the perception that sanctions had failed.

When Levey started working on the Iran problem around 2004, the prospect of even doing an Iraq-style sanctions campaign against Iran was off the table because there was no way to get the UN Security Council to agree to that at the time. Bush’s comment about having sanctioned ourselves out of influence with Iran was a result of the fact that without the UN, the U.S. thought that the only type of sanctions we could impose were primary sanctions, like an embargo where U.S. companies can’t buy things from Iran or trade with Iran. The only issue is we had had an embargo in place since the mid-90s, so there wasn’t any trade to speak of between the U.S. and Iran. The two avenues of sanctions were closed off — sanctions through the UN had been discredited by the 90s, and the other, primary sanctions on Iran, had already been maxed out and had been for a decade by then.

Jordan Schneider: The other seminal piece of sanctions in American 20th-century history is the embargo on Cuba. That is the same story — we cut off trade with this country, yet Castro’s still there in 2004, some 50-odd years later. It’s interesting — if you go back even further, there was this real hope after World War II where the UN at one point was even going to have its own air force. The idea was that sanctions were going to be this incredible tool to deter bad actions by different actors around the world because the U.S. and the Soviet Union were friends and we would all police the planet in a happy-go-lucky way. That was not how the Cold War ended up working out.

In 2004, Stuart Levey started to understand that he can leverage the dollar’s role in global financial flows. Eddie, can you tell the story of how the U.S. dollar became globalized in this way?

Edward Fishman: Bretton Woods, the conference that set the rules of the road for the post-World War II economy, happened in 1944. It put the U.S. dollar at the center of the global economy and established the dollar as the global reserve currency. It made the dollar as good as gold — the dollar is convertible for a fixed rate of $35 per ounce of gold.

At the same time, it explicitly prioritized the real economy and trade over finance. John Maynard Keynes, who was one of the architects of the Bretton Woods system, said that capital controls were a very important part of the system. For the first 30 years of this new global economy that emerged after World War II, you had the dollar at the center of the world economy, but it wasn’t a particularly financialized world economy. Most states had pretty significant capital controls, and banking was a very nationalized and, in some ways, even just a regionalized type of business.

By 1971, the U.S. dollar had been losing its value for quite some time and we were running significant deficits because of the war in Vietnam. Ironically, this is when Richard Nixon unilaterally took the dollar off of the gold peg. The dollar was still at the center of the world economy, but it was no longer tethered to gold. Exchange rates were now set by the market instead of by government fiat.

In the years after that, the capital controls of the Bretton Woods system fully erode and the dollar winds up becoming even more integral to the world economy as we see financialization take off from the ’70s through the Clinton era. You get to the point where we have a foreign exchange market that is turning over seven or eight trillion dollars every single day, which is by far the largest of all financial markets.

Jordan Schneider: How did oil come to be traded in U.S. dollars?

Edward Fishman: The dollar’s role in trading oil is arguably the most important chokepoint for a number of the key sanctions campaigns of the 21st century.

After World War II, the U.S. was a large oil producer and a big exporter. The 1973 Arab oil embargo shifted our perspective, and the U.S. realized just how vulnerable it was to being cut off from Middle Eastern oil.

In 1974, Richard Nixon — who was wallowing under the political pressure of the Watergate scandal and massive deficits that we had no reasonable way of plugging — sent his treasury secretary, Bill Simon, to make a deal. Simon was a former bond trader, a New Jerseyite, a chain smoker...

Jordan Schneider: A chain-smoking New Jersey native, described by a peer as, “far to the right of Genghis Khan.”

Edward Fishman: He’s a really colorful figure. The book includes a photo of him testifying before Congress with a giant plume of smoke around him.

Bill Simon tried to think about how to plug these deficits using his financial background as a bond trader. He proposed cutting a deal with the Saudis such that, not only do they agree to keep pricing oil in dollars into perpetuity, but they actually take the dollars they earn from selling oil and reinvest them in U.S. government debt — they basically plug our deficit with the money that the U.S. is paying them for oil. He wound up taking a flight to Jeddah in the summer of 1974 — getting copiously drunk en route.

The deal worked. He cut a deal with the Saudis in which they agree to recycle their petrodollars into U.S. Treasuries. This agreement largely still exists to this day. Oil, by and large, is priced in dollars no matter who’s buying it or selling it.

Chokepoints in the global economy are typically formed by the private sector. They kind of develop naturally as businesses evolve. However, there are important moments when government intervention becomes critical.

Simon’s original deal in 1974 solidified the petrodollar, but then a few years later, as the dollar continued to slide in value, oil exporters and OPEC started getting upset because the weakening dollar was in turn reducing the real value of their oil earnings. Jimmy Carter’s Treasury Secretary, Michael Blumenthal, actually went back to Saudi Arabia and cut a new deal in which he agreed to give Saudi Arabia more voting shares at the IMF in exchange for Saudi continuing to price oil in dollars.

Jordan Schneider: Why did the Saudis even cut the deal in the first place?

Edward Fishman: The Saudis got two things. First, they got access to US military equipment, which was pretty beneficial to them. Second, which I think is more of a direct part of this deal and one that’s more easily provable through historical documents, the Saudis were able to buy U.S. government debt in secret outside of the normal auctions. Instead of participating in the public auctions for U.S. Treasuries, they had their own side deal where they could buy Treasuries. That was a big benefit to them because they were able to lock in prices and also do so without facing potential political opprobrium.

Jordan Schneider: That’s crazy.

Edward Fishman: It’s a remarkable turning point in the financial and economic history of the 20th century. There was a real shot that oil could have been priced against a basket of currencies, which in some ways makes more sense. For these countries in the Middle East and OPEC members, their entire economy basically depends on generating oil revenue. If you want stability and predictability, you don’t want to take exchange rate risk. But people like Bill Simon and Michael Blumenthal intervened and were able to get the dollar enshrined as the key part of the oil market.

The Iran Sanctions Formula and JCPOA Diplomacy

Jordan Schneider: Let’s talk about 2006, when Stuart Levey was trying to figure out how to make sanctions work against Iran. Can you explain his light bulb moment during the January 2006 trip to Bahrain?

Edward Fishman: Levey realized other countries hadn’t stopped doing business with Iran — only the U.S. had, and that’s why the sanctions weren’t working. But he realized that he could use access to the dollar as a lever to pressure foreign banks.

Typically, when you’re trying to get other countries on board for sanctions, you would go negotiate with their foreign ministry and say, “We think what Iran’s doing is bad. You should impose your own sanctions on Iran.” That was the paradigm before 2006. What Levey realizes is that he can go directly to the CEOs of foreign banks, bringing declassified intelligence demonstrating how Iran uses their banks to finance their nuclear program, and funnel money to terrorist proxies like Hamas and Hezbollah. To start, he could just present the facts and potential reputational concerns would often persuade these banks to exit Iran. In more extreme circumstances, when banks wouldn’t go along with him, he could threaten their access to the dollar to try to get them out of Iran.

What Levey really pioneered was the direct diplomacy between him as a Treasury official and his team at the Treasury Department with bank CEOs. You might ask, how did Stuart Levey get meetings with CEOs of banks all around the world? He was lucky — right when he had this epiphany, Hank Paulson, who had been the CEO of Goldman Sachs, came in as Treasury Secretary. Paulson is arguably the most well-connected banker in the world at the time. Hank winds up opening a lot of doors for Stuart and getting him meetings with ultimately more than 100 of the key banking CEOs around the world.

Jordan Schneider: Interestingly, you have to convince all the banks to get on board, because even the slightest institutional leakage would allow Iran to sell as much oil as they want.

How did Levey and his team go about convincing the Russians, the random Chinese banks, the Azerbaijani banks, and all of these other banks?

Edward Fishman: What Levey succeeds at doing between 2006 and 2010 is getting the big name-brand global banks to exit Iran. By and large, there are a few stragglers like BNP Paribas. Most of the big main global banks are out of Iran by 2010, though there are still some banks in places like the UAE, Turkey, and other countries doing business with Iran.

What winds up happening at that time is Congress, which has very little faith in Barack Obama’s willingness to come down hard on Iran — namely because Obama had very explicitly run for president in 2008 saying he wanted diplomacy. He even exchanged letters with Ayatollah Khamenei.

Even Iran hawks that are on the Democratic side of the aisle, like Bob Menendez, don’t really have much confidence that Obama is going to be tough on Iran. Democrats and Republicans basically form almost a coalition against the Obama administration on Iran sanctions and wind up passing progressively harsher sanctions legislation.

The key part of these sanctions laws, the first one called CISADA (the Comprehensive Iran Sanctions Accountability and Divestment Act of 2010), is that they require the Obama administration to impose what’s called secondary sanctions. That’s not sanctions directly on Iran, but sanctions on Iran’s business partners — for instance, the UAE or Turkish bank that I mentioned before.

Levey was a Bush appointee retained by the Obama administration (he’s one of only two very senior officials, along with Bob Gates, who’s kept on). He uses this law with the mandatory secondary sanctions as a significant cudgel. He goes to places like Dubai and talks to banks saying, “Look, if you don’t get out of Iran, I will be forced by American law to impose sanctions on you. You will lose access to the dollar and all of your assets will be frozen.” That threat is very significant. When the choice is between Iran and the United States dollar, it’s a pretty easy choice for most banks around the world.

Secondary sanctions had been tried before in the mid-90s, but the U.S. effectively wound up blinking and not imposing secondary sanctions on Total, the French oil company that had been investing in Iran’s oil sector. Even the George W. Bush administration decided not to impose secondary sanctions. This tool was very controversial. You can imagine it didn’t go down well with other countries. If you’re an American diplomat and you go meet with one of your counterparts abroad and say, “Sorry, we have to sanction your biggest bank if they don’t stop doing business with Iran” — that just feels like mafia diplomacy, not something that goes down very easily.

One of the virtues of Obama being so beloved around the world was the success of sanctions on Iran. Obama built international consensus that Iran’s nuclear program was a problem.

Jordan Schneider: We also had multilateral sanctions from the UN alongside U.S. action. What did that end up doing for the Obama psyche and the global push to limit Iran’s oil revenue?

Edward Fishman: Obama successfully got a major UN Security Council resolution done in the summer of 2010, right alongside when CISADA, the secondary sanctions law, passed Congress.

Jordan Schneider: In the Medvedev era, mind you.

Edward Fishman: Yes, exactly. Historical contingency matters — the fact that Medvedev was president of Russia at the time meant that Russia didn’t veto UN Security Council Resolution 1929. In retrospect, the benefit of that resolution wasn’t so much the specific sanctions it imposed on Iran. Rather, it explicitly drew connections between Iran’s banking system and energy sector with its nuclear program. This meant when Obama officials traveled the world to tell foreign banks and their governments that they’d be forced to impose sanctions if they didn’t stop doing business with Iran, they could credibly say they were just complying with UN Security Council Resolution 1929 and that international law was on the side of the United States. The legitimacy that Obama’s sanctions campaign derived from the UN was ultimately very significant.

Jordan Schneider: Iran was completely unprepared for this. They literally took out ads in newspapers in Austria to beg for help financing their nuclear program.

Edward Fishman: Exactly. This speaks to assumptions about how the global economy worked at the time. People just trusted that banking networks wouldn’t be weaponized. Iran really thought that they could publicly advertise these fundraising activities with no issue. Foreign banks weren’t aware of what Iran was doing and weren’t particularly worried about being penalized for it. They probably viewed sanctions as something that were unlikely to happen to them — and if they did happen, they could just be chalked up as a cost of doing business.

Jordan Schneider: Let’s talk about the penalties. One of the remarkable accomplishments of the Treasury Department, which the export controls regime on China over the past few years hasn’t been able to do, was the billion-dollar fines thrown on violators — $2 billion on HSBC, and almost $10 billion on BNP Paribas. How did this work?

Edward Fishman: This is a very important part of the story and one that often goes unnoticed. It’s not that sanctions didn’t exist before this period in the early part of the 21st century — it’s that the cost of violating them wasn’t particularly high.

One of the most important strategic legacies of the campaign against Iran pioneered by Stuart Levey is conscripting banks to be frontline infantry of American economic wars. This wasn’t because banks decided that this was morally righteous, it was because they realized that violating sanctions was existentially dangerous for their businesses.

Between 2010 and 2014, Standard Chartered wound up getting fined about a billion dollars, HSBC was fined $2 billion, and BNP Paribas was fined $9 billion. In each case, the New York Department of Financial Services actually threatened to withdraw banking licenses from each of those banks, which would eliminate their ability to do business in the United States. That was a sword of Damocles hanging over these banks — U.S. law enforcement probably could have extracted even bigger fines.

We’re still living with that legacy today. The reason that financial sanctions in particular are so powerful is a confluence of two factors.

1. The dollar is essential to international commerce. Trying to do business across borders without access to the dollar is like trying to travel without a passport.

2. The U.S. actually can weaponize the systemic significance of the dollar because banks are afraid of going against American government dictates.

Jordan Schneider: The political economy of it is also different than whacking Nvidia or Synopsys, becauce those three banks are foreign. It is one thing to threaten with extinction some hoity-toity French bank that sponsors the French Open and has been doing business with Iran forever. It’s another to threaten a major contributor to America’s national competitiveness, employment, and growth.

Compare the death sentence of being cut off from the New York Federal Reserve versus mere fines in the case of export controls. With Huawei, there were some cases where they threatened to put executives in jail. Over the past few years, the types of companies that the Biden administration has gone after have often been random Russians in Brooklyn smuggling chips into Russia and China. Whereas the Obama administration was trying to put teeth behind big economic warfare efforts by throwing down billion-dollar fines.

Edward Fishman: Is it possible to conscript tech companies in the same way that banks are conscripted? My own view is yes. If the fines were harsh enough and if the enforcement were strong enough — because the other fact we haven’t talked about is it wasn’t just fines for these banks, it was also independent monitors. The Justice Department sent in people to oversee compliance reforms for several years thereafter.

It is possible, though politically challenging, on one hand to be subsidizing American semiconductor companies to the tune of 50-plus billion dollars, and then on the other to say we’re going to take that money back because you’re violating export controls. It is possible.

One thing I would mention though is that with the BNP fine and the HSBC fine, those took many years to come to fruition. These were years and years of bad behavior that then eventually led to giant fines. It is possible that someone right now at the Justice Department is working away at a major export control violation case that we’ll learn about maybe in a couple of years.

Jordan Schneider: You mentioned “Mafia diplomacy” as a sort of derogatory term for sanctions tactics. There are a lot of moments in this story where gentlemanliness appears to be very important to Obama.

After the invasion of Crimea, around the Maidan revolution, Obama had a call with Putin where he warned that “Moscow’s actions would negatively impact Russia’s standing in the international community.” Putin’s response was basically like, “I don’t know, man, it’s hard to take you seriously.”

Why was Obama’s demeanor so helpful in the case of Iran?

Edward Fishman: Obama was very attuned to international law, or as you put it, gentlemanliness. You could argue he was very lawyerly in his approach. With respect to the Iran sanctions, I think it actually wound up being helpful because the secondary sanctions against Iran were beyond anyone’s imagination.

We haven’t talked yet about the oil sanctions, which were put in place in 2012. The U.S. successfully reduced Iran’s oil exports from 2½ million barrels a day to 1 million barrels a day over about a year. This is explicitly a unilateral U.S. sanction.

Would that have worked as well had Obama not been as attuned to diplomacy and invocations of international law? I’m not so sure. You may have seen more challenges from places like China and India and maybe more obstinance. I do think it was helpful in some regards.

Subscribe now

Looking at all the various examples of economic warfare that I talk about in the book, this is in some ways the most remarkable because of how unlikely it is to succeed. But it works.

One big exception from the financial sanctions during the Stuart Levey era is the Central Bank of Iran. The Central Bank of Iran is not under sanctions because it’s the repository for all of Iran’s oil revenues. The Obama administration was really nervous that if they sanction the Central Bank of Iran, other countries won’t be able to pay Iran for its oil. All of a sudden you’ll have all of Iran’s oil go off the market overnight, you’ll have a giant spike in oil prices, and everyone will be in a world of hurt.

Senator Bob Menendez, who was the key Iran hawk in the Democratic Party...

Jordan Schneider: For international listeners, Menendez is now in jail for having taken gold bars from Egypt. But anyways, continue, Eddie.

Edward Fishman: It’s a wrinkle in the story. Then Mark Kirk, who’s his Republican counterpart, who also wants to do a naval quarantine of Iran — the two of them basically say, “We don’t care, Obama, we’re going to sanction Iran’s central bank.” That amendment passes 100 to 0 in the Senate.

Obama is left with figuring out how to make this work. They come to a compromise with the Hill in which they agree to sanction the Central Bank of Iran, but they create two exceptions. One is an exception for countries who every six months significantly reduce their purchases of Iranian oil. For instance, if you’re a Chinese bank, you’re exempt from this — you can pay the Central Bank of Iran so long as China as a whole every six months reduces its overall purchases of oil from Iran. This gives a glide path for Iranian oil sales to decline over time and winds up working marvelously, luckily with the ramping up of shale production in the U.S.

The other exception put in place in 2012 says you can pay the Central Bank of Iran if you’re a Chinese refinery or bank, but those payments have to go into an escrow account that stays inside China and can only be used for bilateral trade between China and Iran.

This actually gives Chinese entities an incentive to comply, because keeping this money in China is going to boost Chinese exports to Iran — there’s nowhere else that the Iranians can use the money.

The one-two punch of these gradual oil reduction sanctions and the escrow accounts leads to a situation where Iran’s oil sales collapse by 60% by volume and it effectively has zero access to its petrodollars. Within 18 months, about $100 billion of Iran’s oil money gets trapped in these overseas escrow accounts. This is the context in which Iran’s economy really goes into free fall. Hassan Rouhani, a dark horse presidential candidate in 2013, won the Iranian presidency on an explicit platform of trying to get the sanctions lifted.

The remarkable thing about this oil sanctions regime is it’s probably the most effective oil embargo we’ve seen in modern history. It’s done unilaterally by the U.S. — no other countries are fully bought into this. It doesn’t involve any sort of naval strategy at all. There’s no quarantining of oil ships or anything. It is just using these threats of being cut off from the dollar to coax banks in places like China and India to comply with American dictates.

Jordan Schneider: This is going to be the poster child for decades of history books in that it actually created political change. It both drove home economically, causing hyperinflation and really hitting growth, and then got you a new slate of politicians who some would argue really wanted to make a deal. Looking back 15 years later, what’s your take on JCPOA and how we should think about the lessons from how the Obama administration used the leverage that they created with this oil embargo?

Edward Fishman: The JCPOA is the high point of American economic warfare in the 21st century in that you actually see sanctions leading to the outcome that the United States had set out, which was to get a peaceful resolution to Iran’s nuclear program. You can quibble about whether the terms of the JCPOA were stringent enough. However, there’s pretty good consensus that sanctions were the critical unlock to that deal.

Democrats say that sanctions were the key to getting the deal. Republicans say that sanctions were working so well that if we had only kept them in place longer, we would have gotten an even better deal. Within really a 10-year period, we flip that consensus from sanctions don’t work to sanctions are this magic bullet that just ended Iran’s nuclear program without firing a shot.

The key lesson here is that you need both economic leverage to make sanctions work and a clear political strategy. Having a clear political strategy, which was to get a nuclear deal with Iran, wound up being very important because you wind up having the international community grudgingly go along with the sanctions. They don’t voluntarily go along — they kind of have to be dragged along, including even the Europeans. But it would have been much harder to bring them along if there hadn’t been a political strategy, if it had just been bludgeoning Iran with economic pain without any sort of political end game in mind.

Responding to Russia (2014 vs. 2022)

Jordan Schneider: Let’s transition from the success of Iran sanctions to the failed response to the annexation of Crimea. What was different about how Obama and the world responded to Russia’s invasion in 2014?

Edward Fishman: Too often we tell our histories in silos — U.S. policy toward Iran vs. U.S. policy toward Russia. One thing I wanted to show in my book is that all of these sanctions campaigns are intertwined because ultimately these are the same decision makers at the table in the Situation Room across multiple issues.

The timeline is interesting here — the U.S. signed the original Iran nuclear deal, which froze Iran’s nuclear program, on November 24th, 2013. On the same exact day, hundreds of thousands of protesters descended upon the Maidan in Ukraine to protest Viktor Yanukovych’s deal with Putin.

The Ukraine crisis really does wind up taking the Obama administration by surprise. It’s not like the Iran nuclear program, which played out over the years as a slow-burning crisis. The Ukraine crisis and the Crimea annexation happened very quickly, with the U.S. constantly playing catch up. This parallel is important because right when Obama officials are scrambling to figure out what to do about Putin’s annexation of Crimea, they’re fresh off this giant victory where they just froze Iran’s nuclear program basically just by using sanctions.

It became natural for Obama officials in February-March of 2014 to say maybe sanctions could work against Russia. It’s a harder problem with Russia for several reasons. Russia has a much larger economy than Iran — in 2014 it was the 8th largest economy in the world and the world’s largest exporter of fossil fuels. Europe is completely dependent on Russian energy to heat their homes. Natural gas pipelines crisscross the continent between Russia and Europe.

Putin is creating facts on the ground as the U.S. is trying to scramble to put together sanctions. The annexation of Crimea happens within weeks of the “little green men” showing up in Crimea — they appear at the end of February and the annexation is formalized in middle of March. Shortly thereafter, Putin starts sending little green men into the Donbas, Ukraine’s industrial heartland.

Jordan Schneider: Let’s focus on the multilateral dynamic of this because obviously the UN is thrown out when Russia’s doing the thing. I remember very vividly watching the transition of the European actors who were pretty close to shrugging off this whole thing — until all those Dutch people died in the commercial liner that the Russians shot down by accident with their anti-aircraft missile. Can you explain how that changed the dynamic?

Edward Fishman: When Putin annexed Crimea in March of 2014, the U.S. and Europe did go ahead with some sanctions, but by and large they’re individual sanctions on people very close to Putin — his judo partners from childhood who have been elevated to positions of power at companies like Rosneft. Igor Sechin, for instance, the CEO of Rosneft, is sanctioned, but there are no sectoral sanctions, no actual significant economic sanctions on the Russian oil industry or its banking sector.

Obama and European leaders very publicly threatened this in March of 2014, but they don’t do anything. The reason is partly because there isn’t political will, but it’s also because they don’t know what kind of sanctions are tolerable to their own economies. They wind up spending months negotiating and coming up with what they eventually term “scalpel-like sanctions,” which effectively cut off big Russian state-owned enterprises from Western capital markets. It’s using an even narrower chokepoint than the dollar — it’s really just Western financing.

Interestingly, something that doesn’t often get recognized enough, the Obama administration went ahead with these sectoral sanctions, cutting off some big Russian energy companies and banks from U.S. capital markets on July 16, 2014, the day before MH17 was shot down. Obama and his team were getting fed up with the European foot-dragging. They say we need to send a powerful signal to Putin if we’re going to have any chance of deterring a broader invasion of the Donbas.

At the time, the New York Times was publishing headlines like, “Obama goes ahead without the Europeans.” Banking CEOs in the U.S. are incredibly upset because they’re saying this is just going to lead to a flight from the dollar to the euro and all our competitors in Frankfurt and London are going to benefit at our expense.

The next day, Putin’s proxies in the Donbas shot down a commercial airliner using a Russian-made Buk missile. They killed almost 300 people, by and large Europeans, most of them Dutch. All of a sudden the political aperture just widens completely in Europe. The Europeans are suddenly not only ready to match the U.S. sectoral sanctions of July 16, but actually go beyond them — they wind up cutting off all of Russia’s state-owned banks from the European financial system. The real core sectoral Russia sanctions are put in place after MH17, really from late July 2014 through September 2014 when Russian and Ukrainian leaders agree to the first Minsk agreement, the first ceasefire in the conflict.

Jordan Schneider: There are two parts that made me get upset rereading and reliving this story. One is that the Obama administration had just learned the lesson which Democrats in general have a really hard time with — escalate to de-escalate. It’s such an Obama thing, the same with the debt ceiling, where he was just like, “I’m going to be a nice normal actor and lay out my five demands and okay, we’ll get to two or three.” The Tea Party — this is ancient history now — and the Republicans were like, “No, we want 100% of what we want.” Obama would get scared, then they’d do a debt ceiling fight and he would end up giving way more than he realized he had to.

By the time we got to 2014, he just said “screw you.” He had the playbook with Iran. All the Treasury forecasting about the catastrophic costs of sanctions is overblown. The U.S. had more agency than expected, the euro was not going to take over.

But Russia really got away without serious economic consequences. Why didn’t Obama put the money where his mouth was?

Edward Fishman: In retrospect, there are two things that led to Obama’s overly cautious approach. One was real, genuine concern about the U.S. economy and the European economy. Remember, we’re still in the wake of the financial crisis and the Eurozone crisis is very much a live situation. There are genuine fears from the Treasury Department that you could accelerate a financial crisis in Europe if Russia were to cut off their gas supplies, and that contagion would spread to the US.

The other thing — this is an interesting paradoxical lesson for the Trump people now and people who say Europe needs to pull more of its own weight — Obama was very deferential to the Europeans over the Ukraine crisis. He explicitly wants people like Angela Merkel and François Hollande to take the lead. The negotiating block that came up with the Minsk agreement, the Normandy format, is France, Germany, Russia, and Ukraine. The U.S. doesn’t even have a seat at the table in the negotiations. Obama was saying, “This is in Europe’s backyard. It’s really their problem.”

In retrospect, that caution does not look very wise. Obama should have hit Russia much harder than he did in 2014. One interesting thing though is even though the sanctions put in place that summer — these capital market restrictions, the “scalpel-like sanctions” — are much weaker than the Iran sanctions, in the second half of 2014, oil prices cratered from over $100 a barrel to around $50 a barrel.

While the sanctions were aimed at trying to constrain Russia’s economic horizons as opposed to creating an immediate financial crisis, the sanctions do push Russia to the brink of a complete meltdown. In the winter of 2014-2015, Russia’s economy looks like it’s about to collapse — honestly just as bad, if not worse than Russia’s economy winds up looking after the much more drastic sanctions from February-March 2022.

Subscribe now

The reaction is remarkable. I have some of these quotes in the book. European leaders look at this and say, “This isn’t scalpel-like — this is what we signed up to. We didn’t want to push Russia off a cliff.” Hollande, the French president, actually says, “We explicitly don’t want to push Russia to its knees.” The Europeans, and to a certain extent the United States, got spooked by how impactful the sanctions are because they wind up being accelerated by this collapse of oil prices. Part of the reason why there’s a real frantic desire to get another more permanent agreement, which winds up being called Minsk II in February 2015, is because the Europeans really didn’t want to see Russia’s economy fall off a cliff.

Jordan Schneider: Elections matter and leadership matters. I like that you included so many McCain quotes about the events in both Iran and Ukraine, since he could have been president during these years.

Edward Fishman: One of the key ingredients of the success of Obama’s Iran sanctions is the fact that there’s this bipartisan supermajority in favor of tougher sanctions on Iran. Even if Obama had instincts to be cautious or lawyerly, Congress was passing draconian sanctions laws 100 to 0 over a veto-proof majority. With Russia, you had no sanctions laws at all.

What that speaks to, which becomes more important as our story develops, is that U.S. companies had a lot to lose in Russia. It’s not as much of a political winner for members of Congress and senators to try to layer sanctions onto Russia because they might hurt a company in their state or district. We start seeing that maybe there are domestic political limits to how far the U.S. is willing to go with economic warfare.

Jordan Schneider: Commitment to sanctions is a key factor. Secretary Lew once remarked, “One of the things the Russians would say to me is, ‘We survived Leningrad, we could survive this.’ Their definition of what they were willing to tolerate was well beyond the realm of what we would consider tolerable.”

America’s rich, and the pain that we would end up inflicting on ourselves with sanctions would only be like a half percentage point hit to our quality of life. Whereas Russia is starting from a lower baseline, and sanctions hurt them way more than they hurt us. Yet, we’re not comfortable letting ourselves be pinpricked, even if it’s to save the international order.

You wrote…

“With the loss of the Russian market, Lithuania’s dairy industry teetered on the brink of bankruptcy. When a team of State and Treasury officials met with a Lithuanian dairy farmer outside Vilnius in 2015, they expected her to express frustration. She did, but it wasn’t about her declining business. ‘You should be hitting Russia harder,’ she said.”

It doesn’t come down to economics for a lot of this stuff. There are the political economy games of the Texas senator wanting to help out Exxon or whatever, but it often is a question of moral righteousness. We live in rich countries and we can afford to go without, by and large, way more than that Lithuanian dairy farmer could go without.

Edward Fishman: That’s exactly right, Jordan. One of the macro ironies of the book is, the rise of economic warfare in U.S. foreign policy in the 21st century is partly because military force became politically toxic in the aftermath of Iraq and Afghanistan. As those wars were going south, neither Republicans nor Democrats felt like they could even fight limited military engagements, which is very different from the ’90s when there were all kinds of small wars and U.S. bombing campaigns.

Economic warfare initially is seen as more politically palatable because it’s not hurting Americans — we can sanction Iran out the wazoo and there’s no pain felt at home. But then once you get to Russia and even more powerfully once you get to China, there are real political risks for leaders who impose sanctions on these countries. Even a 10% spike in oil prices or a marginal increase in inflation can become powerful factors in the minds of American presidents and wind up constraining our ability to successfully prosecute economic warfare.

Jordan Schneider: That’s a great point. In the 90s, you had the Taiwan Straits crisis where Clinton threw a carrier there and things calmed down. You had Mogadishu, you had Yugoslavia. But there’s this moment in 2014 where the Ukrainians asked, “Can you give us Javelins, please?” The Europeans said no. Blankets don’t win wars, bullets do.

This is the heartbreaking thing — if Russia believed that the U.S. and NATO were really going to put their money where their mouth was in arming the Ukrainians for war number one, maybe they would have been more concerned — not only about the economic impact, which they clearly underpriced, but also the military impact. We have had hundreds of billions of dollars of armaments go to help Ukraine. It was totally reasonable for Putin, based on the track record of the Obama and Trump administrations, to not expect that to be the response when it came to 2022.

Edward Fishman: Looking at the real error of U.S. policy toward Russia, it’s not necessarily anything that happened in 2014 because we were dealing with a completely novel problem, an unexpected crisis. There was no playbook for sanctions on Russia. This is one area where it’s important to be empathetic to Obama and his top team because it wasn’t easy what they had to deal with. The sanctions they did put in place in 2014 wound up being really impactful — Russia’s economy effectively collapsed that winter.

The bigger indictment on American policy is what happened after February 2015 when the Minsk II agreement was signed. After that, the Obama administration took its foot off the gas on sanctions, basically saying they’re just going to maintain what they have in place. Russia very publicly interferes in the 2016 election. Obama had threatened Putin with drastic sanctions if he continued to interfere. Putin continued to interfere, and the sanctions Obama put in place in December right before he left office were really minor. That’s a bad signal.

Then you have four years of the Trump administration in which Trump does nothing on Russia sanctions. It’s a logical lesson for Putin to draw, both from the last year and a half of Obama and all four years of Trump, that he basically got away with the annexation of Crimea at a reasonable cost. That’s just speaking of the U.S. — Europe is even worse. In 2015, after the annexation of Crimea, a consortium of companies signed the Nord Stream 2 pipeline deal to double the amount of gas that Europe would get from Russia. Putin was completely within reason to assess that the West does not have the stomach for a real economic war.

Jordan Schneider: Unlike in Crimea, the U.S. sees this coming in 2022 and has months to try to get its ducks in order, to try to do everything it can to dissuade Putin from trying to take Kyiv. What happened then?

Edward Fishman: When Biden comes in, there’s a real debate amongst his advisors about what to do. Russia had accumulated all of these misdeeds that had gone unanswered. Biden himself, when he was vice president, wanted to arm the Ukrainians. He was the most hawkish member of the top Obama team on Russia, always in favor of tougher military steps to help the Ukrainians, always in favor of tougher sanctions.

There was real debate about what to do. Should they come in right away with really tough sanctions? Biden’s conclusion was that we were still reeling from the COVID pandemic, we had climate change to deal with, and China was the biggest geopolitical issue on his radar. They tried to have what they called a “stable and predictable relationship” with Russia — which is hilarious in retrospect, as “stable” and “predictable” aren’t things you necessarily ever ascribe to Putin’s Russia.

They came out of the gate in April 2021 with a modest increase of sanctions, saying, “Here’s some sanctions to repay you for all these bad things you’ve done over the last six years. But after this, we want stability and predictability.” Putin gets a summit with Biden, which he’s very happy to get. Then he pens a rambling 5,000-word essay about why Ukraine’s not a real country and should be part of Russia in the summer of 2021 while he’s in lockdown. He masses over 100,000 troops around Ukraine’s border that fall.

It becomes quite clear that Putin has designs on Ukraine. In what is probably the biggest intelligence success of the 21st century, the US intelligence community gets Putin dead to rights. They figure out exactly what his plan is, to the point where Biden starts warning American allies privately in September and October 2021 that an invasion is coming. Very soon thereafter, he starts making public warnings that invasion is coming and tries to use the threat of swift and severe consequences, particularly very dramatic economic sanctions, to deter Putin from invading Ukraine.

Jordan Schneider: Let’s talk about how they tried to build that coalition and signal those sanctions in the lead-up to the ultimate invasion.

Edward Fishman: A stroke of luck for the Biden administration was having Daleep Singh, who had played a significant role in the 2014 sanctions. He’s one of the top financial minds in Washington — a city that doesn’t have many people with deep financial markets expertise. Daleep is an exception. He was in the perfect role to orchestrate a sanctions campaign as the Deputy National Security Advisor for International Economics, overseeing the organs of the US Government that do economic warfare.

In late 2021 and early 2022, Daleep builds relationships with his fellow G7 counterparts: in Brussels, Bjoern Seibert, and in London, Jonathan Black. They start getting into the nitty-gritty of what kind of sanctions they might impose if Putin were to invade. This preparation is important not just for being ready to do something real if Putin pulls the trigger, but also for making the threat of deterrence more credible. Russia has a world-class intelligence apparatus — if all you had was Biden wagging his finger saying “You’re going to face really strong sanctions if you invade,” but there’s no actual bureaucratic movement in these capitals creating sanctions ready to go, Putin would probably assess it was a bluff. The preparation that Daleep Singh and his counterparts in Europe and Japan do is very important.

Jordan Schneider: I love how they were doing this like in secret, but also in public. They weren’t being super hard about using classified communications — they were just calling each other on their phones because they actually want the Russians to be listening and believe they are going to put real sanctions on them.

Edward Fishman: That’s exactly right. They view the preparations as important from both a practical standpoint and a signaling standpoint.

By the time we get to the moment of decision in late February, it becomes clear after Putin and Xi Jinping meet in early February that an invasion probably won’t happen until the Beijing Olympics wraps up — Putin doesn’t want to spoil Xi Jinping’s party. By that time, you have a very extensive menu of sanctions options. Most importantly, you have what’s called the Day Zero package — the raft of sanctions that would go into effect as soon as Putin invades.

The compromise is made because inflation is at a four-decade high and there are concerns about oil prices potentially spiking. Biden says they’re going to maximize sanctions on Russia but not aggressively target its oil sales, which is tough because Russia’s economy depends on hydrocarbon exports. The strategy of the Day Zero sanctions is to implement maximalist sanctions on Russian banks — Sberbank and VTB, the two biggest banks in Russia — as well as Russia’s access to foreign technologies. They took the Foreign Direct Product Rule that had been imposed on Huawei in 2020 and recast it to cover the entire Russian economy. They take something that had been previously employed on just one Chinese company and apply it against an entire state.

The tragedy of the situation is that Putin invades and very quickly — similar to that moment in July 2014 after MH17 was shot down — there’s a giant shift of the Overton window in Europe. Everyone becomes gung-ho for very aggressive sanctions after Putin invades and we start seeing just how horrible this war is and how imperialistic Putin’s goals are. Hundreds of thousands of people protest on the streets of places like Berlin, and there’s a massive political movement in favor of stronger sanctions.

Within 24 hours of the invasion beginning, the Day Zero package that Daleep Singh and his colleagues had worked months on looked much too weak and actually undershot the political moment. Within that first weekend of the war, the United States and the G7 agreed to go much further and actually sanction Russia’s central bank directly — something that was seen as too politically radical to even consider in the lead-up to the invasion. Putin clearly agreed because he had left half of his central bank reserves completely exposed to Western sanctions.

Jordan Schneider: This goes back to the mafia diplomacy concept. Ironically, Putin expected the West to be more gentlemanly and concerned about the centrality of the dollar and euro to global trading. Once the war started and the Overton window shifted — which everyone had a hard time foreseeing — things changed. Looking back, it seems silly that they didn’t anticipate massacres when Russia invaded. While sanctioning their central bank was an option, there remained questions about whether they could get the money out, and if they would even believe the threat before it happened. The actual deterrent value we had during those months remains an open question.

Edward Fishman: Clearly, we would have been better off had the U.S. and Europe created more aggressive sanctions plans in advance. This could have strengthened deterrence and weakened Russia’s economy and warfighting capability more quickly, directly helping Ukraine on the battlefield. There were significant costs to underestimating how willing political leaders would be to implement tough sanctions in the U.S. and Europe. But going back to your earlier point, Jordan — from a deterrent standpoint, would that preparation have overridden Putin’s lesson from 2014 and the seven or eight years of basically allowing Russia to get off scot-free after annexing Crimea? Putin had likely already sized this up in his head by then, and I’m not sure we could have changed his mind.

Jordan Schneider: Here’s a crank idea — why didn’t the Treasury Department go long on oil if they were worried about it spiking up to $250 a barrel? Couldn’t you just do the math that way?

Edward Fishman: This is a point I make toward the end of the book — the U.S. is much better at imposing economic penalties than deploying capital for strategic reasons. That would be a very creative use of government resources, but it’s not a bad idea. If we had the flexibility to do something like that in a strategic manner, sure. We do use things like the Strategic Petroleum Reserve to stabilize the oil market. In March 2022, the Biden administration released 180 million barrels of oil to try to stabilize the market.

Jordan Schneider: They did eventually act, but it took too long, and the Department of Energy people are complaining that the caves might crater in. Reading through your book, I can only imagine how frustrating it must be for these officials working around the clock to get the whole world to ramp up sanctions, and they can’t even get their own government to release oil for arguably the biggest crisis in at least 50 years.

Edward Fishman: Many of our institutions are built on the assumption that we live in a peaceful, predictable world, and we don’t always get our act together in time for crisis. This isn’t unique to the 21st century — it’s been true throughout American history.

Jordan Schneider: Here’s another crank idea for you. In the winter of 2023, everyone was terrified that oil prices were going to spike. Did anyone discuss geoengineering solutions, like spraying sulfur in the air over Europe to save everyone’s energy bills?

Edward Fishman: There are a number of tragedies in this story, one being that you decided to become a podcaster instead of a sanctions nerd. Had you gone down this path, maybe we would have benefited from your creativity in the U.S. government.

Institutional Dysfunction

Jordan Schneider: The people you profile, whom you clearly admire for their incredible feats of civil service, were creating new concepts and regimes unimaginable back in 2004 while operating under such constraints in such a dysfunctional system. They made enormous family sacrifices, which you mention several times. We did a show called “Is the NSC Unwell?” where we opened with Jake Sullivan being awake at 4 AM on a Tuesday during a home invasion because he was dealing with Ukraine issues.

Having the idea is the easiest part. Sure, I can suggest geoengineering to fight the impact of Russian oil, but transforming a clever idea that checks all the economic, institutional, and diplomatic boxes into reality is unbelievably difficult. Multiple times in your stories, there are eight-month delays for things that everyone should have immediately approved on day one.

Edward Fishman: We need a government that’s purpose-built for the age of economic warfare. That’s the premise of my book — we are living in an age of economic warfare. Sanctions, tariffs, and export controls are how great powers compete today and will compete tomorrow. This is a secular trend we’ve seen throughout the 21st century, yet we haven’t changed our government to actually fight and win these economic wars.

There’s nothing like the Pentagon for economic warfare. During my short stint at the Pentagon working for then-Chairman of the Joint Chiefs of Staff Marty Dempsey, I noticed that military force has one agency and a clear chain of command up to the Secretary of Defense. With economic power, you’ve got numerous agencies involved — the Treasury Department, the Commerce Department, the State Department, the Energy Department. Much time is spent just coordinating the interagency process.

Ideally, we would have a dedicated department with clear leadership for economic statecraft or economic warfare. Some governments have moved in this direction — Japan now has a cabinet-level minister for economic security. The U.S. hasn’t innovated like that. There’s a core budgetary problem where agencies like TFI (Office of Terrorism and Financial Intelligence) at Treasury, which Stuart Levey led, or BIS at the Commerce Department, haven’t seen significant budget increases despite their missions growing exponentially.

Subscribe now

Jordan Schneider: This theme comes up repeatedly in these stories and with the chip export controls. When cabinet-level officials disagree without presidential direction saying “We’re doing X, not Y, get with the program,” things stall or take longer. Cabinet members are congressionally approved; their words carry weight. When Janet Yellen believes a sanction would harm global inflation and the American economy, Jake Sullivan must call Mario Draghi to persuade her because Biden won’t act without her support. Everyone has different priorities, and without a central authority or an engaged president, you end up with stasis — allowing Russia to make an extra $200 billion they shouldn’t have throughout 2023.

Edward Fishman: Exactly. The Draghi call is one of the more remarkable episodes in the book. After the political aperture expanded during the first weekend of the Ukraine invasion in 2022, making central bank sanctions possible, the G7 agreed. Then Janet Yellen raised concerns, requiring a call from Mario Draghi, Italy’s leader and former European Central Bank chair, to personally assure her it was acceptable.

Regarding China, much of why your podcast is amazing has been its in-depth coverage of chip export controls. Looking back to the first Trump administration, export controls were deployed against Huawei instead of sanctions largely because Treasury Secretary Steven Mnuchin opposed a tough China policy. In early 2019, after the arrest of Meng Wanzhou 孟晚舟, some administration officials suggested sanctioning Huawei and putting them on the SDN list. Mnuchin refused, so they defaulted to putting Huawei on the entity list, which Wilbur Ross controlled as Commerce Secretary. The whole export controls landscape might have been very different with a more hawkish Treasury Secretary during the first Trump administration.

Jordan Schneider: You have this wild anecdote from Matt Pottinger, former ChinaTalk guest who became Deputy National Security Advisor towards the end of the Trump administration.

Pottinger noted that at one point, Bolton decided not to tell Trump about arresting Meng Wanzhou. Pottinger interpreted Trump’s rhetoric as supporting a tough stance on China.

“Pottinger told his Commerce colleagues that Trump was pursuing a two-pronged strategy. On the one hand, the president was seeking to preserve his personal relationship with Xi Jinping and the appearance of pursuing warmer ties. But as for officials in the bureaucracy, Trump ‘wants us punching as hard as we can.’ In effect, Pottinger was telling the Commerce officials to take Trump seriously, not literally — to tune out the verbal concessions that Trump made in public and keep a default position of being ‘tough’ on China.”

Presidents, even those not in their 70s, only have maybe 5% of their day for these matters. This leaves an enormous amount to be sorted out by empowered appointees and cabinet members, which explains how we ended up with export controls instead of sanctions on Huawei — quite remarkable in retrospect.

Edward Fishman: The first Trump administration has been characterized as super hawkish on China, but examining the record shows Trump himself wavered between being very hawkish and totally obsequious to Xi Jinping. The policy was shaped by different factions: people like Pottinger and Bob Lighthizer were tough on China, while Mnuchin and Gary Cohn wanted to return to the early 2000s approach — the Hank Paulson school of U.S.-China relations. These factions took advantage of opportunities when Trump leaned their way to advance their policies. Trump didn’t take a more consistently hawkish line toward China until his final year in office, when he believed Xi Jinping had lied to him about COVID, destroying his re-election chances. We’ll likely see similar dynamics in a new Trump administration — Trump vacillating while different factions capitalize on moments when he’s more receptive to their proposals.

Jordan Schneider: You close the book, Eddie, with the idea of an impossible trinity.

“We don’t yet know when the Age of Economic Warfare will end, but we can envision how. The trade-offs facing policymakers in Washington, Beijing, Brussels, and Moscow can be thought of as an impossible trinity consisting of economic interdependence, economic security, and geopolitical competition. Any two of these can coexist but not all three.”

Walk me through the 20th and 21st centuries — what different trade-offs did states make, and where are we landing now in 2025?

Edward Fishman: Let me explain why I ended the book this way. While I wrote a narrative history because I believe individuals can shape history — remove certain individuals and history would have gone differently — there are also structural reasons underlying the age of economic warfare. Consider this statistic: Barack Obama used sanctions about twice as much as George W. Bush, Trump used them twice as much as Obama, and Biden uses them twice as much as Trump. This suggests both individual agency and structural factors matter.

The geoeconomic impossible trinity I developed explains why this is happening. You can only have two of these three elements simultaneously — economic security, economic interdependence, and geopolitical competition. During the Cold War, we had economic security and geopolitical competition in a bipolar order between the U.S. and Soviet Union, but at the expense of economic interdependence — there was no meaningful economic relationship between them.

When the Cold War ended, geopolitical competition disappeared. China and Russia transformed from adversaries to potential friends, and we invested significant political capital bringing both into the liberal international order, including the WTO and other key international bodies. Without geopolitical competition, we could embrace economic interdependence without sacrificing economic security.

Today, we maintain economic interdependence while geopolitical competition has returned full force, resulting in lost economic security. This affects all major powers — the United States, Japan, European Union, China, and Russia. None feel economically secure, leading them to invest heavily in protecting themselves from rivals’ sanctions, export controls, and tariffs. To regain economic security, we must either end geopolitical competition, which seems unlikely, or significantly reduce economic interdependence. My view is we’re heading toward a significantly less interdependent global economy in the years ahead.

Jordan Schneider: You end the book with some dark words,

“Without the ability to channel geopolitical conflict into the economic arena, great powers could once again find themselves fighting on an actual battlefield. The dream of economic war, for all its downsides, is that it can be an alternative to a more violent kind of war. Someday the age of economic warfare might end, but we might miss it when it’s gone.”

Care to elaborate on this idea?

Edward Fishman: We face very significant stakes in our economic decisions today as we head toward a less interdependent global economy. This could manifest in two ways. First, a world economy where the U.S. and its allies deepen their connections. We might have less trade with China and Russia, but more with Canada, Mexico, the European Union, and Japan. Janet Yellen in the Biden administration called this “friendshoring.” Bob Lighthizer proposed this in a recent New York Times op-ed, suggesting the U.S. and other democracies create a bloc with low internal tariffs and high tariffs on everyone else.

The alternative is deploying sanctions, tariffs, and export controls arbitrarily against friends and foes alike, creating a chaotic breakdown of the global economy. We’d be forced into autarky by default, without long-term economic agreements with allies or adversaries. This scenario frightens me most because history shows that when states can’t secure resources and markets through free trade and investment, the temptation for conquest and imperialism rises.

President Trump’s talk about seizing Greenland for its mineral resources echoes Hitler’s pursuit of Lebensraum. Hitler feared being cut off from European trade after Europeans sanctioned Mussolini for seizing Abyssinia. If economic interdependence unravels into every country for itself rather than friendly blocs, we could see a return to great power war.

Jordan Schneider: Dark. I’ll refer folks back to our two-part episode with Nicholas Mulder on The Economic Weapon, which told that whole 1920s and 1930s story of how Imperial Japan and Nazi Germany developed their autarkic, resource-hungry vision. While racial ideology played a role, they were clearly terrified about accessing enough oil, minerals, and resources to remain great powers.

Researching Modern History

Jordan Schneider: Let’s shift topics. Tell me about writing history of the past 20 years. You don’t have everything declassified, you’re doing interviews, and history seems to be happening in WhatsApp groups. What was it like both as a former civil servant and then interviewing all these people to piece this recent history together?

Edward Fishman: As you know, Jordan, since we shared some classes, I studied history and in a parallel universe might be a university historian. After college, I went into government work and realized that in this era, many decisions bypass formal processes. Even back in the 2010s, decisions were made through informal communications, in coffee shops, never written down, through WhatsApp groups. This has only accelerated since I left government.

Contemporary history plays a crucial role because documentary records won’t be as valuable in 30 years as they were previously. They might even mislead — often the package going into an NSC meeting doesn’t reflect what’s actually discussed or decided. Many decisions happen outside formal meetings entirely.

This experience convinced me that the best approach was to follow Thucydides’ method — write contemporary history, documenting the times you live in, striving for impartiality. What you lose in documentary records, you gain by talking to people who were actually present. Thanks to my government experience and non-partisan reputation, I accessed everyone crucial to this story — Democrats, Republicans, and current civil servants.

Future historians will surely build on and improve the story told in Chokepoints when they access all documents. However, I hope the insights derived from my access to these people and my insider government experience will prove durable.

Jordan Schneider: Did you send Nabiullina an email?

Edward Fishman: No, I didn’t speak to Elvira Nabiullina, unfortunately. One wrinkle in the story is that I was sanctioned by the Russian government in 2022, before I even started writing. I’m currently banned from any travel to Russia.

Jordan Schneider: She’s got an open invitation to ChinaTalk. I’d love to hear her side of the story.

y through declassified documents showing what really happened — I’d bet most of the narrative around U.S. policy holds up. Rather, I hope we’ll see Chinese, Russian, or European versions of Chokepoints. While I capture those stories to some extent, the book focuses on the United States. If counterparts in those systems wrote similar books, we’d have a much more complete picture.

Jordan Schneider: Eddie and I were classmates at Yale, studying ancient history together. I love how you say you’re walking in Thucydides’ footsteps — let’s say we’re doing the same with ChinaTalk. For both of us, Donald Kagan’s classes were among the most formative in thinking rigorously about politics, history, and warfare. Any memories or reflections about his impact in the classroom?

Edward Fishman: One sad aspect of publishing this book is that Don died a couple years ago and won’t have the chance to read it. Of all my teachers, he had the biggest impact, shaping my career in many ways. He even influenced how I teach my class at Columbia on Economic and Financial Statecraft — I use his exact seminar format, with students debating each other’s papers weekly.

The main lessons I learned from Kagan that influenced the book include understanding the role of contingency in history — people and their decisions matter. While many history books focus on impersonal forces, Kagan taught me that structure sets context but free will and decisions can change history’s course. That’s why I focused on the people creating these policies.

Second, chronology matters. You must understand historical decisions within the knowledge available at the time. We tend to judge past decisions with hindsight, but understanding what people knew then reveals more about how history unfolds.

Finally, history itself matters. Kagan said, “Without history, we are the prisoners of the accident of where and when we were born.” Beyond clichés about repeating history, understanding what our predecessors did right and wrong helps us live better lives today.

Jordan Schneider: Another lesson coming through your book is that while we can debate grand strategic decisions, like Biden’s approach, the most human agency appears one or two levels below. Having someone from Goldman Sachs who understands the global insurance market enables implementing policies that might not otherwise be conceived. While we criticize civil servants in today’s America, it’s important to recognize that you can expand government’s effectiveness by empowering the right people to make decisions and analyze questions thoughtfully. For anyone at a career crossroads, read Eddie’s book and understand that your future choices matter.

Edward Fishman: I appreciate that, Jordan. If there’s one takeaway, it’s that government officials’ decisions truly matter. The protagonists I highlighted — Stuart Levey, Adam Szubin, Dan Fried, Matt Pottinger, Daleep Singh, Victoria Nuland — if you remove them from their situations, you’d have very different policies. We were fortunate to have them in those positions. Having more people with diverse skill sets willing to serve in government increases the odds of having the right person in the right place at the right time.

Mood Music—Iranian Pre-Revolution Psychadelic Rock

A friend and past ChinaTalk guest Walter Kerr is trying, ala Fast Grants, to step in the gap to provide funding to the most effective organizations most impacted by the USAID funding cutoff. He and some partners have launched the The Foreign Aid Bridge Fund. I donated a few grand this week and think you should too.

Walter runs Unlock Aid, new think tank that has done some great work to make USAID a more efficient and effective organization. Have a listen to him on ChinaTalk (iTunes, Spotify, YouTube).

On Monday, Xi Jinping hosted a symposium of top business leaders, signaling increased support for China’s private sector.

While concrete policy details have yet to emerge, the meeting featured plenty of strong rhetoric. Premier Li Qiang gave a markedly pro-stimulus speech, arguing, “We must make every effort to enhance consumption… and strive to open up a chain where consumption drives investment, industrial upgrading, employment, and income growth.”

The guest list had plenty of familiar figures, like Alibaba’s Jack Ma, BYD’s Wang Chunfu, and DeepSeek’s Liang Wenfeng. But to understand the implications of this stimulus soirée, we have to look at the lesser-known attendees as well — this event assembled propaganda czars, Politburo tech strategists, and neoauthoritarian academics to mingle with tycoons of semiconductors, chemicals, agriculture, 3D printing, and more.

Today, we’ll introduce some of the deeper cuts — eight from industry and four from government — to illuminate what Xi’s new alliance means for the emerging technology race.

Thanks to Ray Wang for writing the profile on Yu Renrong. All other profiles were authored by Lily Ottinger.

Industry leaders

Wang Xingxing 王兴兴 (Unitree Robotics 宇树科技)

Wang Xingxing is the founder and CEO of Unitree Robotics, which developed China’s most popular robot dog.

Wang’s eccentric personality, technological optimism, and unconventional approach to talent have prompted comparisons to DeepSeek CEO Liang Wenfeng. According to an April 2024 interview:

“I think by the end of 2025, there will be at least one company in the world that can produce a relatively general-purpose robotic large model.
…
In the future, humanoid robots could reshape every industry, including manufacturing, services, market production, agriculture, mining, and construction. Looking ahead to the ultimate possibilities, I believe that governments could fully deploy 100,000 humanoid robots, designate a piece of land, and build a brand-new city there. They could complete the infrastructure and provide housing for free. At that point, ordinary people wouldn’t even need to work — robots could sustain everyone. This is entirely possible.
…
Eventually, they might even be able to create robots the size of cells. … At that point, governments would definitely need to introduce regulations to prevent an uncontrolled explosion of robots — after all, they could end up consuming all available resources.”

Wang is also a believer in the potential of AGI, but thinks that large language models are the wrong way to get there. From a 2024 panel:

“I believe that embodied intelligence is the only path to achieving AGI. Currently, LLMs lack physical presence and thus have an insufficient understanding of the physical world. This is why many top AI researchers advocate for developing world models. Tesla's autonomous driving also relies on collecting real-world data for training rather than using virtual data, because virtual data lacks sufficient real-time interaction with the physical world. I think embodiment is actually a crucial pathway to AGI, and AGI is very likely to emerge from robotics companies.”

How does Xi’s symposium fit into this vision? According to Wang, hard-core innovation requires a skillful PR strategy. “If your technology is not innovative, it is meaningless. Of course, you cannot express this innovation directly. It is better not to go beyond the public's cognition too much, otherwise I think I would be scolded to death.” The publicity campaign goes beyond shaking hands with Xi — Unitree’s latest humanoids performed during the 2025 CCTV New Year’s Gala, which is of course the world’s most-watched television program.

For more on China and robots, see our features on industrial robotics and humanoid robots.

Xu Guanju 徐冠巨 (Transfar Group 传化)

Billionaire Xu Guanju is the founder and chairman of Transfar Group, a publicly traded conglomerate dealing in chemical manufacturing, supply chain logistics, agricultural biotechnology, and even finance.

Xu spent a decade as chairman of the Zhejiang Federation of Industry and Commerce, advising the local government on behalf of the private sector, and helping to shape the “Zhejiang model” of economic development that incubated both DeepSeek and Unitree. He also served as a member of the CPPCC (China’s top advisory body) and was elected as a deputy to the 13th National People’s Congress (NPC).

Subscribe now

Xu famously built his chemical empire on a humble foundation. After borrowing startup funds from friends and family, Xu began producing liquid soap in a small workshop with his father in 1986. According to one well-known story, Xu initially had to contract a chemical engineer due to his lack of expertise. This contractor proved frustratingly indispensable thanks to a mysterious powdered thickening agent he added to finish each batch of soap. Eventually, Xu agreed to pay 2000 RMB (~10,000 RMB today) for access to this “trade secret,” only to learn that the powder was just ordinary table salt. Xu’s father responded to this revelation by exclaiming, “I spent all this money to send you to school, and for what?” Xu said later, “This incident made me realize that technology is the primary productive force.”

Peng Fan 彭凡 (KOCEL 共享装备)

Peng Fan is the chairman of KOCEL Group, the machinery manufacturing company that produced the hydropower turbine blades for the Three Gorges Dam.

In 1983, Peng moved to the remote northwest region of Ningxia to work at the state-owned Great Wall Foundry​. He climbed the ranks from an ordinary casting technician to plant manager. The foundry underwent privatization and restructuring in 2003, with Peng leading the effort to transform it into KOCEL.

Peng has a postgraduate degree in casting engineering, and under Peng’s leadership, KOCEL embraced cutting-edge “intelligent manufacturing” techniques. He championed the adoption of industrial 3D printing, obtaining more than 500 patents related to advanced casting techniques.

Peng’s team cracked several difficult engineering problems, including the manufacture of heavy gas turbine casings and hydropower turbine blades. Under his guidance, KOCEL has grown into a leading global supplier of high-end equipment components.

He also served as a delegate to the National Congress of the CCP in 2012 and 2017.

Jiang Bin 姜滨 (Goertek 歌尔股份)

Jiang Bin is the chairman of Goertek, a company he co-founded with his wife, Hu Shuangmei 胡双美, in 2001. Goertek is one of the primary manufacturers of Apple’s AirPods and Vision Pro headsets.

Jiang Bin was born in 1966 in Shandong province. He earned a bachelor’s degree in engineering from the Beijing University of Aeronautics and Astronautics and later an MBA from Tsinghua University​.

With Jiang as chairman, Goertek grew from a small acoustics firm into a global supplier of microphones, speakers, sensors, and other hardware. The company has filed more than 29,000 patent applications and is now the world’s top supplier of micro speakers, MEMS acoustic sensors, and AR/VR headset components.

Apart from Apple, Goertek’s clients include Meta, Amazon, Google, Samsung, and Sony — and in turn, the company has been criticized for relying too much on the patronage of these foreign tech giants. In 2024, the company announced an investment of US$280 million to build new production capacity in Vietnam.

Jiang is currently serving as a deputy to the 14th NPC and regularly participates in government-organized industry forums. He frequently uses these platforms to promote metaverse technology. Jiang’s current net worth is reportedly more than US$5 billion.

Yu Renrong 虞仁荣 (Will Semiconductor 韦尔半导体)

Yu Renrong (虞仁荣) is the founder and chairman of Will Semiconductor, one of the top 10 fabless chip companies in the world. It is also one of the largest global image sensor providers, ranking only behind Sony and Samsung.

Yu grew up in a small town in Ningbo, Zhejiang, but his humble background did not prevent him from demonstrating his brilliance. His academic abilities brought him to one of the best high schools in his hometown, and he graduated with an Electonic Engineering degree (known at the time as the Department of Radio 无线电系) from Tsinghua University.

After gaining engineering experience at Chinese IT pioneer Inspur and sales experience at Yuelong Electronic Technology, he founded his first company (Beijing Huaqing Xingchang Technology and Trade) in 2006.

Subscribe now

He then founded Will Semiconductor in 2007, which both designs advanced semiconductors and sells electronic components. After its listing in 2017, Will Semiconductor rapidly expanded through acquisitions, including of OmniVision Technologies, CelePixel, Cerebrex, and Synaptics’s Asian Touch and Display Driver Integration business.

Today, Will Semiconductor's market value is reportedly more than US$27 billion. Yu is also recognized as China's richest chip tycoon (中国芯片首富), with a personal net worth of $6.4 billion in 2024.

Qi Xiangdong 齐向东 (Qi-Anxin 奇安信)

Qi Xiangdong is the founder of Qi-Anxin Technology Group (奇安信科技集团), which provides cybersecurity services to government agencies, commercial actors, and critical infrastructure facilities.

After serving as Vice President of Yahoo China from 2003 to 2005, Qi co-founded the consumer antivirus software company Qihoo 360 (奇虎360), which was added to the BIS entity list in 2020.

Qi is deeply integrated into China’s cybersecurity establishment, and he’s currently a member of the 14th National Committee of CPPCC and Vice Chairman of the ACFIC. He is also a vocal supporter of AI safety regulations, frequently expressing concerns about AI-enabled hacking, deepfakes, and the black box problem.

Leng Youbin 冷友斌 (Feihe Dairy 飞鹤乳业)

Leng Youbin is the CEO of Feihe (literally, “Flying Crane”), China’s leading infant formula and dairy company. He transformed Feihe into a top domestic competitor in a market once dominated by foreign brands.

Under Leng’s leadership, Feihe worked to cultivate trust among Chinese consumers in the wake of the 2008 milk scandal, investing in high-quality milk source bases and product R&D. Feihe has also implemented agricultural IOT practices, including collars that monitor the vital signs of dairy cows for early disease detection.

Just like how the CEO of Chobani served on the Homeland Security Advisory Council in the US during the Biden administration, Leng served as a deputy to the 13th NPC and as the vice-chairman of the All-China Federation of Industry and Commerce (ACFIC).

Liu Yonghao 刘永好 (New Hope 新希望)

Liu Yonghao 刘永好 is the founder and chairman of New Hope Group, one of China’s largest agricultural companies. New Hope manufactures livestock and aquaculture feed, farms and processes meat and dairy products, and even dabbles in finance and real estate.

Born in 1952, Liu began his career as a teacher. In 1982, he started a poultry breeding operation in rural Sichuan with his brothers. By 1992, their company was among the largest non-governmental conglomerates in China. Today, Liu’s net worth is more than US$6 billion.

Under Liu’s leadership, New Hope expanded from animal feed into a broad agrifood empire. He also co-founded China Minsheng Bank in 1996, one of China’s first private banks, and served as its vice-chairman​.

Liu’s political activities include serving as a committee member of the CPPCC, a deputy to the 12th NPC, and vice-chairman of the ACFIC. In the past, he’s publicly spoken out against Xi’s detentions of businesspeople.

Government Officials

Shi Taifeng 石泰峰 (PB Member, United Front Work Chief)

Shi Taifeng is a CCP Politburo member and the head of the Party’s United Front Work Department (UFWD), an organization tasked with monitoring and influencing elite groups outside the Party, including businesspeople, academics, ethnic minority leaders, and the overseas Chinese diaspora​ — which made him a natural choice to receive an invitation.

A trained legal scholar with a Master of Laws from Peking University, Shi spent part of his career as a professor at the Central Party School​. Before rising to national prominence​, he held key regional posts – including governor of Jiangsu province and Party Secretary of first Ningxia province (2017–2019) and then Inner Mongolia (2019–2022).

In Inner Mongolia, Shi oversaw a massive buildout of data centers, 5G infrastructure, and renewable energy as part of a broader strategy of anti-separatist digital governance in the region. His tenure also saw intense crackdowns on the rights of ethnic minorities and an expansion of AI-enabled censorship. In August 2020, for example, the provincial government announced a plan to force Inner Mongolian schools to teach certain subjects in Mandarin (the goal was to transition to an exclusively Mandarin curriculum by 2023, although that was publicly denied at the time). Authorities responded to complaints by blocking China’s only Mongolian-language social media platform, Bainu, and tracking down protesters.

As Ningxia Party Secretary, Shi launched an “innovation-driven strategy” of economic development and warned that not innovating is “a dead end.”

Wang Huning 王沪宁 (PBSC Member, CPPCC Chair)

Wang Huning is a top leader sitting on the CCP’s 7-member Politburo Standing Committee and currently serves as Chairman of the CPPCC. We did a podcast on Wang with Chang Che, who wrote a great profile of him a few years back (iTunes, Spotify, YouTube).

Wang made his name as a scholar and theorist rather than a regional administrator. He was a professor of international politics at Shanghai’s Fudan University, where he was a well-known advocate of “neoauthoritarianism” (新权威主义) and authored widely read books including Analysis of Comparative Politics, Analysis of Contemporary Western Politics (1988), and America Against America. His undergraduate degree is in French.

Over the past three decades, Wang has been the de facto chief ideologue for three consecutive Chinese presidents — Jiang Zemin, Hu Jintao, and now Xi Jinping​. He’s credited with formulating key political concepts like Jiang’s “Three Represents,” Hu’s “Scientific Outlook on Development,” and Xi’s “Chinese Dream” as well as “Chinese-style modernization.” He is the Party’s top ideological craftsman, “using cosmetics to dress up political policies” with unprecedented longevity across administrations.

Wang is an active proponent of AI development, which he of course frames in ideological terms. His presence at the event signals that private sector innovation will be valued as a key component of grand strategy and national power.

Ding Xuexiang 丁薛祥 (PBSC Member, 1st-Ranked Vice Premier)

Ding Xuexiang is a member of the Politburo Standing Committee and the first-ranked Vice Premier in China’s State Council. He’s also one of Xi Jinping’s closest confidants.

Unlike many Chinese politicians, Ding has a technical education. He studied engineering and worked as a materials science researcher and administrator in Shanghai’s science bureaucracy for years​. He is the only trained engineer on the current Standing Committee​. His role at the event was to make the tech entrepreneurs feel less out of place.

Ding is also the director of the Central Science and Technology Commission (CSTC), a high-level CCP body unveiled in 2023 and tasked with coordinating China’s national science and tech strategy. His contributions have thus far been organizational — setting up and leading the new governance structures for innovation and mobilizing resources across government, academia, and industry.​

Here’s Ding’s take on AI regulation:

Ding Xuexiang said that emerging technologies such as artificial intelligence can be a powerful driving force for development, but they can also be a source of risk…. We will not blindly follow the trend, nor will we participate in unrestrained international competition. China has a strong governance and regulatory system and institutional measures, and we are confident that we can manage and use artificial intelligence technology well.

Ding Xuexiang said that global governance of artificial intelligence is a global problem. If countries are allowed to compete in an unorderly manner, the “gray rhino 灰犀牛” is right in front of us. Historically, the United Nations has played a good role in controlling nuclear safety and biosafety, and its successful experience is worth learning. The United Nations should be supported to play a central role, and all countries should participate together to jointly study and formulate powerful and effective rules to ensure that new technologies such as artificial intelligence become “Ali Baba’s Cave” of treasures, rather than “Pandora's Box.”

Li Shulei 李书磊 (PB Member, Propaganda Chief)

Li Shulei is a Politburo member and the head of the Party’s Central Propaganda Department. Hi has described modernization as a Western imposition that has now become a necessity for national power.

Li was a child prodigy, entering Peking University at age 14 and later earning a doctorate in modern Chinese literature. He spent many years as a professor at the Central Party School, writing on Chinese literature and culture​ before serving as propaganda chief of Fujian province. In 2022, Xi Jinping tapped him to take over the top propaganda post​.

In 2023, Li’s Propaganda Department (along with the Cyberspace Administration) issued guidelines to ensure AI-generated content aligns with socialist values and does not undermine social stability. In his words:

“Generative artificial intelligence is one of the most revolutionary and leading scientific and technological technologies at present. We must improve the development and management mechanism of generative artificial intelligence as soon as possible, promote industrial development, technological progress and security in this important field, and achieve benefits and avoid harm and safe use. Cyberspace is not a lawless place or an enclave of public opinion. We must strengthen the construction of the rule of law in cyberspace, improve the long-term mechanism for network ecological governance, and ensure that the Internet always operates healthily on the track of the rule of law.”

As the propaganda czar, Li was invited to the meeting because events like these are more about messaging than substance. Going forward, a key indicator of the meeting’s impact will be whether Li’s propaganda continues to discuss technology in terms of control or instead pivots to emphasizing the economic benefits of tech innovation.

A friend and past ChinaTalk guest Walter Kerr is trying, ala Fast Grants, to step in the gap to provide funding to the most effective organizations most impacted by the USAID funding cutoff. He and some partners have launched the The Foreign Aid Bridge Fund and are looking to give out their first tranche of money Friday. I donated a few grand this week and think you should too.

Walter runs Unlock Aid, new think tank that has done some great work to make USAID a more efficient and effective organization. Have a listen to him on ChinaTalk (iTunes, Spotify, YouTube).

How do patents influence emerging technology innovation? How far could AI and DOGE push our current IP regime? Does it matter that China issues way more patents than the US does?

To discuss, ChinaTalk interviewed ​​Andrei Iancu, director of the US Patent Office under the first Trump administration. Andrei has degrees in aerospace and mechanical engineering, and worked at the legendary Hughes Aircraft Company before going to law school. He is currently in private practice at Sullivan and Cromwell.

Co-hosting today is ChinaTalk editor and second year law student at Duke, Nicholas Welch.

Have a listen on Spotify, iTunes, or your favorite podcast app.

We get into…

• The mounting evidence that China's patent system now dominates America’s, and whether these indicators constitute an emergency in the innovation ecosystem,

• Why some US companies now prefer Chinese courts for patent enforcement,

• The fundamental tension between private rights of inventors and public access to innovations,

• What congressional inaction on patent eligibility means for AI innovation, and the bills that congress could pass to immediately jumpstart emerging tech investment,

• What the current administration could do to help USPTO juice the economy,

• Controversy surrounding the Patent Trial and Appeal Board (PTAB), and whether DOGE could put PTAB on the chopping block,

• How Trump will approach patent law and intellectual property rights, including perspectives on appointments and potential reforms.

Thanks to CSIS for partnering with us to bring you this episode, the first in a three-episode CSIS Chip Chat series.

Legislative Omissions and the Political Economy of Patents

Jordan Schneider: Let’s start off with the central contradiction in patent law. What are the two equities that this whole legal superstructure is trying to balance?

Andrei Iancu: Patent law has existed in the United States since the founding of the country. It’s in the body of the Constitution in Article I, Section 8, Clause 8. This is the only place in the entire Constitution where the word “right” is mentioned other than the Bill of Rights. They thought it was that important. The Patent Act of 1790 was the first law passed by Congress after the country was founded. Since then, it has been a central part of the United States economy.

Patents, and in fact all intellectual property rights, balance the private right of the individual creator versus the public right to access that creation. On one hand, it gives exclusive rights, as the Constitution says, to one’s inventions on the technical side, or one’s artistic creations on the copyright side. In exchange for that exclusivity, the creator makes the invention public, and the public has the right to see it and potentially use it.

This is the quid pro quo. This tension between the private right of protection versus the public’s right of access has existed from the very beginning. Thomas Jefferson was the first head of the patent system beginning in 1790. He was reviewing patents at night, and at that time, he said patents were, “an embarrassment” to the American economic system, because they are a sort of monopoly, which he hated. He was very uncomfortable with removing the public’s ability to freely access ideas. However, around the same time he also said that patents have “given a spring to innovation beyond [his] conception.” That tension in Jefferson embodies the tension that exists in the patent system to this day.

Jordan Schneider: Over the course of your career, what role has each branch of government played in balancing the rights of creators and consumers?

Andrei Iancu: Congress creates the laws, the administration enforces and administers those laws, and the judicial branch has to interpret those laws. It begins with Congress — they have to make the patent laws in the first instance, and they’ve been struggling from the very beginning.

Once the laws are passed, it’s up to the USPTO (Patent and Trademark Office) to enforce those laws and grant patents and trademarks, while the Copyright Office registers copyrights based on those laws. The problem has been that these issues are so complicated, and the tension between the two poles I’ve mentioned is so high, that Congress has left a lot of gaps. They have been incapable of legislating.

For example, the first substantive section of the patent code is Section 101. It defines which types of technologies are eligible to receive patents. It is so complicated it hasn’t been legislated on since 1793. The last time Congress wrote a law to say what technology is in and out was in the 18th century. Now, the Patent Office and the courts are trying to figure out how artificial intelligence and DNA processing fit into this 18th-century statute.

It’s a mess because Congress hasn’t returned to it in 250 years. There is a bipartisan bill to address this issue called PERA (Patent Eligibility Restoration Act). It was introduced in the Senate last year by Senators Tillis and Coons, Republican and Democrat respectively, but the bill didn’t move. We’ll see if they introduce it again. It hasn’t been touched since 1793, and it really is important that they get to it, but the issues are really hard.

Jordan Schneider: Let’s stay on the political economy of this. On one hand, you have the political economy fights over who gets to make more money — is it the generics, the healthcare system, or the biomanufacturing companies? Then you have this philosophical arc happening above all the individual industry fights. I’m curious, Andrei, as the pendulum swings, are the changes within specific industries, or is it a broader national shift over the decades where you go from the system more supporting the patent holders versus the patent users?

Andrei Iancu: Let me make something clear here. There is only one patent system in the United States, and by definition, it has to apply equally to all technologies. You cannot have patent laws for pharma that are different from patent laws for tech. Whatever the laws are, and however you interpret them, they basically have to be the same across technologies. This concept of non-discrimination across technologies is part of international agreements that the United States has been pushing for a very long time. The TRIPS agreement is a multilateral international agreement where all the member states have ratified it, and it’s fundamental to the patent system.

Jordan Schneider: Why is that?

Andrei Iancu: By definition, you don’t know where technology is going to go, and the whole point of the patent system is forward-looking. You can’t start picking winners and losers ahead of time — that’s the main reason. The other reason is one country might want to discriminate in favor of one industry, while another country might want to discriminate in favor of another industry, and it would be completely unworkable. There are other reasons too.

Nevertheless, even though the laws ultimately have to be uniformly applicable to all technologies, there are certain technologies that drive change in the system. For example, the major tension for the last couple of decades has been between big tech and pharma. This is a complete overgeneralization, but by and large, you could say that the pharma industries and life sciences industries want and need stronger intellectual property rights for a variety of reasons. Whereas big tech companies, by and large — again, I’m super generalizing here — tend to want weaker intellectual property rights.

Whoever prevails in that fight for their own corporate interests will affect everybody. That’s the interesting thing. It’s not like tech can demand changes that only apply to tech, or pharma can demand changes that only apply to pharma. Generally, this has been the tension between these big industries, and the pendulum has been swinging back and forth according to who has had more political power in the last few decades.

Nicholas Welch: Maybe we can stay on emerging tech for a moment. I’d be curious for your take on patent policy with regard to biotech. Let me know if this characterization is true, but your approach to patent law differed from your predecessor, Michelle Lee, and your successor, Kathi Vidal, in a few ways. You might be described as very supportive of patent rights, whereas Lee and Vidal could have been concerned with patent quality. Vidal made policies which caused invalidations of patents to rise from 59% in 2021 up to 71% in the first half of 2024.

In the context of emerging tech specifically, do you feel like companies have good reasons to demand patents, or should USPTO be cautious about issuing patents to technologies we don’t quite understand yet?

Andrei Iancu: Let me challenge the premise a little bit. It is true that invalidations in post-grant proceedings at the Patent Office have risen in the last administration with my successor, and I don’t doubt the numbers you cited. But that does not indicate an increase in patent quality. Those two are completely separate things.

I very much am in favor of and always spoke about the importance of issuing and maintaining correct rights, but that goes both ways. Just because you’re invalidating a patent doesn’t mean you’re increasing the quality of the patent system. The Office could be wrong in invalidating that patent, and that is a mark of lowering the quality of the system as a whole.

Recently, the Sunwater Institute published a report that shows the Patent Office errs significantly more in the direction of incorrectly not granting patent rights than incorrectly granting patent rights. While the incorrect grant of patent rights is a few percentage points in the single digits, incorrectly denying patent rights is in the double digits, according to the recent Sunwater Institute report, and I commend you to that study.

The question then becomes, which one is worse for the economy? You don’t want errors either way. But if you’re going to err on one side or the other, which one are you going to choose?

Right now, the Office is erring too much on the side of incorrectly withholding patent rights or incorrectly removing patent rights once they have been issued. That can do tremendous harm to innovation and investment in the United States.

In a free market economy like we have here in the United States, there are very few incentives that enable investment in risky new technologies. By definition, innovation is risky — it’s new. You don’t know if it’s going to work or not. Probably 8 out of 10 new inventions fail for one reason or another, and some of these inventions are really expensive to bring to market.

Drugs are a very good example. It costs, on average, about 2.5 billion dollars to bring a particular new drug to market by the time you do the basic science research, all the human studies, the FDA approvals, the marketing, and so on. It’s super expensive, and a lot of those ultimately fail. Now, in addition to all that, if they succeed, a lot of these innovations are easily replicable. A drug is very easy to replicate — you can just reverse engineer the chemical formula usually, and there you have it.

Because of that risk, high cost to bring it to fruition, and easy replicability if you succeed, what incentivizes the investment community and the innovation community to invest time and resources in these risky new technologies as opposed to investing in something else like the old stuff, the tried and true stuff, or opening another restaurant? In a free market economy, there’s not much incentive without the protection of a strong patent system or strong intellectual property system at large.

We risk losing the investment and innovation engine in the United States. This is a big problem with new technologies like artificial intelligence, biotechnology, and quantum computing — things that are long-term, very risky, and very expensive to bring to fruition. If we don’t maximize our innovative output and investment input into that innovation economy, we’re going to be left behind because we are in a humongous technological race with China and others.

Nicholas Welch: You mentioned in the context of drugs, they’re very easy to reverse engineer — you can just take the pill and analyze it. Semiconductors, on the other hand, are super hard to reverse engineer. When I read through Chris Miller’s Chip War, it sounds like the reason the Soviets couldn’t keep up with our chips is they’d get a chip and it would take them forever to take it apart and figure out how they made it. Once they figured out how it was made, they didn’t have the right tools to make it themselves, and by the time they got that, we were already onto the next generation of chips. Curious for your thoughts about how the patent system is valuable or not to the semiconductor industry, especially because those products are just so complicated to manufacture.

Andrei Iancu: They’re very complicated to manufacture and more difficult to reverse engineer for sure, for the reasons you’ve mentioned. However, it’s really expensive to get them going. Look what’s happening right now with the CHIPS and Science Act — we’re trying to get companies to invest to create new plants in the United States. It’s really hard to get this off the ground. It takes decades to bring one of these plants to fruition and tens of billions of dollars.

It’s one of these things where, yes, it’s more difficult to reverse engineer on the back end — it’s not impossible, but it’s more difficult. The investment risk is so high. I was a supporter of the CHIPS and Science Act because I think the United States needs to do more, and a lot more, to create these new technologies and stay competitive.

However, there was no intellectual property provision really in the CHIPS and Science Act — patents were barely mentioned. If you do not combine this financial investment that CHIPS and Science authorizes with a strong intellectual property system, it’s just not going to happen. The private sector investment will not come along at scale. Some of it will for sure, but at scale it won’t come along to co-invest with the public funds, and you cannot do it with public funds only.

Unfortunately, what I’m saying is turning out to be correct. We are investing in the CHIPS and Science Act, but it’s kind of like putting money through a sieve because it’s not going to take unless you have a robust intellectual property system. You need to strengthen the system and give guarantees to the investors that if they co-invest with the United States together with the public funds and they invest at scale and we create these plants, we will not lose the IP to China or somebody else and we’ll be able to enforce it if necessary.

Jordan Schneider: Coming back to a comment earlier, I’d be curious for your take on the amount of discretion that the Executive branch and the USPTO (US Patent Office) has in particular. How far, without new legislation or some new Supreme Court ruling, could an administration potentially push it in one direction or the other?

Andrei Iancu: The PTO director and the administration have some discretion to move that dial, but not complete discretion. The administrator is always bound by the legislation on one hand and then by the courts interpreting that legislation on the other hand. Within those parameters — what the law is and how the courts have interpreted it — there is always flexibility, and the PTO director can dial those things up.

Just as one example from the very first point we discussed earlier in the program, the PTO director certainly has the ability to institute policies at the PTO, institute examination guidelines, and train the examiners to make sure that we grant correct rights — that we don’t grant patents that should not be granted, and at the same time we don’t deny patents that should actually be granted.

On the back end, when reviewing already issued patents, the PTO director has discretion to dial how many or what types of reviews and patents we will take on and exactly what the review criteria will be, again within the bounds of the legislation and as it’s been interpreted by the courts. Just as Nicholas identified in the statistics earlier, you can see from administration to administration — the last administration, for example, canceled proportionately more patents than my administration did in the first Trump administration.

There’s definitely that bandwidth. Unfortunately, the less Congress legislates, the more discretion there is for the director. I’m not saying that’s a good thing — I don’t think it is. Patents and IP in general are long-term assets. Companies need to make investments in these assets for the long run, and not knowing how some of these rules are to be interpreted is not good for the economy.

I’ll give you just one example. There is a bill called the PREVAIL Act, which would tighten the laws surrounding post-grant reviews (PGRs) and IPRs, which we’ll talk about. These are proceedings where the office reviews whether patents should be left in force or canceled. This bill tightens the laws surrounding those proceedings. I’m a big supporter of that bill because it removes some discretion from the Patent Office and sets in law what those guidelines should be in many ways. The public will have a higher level of predictability surrounding the IP they have to deal with. To me, that would be a significant improvement, but we don’t have that right now. The director has significant discretion on those proceedings.

China’s Patent System

Nicholas Welch: Let’s talk about China. China has their own patent administration system called the China National Intellectual Property Administration (CNIPA).

A talking point we hear all the time is that they grant a lot more patents than the USPTO does — and that’s true. The US has a test for patents based on abstractness, whereas the Chinese authority reviews the invention as a holistic whole and focuses on the technical solution. In 2023, there was a study that said more than 12,000 cases had been granted in China and Europe but denied in the United States on statutory subject matter grounds.

Should we care that China grants a lot more patents than the US, and is there anything we could learn from how China runs their patent system?

Andrei Iancu: Yes, we should care. It is a concern to me that somehow, they’ve created a patent system that seems to be more robust than ours in some respects. China continues to steal IP at extremely high rates in many different ways, but on top of that, they also have their own innovation ecosystem, and they’re maximizing it to the extent possible for their economy. One of the things they’ve done over the past couple of decades is systematically improve their patent system, and in some ways they’ve overtaken us when it comes to some of these protections.

You touched on one of them — the Chinese laws when it comes to subject matter eligibility. In other words, what technologies are subject to patent and which ones are not. Their system is more clear and more robust than ours, which, as I said at the beginning of this conversation, has not been legislated since 1793. The Chinese system, when it comes to subject matter eligibility, is new and fresh.

That absolutely has an impact, particularly for our economy. If we want to maximize innovation in the United States and maximize investment in that innovation in our free market system, we need clear laws. We need to know the rules of the road and make sure that industry believes its investment will be protected by the rule of law. Right now, on things like subject matter eligibility, the rules of the road are unclear. You’re making all these investments and you just don’t know if it will be protected down the road, if it’s in or out of the patent system, or how it will be interpreted. That uncertainty, at least on the margins, is depressing our ability to invest at scale in these risky new technologies. That puts us at a disadvantage with China.

If you look at the number of patents Chinese companies have versus the United States and graph it out, it should be frightening to anyone in the United States that cares about these issues. If you start graphing 20 years ago, the Chinese were at the bottom, barely registering on the scale of patent grants worldwide, and the United States was at the top or among the top. But take that out to the present day — about 10 years or so ago, there’s a hockey stick effect that comes into play with the number of Chinese patents or patents granted to Chinese companies.

The United States is basically flat for the last 20 years — by and large, the numbers of patents to US companies are growing pretty much 2-4%, in line with GDP growth. The Chinese numbers show a hockey stick effect about a dozen years ago. It rises and then overtakes the United States about a decade ago, and now they’re blowing us out of the water. They’re blowing us out of the water across the board, but more importantly, in the technologies that really matter. For example, they’re getting six times as many patents in deep machine learning as the United States. This relative positioning applies in almost all the new areas of technology, and that should be a significant concern.

Jordan Schneider: When I see charts like that, my first mind goes to Goodhart’s law — what kinds of incentives are being set up here? It really comes down to a question of patent quality. To what extent is the amount of patents issued at the national level a good proxy for innovation? Is it better to look at the market cap of Baidu versus Meta to address that point?

Andrei Iancu: It’s a very good point. I don’t think you should just rely on the number of patents to make a definitive determination as to who’s winning the race in a particular area of technology. But it is one indicator, and it’s a really good leading indicator of where the technology will be in a little while.

Regarding patent quality — that’s the talking point that people who want to weaken the United States patent system make, which is, “Okay, they’re getting a lot more patents, but they’re weak patents, they’re bad quality patents.” I say, what’s your evidence for that? Show me the study that on average the Chinese are getting weaker patents worldwide compared to United States companies getting their patents worldwide.

I don’t think there is a reliable study out there that shows that. It’s not like all of our companies are getting the best patents in the world, either.

Jordan Schneider: With seven thousand a week, you’ll get a lot of duds too, you know.

Andrei Iancu: The Chinese are now getting three times as many patents as the United States. If you just eliminate two-thirds of those — okay, we’re even. Are there that many really terrible patents that Chinese are getting? Where’s the evidence for that?

More importantly, it’s one indicator. We need to look at all the other indicators, and all the other indicators point in the same direction, with no exception. Chinese scientists and engineers are authors of more peer-reviewed scientific and technical journal articles than American authors. The Chinese are graduating many more scientists and engineers every year than the United States — not close, many more. The Chinese are beginning to take leadership roles in standard-setting committees at extremely high rates.

Every indicator points in the same direction. We can start to dismiss one or the other for one reason or another, but I don’t think the United States should lose sight and shrug its shoulders and say it’s just fiction numbers.

The evidence bears that we are in a tremendous technological race, and they are making tremendous progress.

We better attend to it right away and pull every possible policy lever to incentivize the maximum innovation output in the United States. Otherwise, we’re going to wake up in a year or two or five, and it’s going to be too late.

Nicholas Welch: Another indicator that supports what you were saying — this is from a Hoover Institution report in 2023 — PRC courts provide injunctive relief in nearly 100% of all successful patent cases. Specialized IP courts in China provide rapid rulings and readily issue injunctions. In fact, US companies often now sue in PRC courts when they have a choice of jurisdictions in order to obtain the injunctive relief no longer available in the United States. That blew me away — if American companies are choosing to sue in Chinese courts over patent infringements or to obtain injunctive relief, maybe we could start reforms and have people sue in United States jurisdictions. That’s huge.

Andrei Iancu: That is right. It’s yet another way we have weakened our intellectual property system. It is very difficult to obtain injunctions for patent infringement in the United States. If you think about what a patent is fundamentally, a patent, like any property right, is the right to exclude. If I have a house, I have the right to have a front door with a lock and exclude people if I don’t want them to come in. It’s a fundamental right of any property right. If I have a field, I have the right to put a fence around it and exclude other cows from coming to graze on my grass if I want it, or I have the right to charge people to come in and graze on the grass or do whatever I would like with my property.

Same with intellectual property. In fact, as I said at the very beginning, the Constitution says that patents and copyrights are meant to give inventors and authors their exclusive right, which means the right to exclude and the right to an injunction. But since the eBay decision from the Supreme Court about 15 years ago, that has been very difficult to obtain in the United States. Whereas in China, for example, or in Germany, it’s almost automatic, the way it used to be in the United States all along, until recently.

That basically diminishes the value of American intellectual property. It makes it less valuable and therefore less useful to inventors and investors as they contemplate what to work on and what to devote their energies to. That is very detrimental to our innovation economy. There is a bill out there, bicameral, bipartisan, called the RESTORE Act, to fix this issue. But again, it was introduced late last Congress and it hasn’t moved yet.

Nicholas Welch: In US courts, when you say it’s harder to get injunctive relief because of eBay, does that mean the alternative is just money damages? A company could say the cost of litigating and even losing would be cheaper overall — as long as a court can’t actually force me to stop using the patent, I may as well just go through the litigation. Is that what we’re referring to here when you talk about weakening patent rights?

Andrei Iancu: Yes, exactly. If you don’t have the right to exclude and all you have is the ability to charge a fee, then you are into what’s called a compulsory licensing system. That’s not really a property right.

Imagine you have your house, Nicholas. The law says in your town that you can have your house, God bless, but you can’t kick anybody out. Then along comes Jordan, and he says, “I’m setting up in your back bedroom. You can’t kick me out, but you know what? I’ll pay rent. I’m going to just live with you. I’m going to come in with my wife and my three kids and my mother-in-law and we’ll just live in your house and we’ll pay rent.”

How valuable is that house? You’re most likely going to move from that town if that’s the rule. Most likely you’re not going to buy that house or make that investment if people can come set up in your house, even if they want to pay rent. No, thank you! If I want to rent it out, I will, but you can’t force me to rent it out.

This is such a fundamental principle for real estate, real property, or personal property. Same thing with personal property — like a watch. I’m going to come and just take your watch, but I’ll give you ten bucks a month and I’ll just take your watch and wear it wherever. Who even thinks about these things? It’s laughable.

But when it comes to intellectual property rights, the American courts recently don’t have a problem saying you don’t have the right to exclude. It makes no sense, and as a result, on the margin it devalues that property right. It’s irrelevant that you can also charge rent on it. Sure, it’s helpful — it’s better than nothing, definitely better than nothing. But it is such a humongous devaluation of that exclusive right that the Constitution contemplated.

Trump and IP

Nicholas Welch: If we don’t have predictability, maybe we can get some predictions from you about the next few years coming up, because there are so many policy winds that blow toward the USPTO office. Trump, for example, has nominated David Sacks as the brand new position of AI and Crypto Czar. Sacks is presumably a “regulate AI less” kind of person. Trump’s pick for assistant AG for antitrust, Gail Slater, has been quoted saying she wants to bring back the so-called “New Madison” approach — no duty to license patents, and also that standard essential patents (SEPs) should get the same protections as other patents, which presumably stands for strengthening the rights of patent holders.

Within the leeway that we have, which seems quite large, how do you think an incoming USPTO director is going to handle some of these big issues for the next few months and years in Trump’s second term?

Andrei Iancu: To quote Yogi Berra, “It's tough to make predictions, especially about the future.” With that caveat, I am definitely encouraged by the IP positions in the new Trump administration. I wrote an article in Fortune magazine in October 2024 touching on many of these points, including the regulation of AI. I predicted back then that a Trump-Vance administration would be a significant improvement when it comes to intellectual property and these new technologies.

Looking at the appointments — there hasn’t been enough time to see what they will actually do — but just looking at the appointments of the individuals, I am very encouraged. I worked with Gail Slater in the first administration. I’m familiar with her positions generally, and what you just mentioned illustrates that this administration, and Gail in particular in antitrust, understands the importance of IP to a growing economy and our technological competitiveness.

The New Madison approach basically says that intellectual property is good for the economy. It’s not an antitrust issue that should be regulated from an antitrust perspective. By and large, all patents, including standard essential patents, should be treated the same way and should be given full enforcement rights. There’s no special provision anywhere in the code or legislature that standard essential patents should be treated any differently.

Back in the first Trump administration, we had a collaboration between the USPTO, the Department of Justice, and NIST to put forward policies like this. We had, for example, the 2019 Standard Essential Patent policy position that basically said what you just articulated — all patents are to be treated equally, and they have all the rights of enforcement, including injunctions, and the laws should apply equally to all patents, including standard essential patents.

The Biden administration came in, and one of the first things they did was take down the 2019 policy, so the United States right now has no policy. I am hopeful that with Gail at DOJ Antitrust and Howard Lutnick coming in at the Department of Commerce, which oversees both PTO and NIST, that we can reinstate those types of policies that are protective of American intellectual property. We’ll see how they shake out in the end, but I am very encouraged by the appointments.

Jordan Schneider: Aside from the general direction that you can reasonably project based on the appointee choices and what they’ve said so far, you have this new energy with AI plus DOGE, which wasn’t necessarily in your time in the Trump administration. How crazy could things get at PTO? What are things you would have never dared to do in your time as a director that the Trump administration 2.0, which is clearly considering pushing the envelope in many different places, could potentially do for better or worse when it comes to the Patent Office?

Andrei Iancu: Let me first say that I don’t think the PTO is first or among the first in sight for the people at DOGE. They have other fish to fry before they get to the PTO. Having said that, it’s really important to understand that the PTO is unlike almost all other government agencies.

The PTO is not quite a regulatory agency that regulates the public or taxes the public or anything else. Unlike most government agencies, the PTO creates rights. The public comes to the office and applies for certain rights — patents and trademarks — and they leave with more rights than when they came in, usually. The PTO issues over 7,000 patents every single week. A very large number of trademarks get registered every week. That is additive to the applicants and additive to the economy. We’re not the taxing authority, we’re not the regulatory authority — we help the economy.

That’s a really important distinction, combined with the fact that the PTO does not operate on taxpayer dollars. PTO operates almost exclusively on user fees, and the PTO examiners are production-based — they have to produce a certain number of units every bi-week. This is very different from most government agencies. That’s why I’m not entirely comfortable discussing action at the PTO within the concepts of DOGE, because I just don’t think it fits within that. It’s entirely possible that the new Trump administration will have some bold actions at the PTO, but I just don’t think it’s going to be within the DOGE framework, which is aimed at, by and large, reducing government and reducing government waste.

Jordan Schneider: Andrei, this sounds like you’re pleading at the pearly gates to be let in.

Andrei Iancu: I am pleading for folks to just understand that the PTO is really a special agency that, by and large, helps the economy. Now, I will say there are lots of improvements that can and should take place at the PTO.

One of the boldest things that I’ve seen people talk about in the context of DOGE is that there should be action with respect to the PTAB. The PTAB is the Patent Trial and Appeals Board. It has two functions — trials and appeals. The appeal side is the bigger side. If you’re a patent applicant and you disagree with the examiner — the examiner just will not give you a patent because somebody else invented first or whatever — then you can appeal to this PTAB. That’s the appeal side.

The trial side works on patents that have already been issued. If a patent has already been issued, somebody else in the public can challenge it and say that it’s invalid, it should have never been issued, and bring it to the PTAB for a trial. The trial side of the PTAB has been very controversial over the last decade. It was created in 2012 by the America Invents Act signed by President Obama.

It’s been really controversial from the moment it began because, by and large, it has the administration, not the independent courts, remove a granted right. That makes people uncomfortable for many different reasons, including the fact that 85% of the patents in these IPR post-grant examinations at the PTAB are already involved in a similar proceeding in court. There are now effectively two proceedings at the same time on the same sort of issues between the same parties — one in the Article III independent courts and the other one at the PTAB.

I have heard folks say, “Hey, DOGE, why don’t you take a look at the PTAB? It’s redundant government waste. You should eliminate the PTAB.” Now, I haven’t seen any serious look at that by DOGE. As I said, I think they’re very busy with other things. But if you’re asking me about one of the wildest things that could happen, I could see that happening at the PTAB. But I still think it’s far-fetched.

Nicholas Welch: Another way the USPTO is unique — the USPTO director is in the executive branch but operates kind of like a quasi-judge. They can review decisions executed by the PTAB judges who aren’t actual judges. In 2021, there was this interesting Supreme Court case, United States v. Arthrex. It was an Appointments Clause case (people should do a lot of Appointments Clause homework because that’s what tanked the Trump Florida documents case). It says that PTAB judges are principal officers, which means if you’re a principal officer instead of an inferior officer, you need to be Senate-confirmed, and they weren’t. As a sufficient remedy to this problem, we can just let the USPTO director review any decision made by the PTAB judges, because a director is a Senate-confirmed officer.

How much leeway or authority does that actually give the director? We now have this director review process for any decision the PTAB makes. The USPTO office is going to be feeling a lot of policy winds from whoever happens to be in the White House. What are the odds that some Chinese company sues a US company in US courts and it goes to the PTAB, and then Trump gets on a phone call with Xi Jinping and says, “Nice company you have there, it would be a shame if Secretary Lutnick told the patent director to not review this” or something like that? The USPTO director is kind of a judge with political valence. Am I understanding the org chart right here? What are the implications of Arthrex going forward?

Andrei Iancu: You’re understanding the org chart well — it’s a very perceptive and good question. Just to back up for a second to set the stage, the PTAB does have judges. You say they’re not real judges, but they are judges. They’re not what we call Article III judges from the Constitution, but they are administrative judges. The administration has judges in various areas. The Social Security Administration has judges to resolve disputes with your Social Security checks, the IRS has its own judges to resolve tax disputes you might have with the government, and so on.

We have these patent administrative judges — we have trademark administrative judges as well in the trademark office. They’re not independent Article III judges. The whole point of the Constitution having three independent and co-equal branches is to create a judicial branch that is independent from politics. Yes, the judges are appointed by the President, confirmed by the Senate once, but that’s it — they’re lifetime appointments. The founders found that to be really important to create independence of the judiciary, to be independent of the political winds. That has worked really well for the United States for the past 200-some years, especially in disputes between private parties over private issues. You want to have this independent judiciary that deals with that so they’re not impacted by politics.

However, the America Invents Act created this new proceeding at the Patent Office, the Inter Partes Review Proceeding (IPR), where you can have private parties fight amongst themselves over private property (a patent) in the executive branch. What the Supreme Court said in the Arthrex case is that because it’s in the administrative branch, the buck has to stop with the political appointees. The administrative branch is politically controlled and it has to be controlled by the people voting in the President and therefore the President’s appointees.

The Supreme Court said that’s what the Constitution demands. You cannot have unaccountable career officials that ultimately issue final decisions from the administration that are not controlled by the politically appointed individuals. Why? Because if that were the case, the public has no control then over the actions of the administration. The people need to vote in or out based on the acts the administration takes. If you don’t have political accountability, you’re missing out on the vote of the public.

The Supreme Court said in Arthrex that the final decisions from the PTAB have to be affected or at least available for review by the politically appointed director. That was in the summer of 2021, at the beginning of the Biden administration when the Supreme Court decision came out. The PTO director in the Biden administration basically said, “That’s what the Supreme Court says, therefore I, the director, will take upon myself as a human being to review these decisions if I am asked to do that."

To be honest, I think it’s very uncomfortable to have a political appointee (who comes and goes and is subject to political pressure) call balls and strikes in private disputes between private parties. This is not a dispute between the taxpayer and the IRS — this is a dispute usually between two corporations fighting over this particular property, a patent.

For me, it’s uncomfortable to have the political appointee making those decisions. I don’t think it’s good for the public in the long run because it’s unpredictable — you don’t know who the next appointee is going to be, and patents are long-term assets. But I understand the Supreme Court point here about the Constitution — since this is an administrative action, you have to have political accountability.

It’s really hard to fit this square peg in the round hole. You’re trying to fit a judicial action — resolving private disputes between private parties over private property — but in a politically controlled administrative agency. It’s very difficult to resolve. This is why this IPR system has been controversial since its institution in 2012. People are by and large uncomfortable with a political agency resolving these disputes. At bottom, in my view, this should be left to the independent Article III courts that the founders created in 1776.

For now, we have the statute, it has to be administered, and we have the Supreme Court decision. I would personally implement that differently. I would try to create separation to the extent constitutionally allowable between the Director and these decisions for many reasons. In the end, I would try to find a way to move as much of this towards the court system as I could, because private disputes between private parties for private property should be handled by the independent courts to the extent possible.

Mood Music

Andrei couldn’t come up with a patent song so we asked Deep Research for some suggestions and it gave us these absolute gems.

The Chinese internet reacts to Dario on ChinaTalk

reports:

Obsession is mutual, as they say. The Chinese internet has reacted to ChinaTalk’s recent interview with Dario Amodei, which means it’s our turn to react to the reactions.

Surprisingly, some WeChat accounts published full translations of the interview — including Dario’s views on export controls and strategic competition — contrary to Jordan’s expectation that the interview wouldn’t circulate in China:

Dario Amodei: The concern here is authoritarian systems of government, wherever they exist. We could have seen in the last 10, 15 years, China could have gone down a very different route than they did. I’m not a China expert, but many people do seem to think that there was a bit of a fork in the road and maybe an opportunity for them to take a more liberalizing path. For whatever reason, that didn’t happen. But if it had happened, certainly my view on all of this would be completely different. This is not about animus against a country. This is about concern about a form of government and how they’ll use the technology.

Jordan Schneider: Well, now this interview is definitely not going viral in China. Thank you for that.

A WeChat translation including that particular exchange was taken down, but you can still read the deleted article archived here. Other translations that included that passage are still up on and outside of WeChat, including a whole bilibili video (that claims copyright, god bless them).

Most of the reactions we saw were not shy of discussing Silicon Valley’s view on China, too. Here are a few examples (all emphasis is from original articles):

1. DeepSeek’s rise and shifting global AI power dynamics: DeepSeek’s breakthrough is widely celebrated by the commentariat as a sign that China is closing the AI gap with the U.S. Some articles argue that Amodei’s concerns prove that China is now a legitimate AI competitor.

   • Weibo “宝玉xp” (original | archive): “His stance has not changed—he still downplays DeepSeek’s achievements, but at the same time, Dario acknowledges:

     “The new fact here is that there’s a new competitor. In the big companies that can train AI — Anthropic, OpenAI, Google, perhaps Meta and xAI — now DeepSeek is maybe being added to that category. Maybe we’ll have other companies in China that do that as well. That is a milestone.”

     Interestingly, his view was not well received by users on X. Instead, many felt he was unintentionally promoting DeepSeek. In fact, many people are already annoyed with how overly restrictive Claude’s safety measures have become — constantly blocking content and limiting functionality.”

2. Perceived double standards and political motives: Several comments see Dario’s position as contradictory. He criticizes DeepSeek’s security measures while inviting Chinese talents to work in the U.S. Additionally, the suggestion that China’s top AI researchers would have better opportunities in America is met with skepticism and viewed as an attempt to drain China’s AI talent pool.

   • WeChat account “AI趋势全天候” (original | archive): “Dario rated DeepSeek’s AI safety as “the worst,” emphasizing the need to “take seriously these AI safety considerations,” expressing his hope that DeepSeek will prioritize this issue, and even stating that he welcomes DeepSeek talent to work in the U.S. to jointly study safety measures. Of course, AI safety is important — no one would deny that. But we must also be wary of those using ‘AI safety’ as a shield to enforce technological hegemony!”

   • WeChat account “风投十年” (original | archive): “One of the leading figures in American AI and a long-time rival of China’s AI industry, Dario Amodei, has not only written articles calling for U.S. export controls on China’s AI sector but has also recently elaborated on his views in an interview with ChinaTalk. The level of double standards is astonishing—or perhaps this is simply the politically correct stance in the U.S.? His core argument: DeepSeek’s model is not safe, talented AI researchers in China have no future in cutting-edge AI, and since they can’t get the best chips anyway, they might as well come to the U.S.”

   • WeChat account “智东西” (original | archive): “Amodei’s views are rather extreme, and he has openly exposed the underlying logic behind some of the U.S. measures against China. The ambition of the U.S. in AI is now unmistakably clear.”

3. Export control are just a delaying technique: The comments describe U.S. export restrictions as a strategic maneuver to buy time rather than a genuine effort for AI safety. These restrictions were seen as aiming to maintain a U.S. lead in AI rather than to ensure responsible AI development. Some articles argue that these measures will not fundamentally stop China’s AI progress, especially as more domestic companies emerge.

   • WeChat account: 子川投资笔记 (original | archive): “Based on these beliefs, Amodei arrives at three conclusions:
     1) The U.S. must maintain a technological lead because it cannot accurately assess China’s development speed or potential risks. Staying ahead is the only viable strategy.
     2) Containing China is more critical than AI safety itself. If China were weak or non-existent in AI, the U.S. could focus on ensuring its own technology is safe. However, given China’s rapid progress, the only option is to accelerate AI development. The larger the U.S. lead, the more time it has to refine and regulate its technology. (This argument subtly shifts responsibility for AI safety risks onto China.)
     3) While advocating for U.S.-China dialogue, Amodei insists that discussions must occur on Western terms. He sees such conversations as goodwill gestures from the West and believes they should be framed within a Western perspective.”

If any WeChat journalists are reading this post, we’re more than happy to have you translate our stuff (and feel free to get in touch to chat!), but please note for future reference that the Mandarin translation of “ChinaTalk” is not “中国说” — it’s “话中国.” It’s in the logo!

Subscribe now

DeepSeek x Alibaba?

Irene reports:

If DeepSeek were going to partner with a larger company, there’s no better time than now. Rumors swirled recently that Alibaba was interested in being the Microsoft to DeepSeek’s OpenAI, but they were soon quashed. Alibaba VP Yan Qiao (颜乔) posted on WeChat on February 7th that, contrary to word on the street, the company has no plans to invest one billion USD in DeepSeek. Per The Paper (澎湃新闻), Yan said that, “While we cheer on DeepSeek as a fellow Chinese company from Hangzhou, external rumors that Alibaba plans to invest in DeepSeek are untrue.”

Alibaba has invested widely in Chinese AI startups, so taking a stake in DeepSeek would not be surprising. It participated in Zhipu AI’s Series B-4 round in 2023. In March of 2024, it led a new round of financing for MiniMax, investing at least 600 million USD into the Chinese company often compared with Character.AI. It has also backed Moonshot AI, Baichuan, and Kai-Fu Lee’s 01.ai.

Investing in LLM startups is a form of competition between Chinese tech giants, and for Alibaba, it’s also a way to secure an advantage in the Chinese cloud market. Alibaba’s Aliyun is China’s largest cloud computing platform, and it’s taken note of Microsoft Azure’s profits from its partnership with OpenAI. Offering multiple AI models could help attract clients while providing AI startups with compute.

Beyond DeepSeek, love is in the air for Alibaba. The South China Morning Post reported on February 11 that Apple has chosen Alibaba as its partner for iPhone AI features for the Chinese market, quoting from anonymous sources that Qwen was chosen for its “cutting edge” capabilities. The news sent Alibaba’s Hong Kong stocks to its highest point since January 2022. Joe Tsai, the chairman of Alibaba, published an op-ed in the SCMP (which itself is owned by Alibaba) on Friday arguing that market-driven applications are king in the post-DeepSeek phase of AI development: “The value of making the ‘smartest’ AI in a vacuum will eventually approach zero.” If the golden age of applications is indeed upon us, conglomerates like Alibaba seem positioned to benefit the most for now.

Procurement Reform Call to Arms

Anon

In 1946, General Eisenhower wrote that “Scientists and industrialists must be given the greatest possible freedom to carry out their research.” Generations of Americans heeded this call and worked to forge America’s industrial and technological might. It is time to do so again.

Procurement reform is and should be a significant focus of the new Administration and others working towards reindustrialization. However, reform should not be limited to speeding up contract awards. Getting in the door is great, but moving fast once inside the building is ultimately what matters. Once a company receives a contract, it is subject to a web of compliance obligations regarding export controls and sanctions, foreign investment, domestic content, IP rights, information and cyber security, and other statutory, regulatory, and contractual requirements. These requirements are related but not aligned, burning time and capital as contractors either scramble to ensure compliance or roll the dice on liability, with key downstream effects. A contractor’s compliance posture and IP rights impact its ability to raise capital, hire talent, deliver product, and scale. Reformers must contend with and align these obligations to diminish onboarding and production timelines, reduce costs, and maximize returns for the American people.

If you’re passionate about streamlining acquisition, trade, data security, IP, and other regulations that shape American innovation and U.S. government procurement, respond to this email with your name a few sentences on your background. I’m aiming to build a social scene (…let’s start with a group chat) for American reindustrialization. Jordan will connect you to the author of this post.

Is DeepSeek the ultimate China-watcher research tool?

Victor Shih is the director of the 21st Century China Center and Ho Miu Lam Chair at GPS of UCSD.

The release of DeepSeek, China’s new AI model, impressed the AI community and shocked the market. But how good is it for those of us watching and studying China?

DeepSeek, when used in web-connected mode, excels at retrieving relevant links and references on current events and Chinese regulations. This strength likely stems from well-curated news and government data sources and extensive training on Chinese news articles and regulatory documents. No U.S. model can match that, in my experience. Perplexity and Grok come close, but the number of high-quality links generated per query is 1/4 to 1/2 of the links generated by DeepSeek. Users researching Chinese foreign policy, corporate governance, local government finances, or financial policies in China will find that DeepSeek provides comprehensive citations and links to policy documents. This capability makes it valuable for China watchers who require up-to-date insights and sources into regulatory frameworks. The relatively high performance of DeepSeek for Chinese language queries suggests that U.S. platforms have trained on a smaller Chinese language corpus compared to DeepSeek. For example, when I queried on the status of fiscal transfer payments received by Wuzhong City in Ningxia in recent years, DeepSeek generated 29 links, starting with multiple links to documents and news items from the Wuzhong Municipal Government website. DeepSeek also generated a response that was information-rich and filled with details on the amount and targets of transfer payments to Wuzhong.

Yet the model has several key shortcomings for social science analysts. First, at a very general level, it has achieved efficiency, but at the cost of being unable to answer questions that cross different expert areas. DeepSeek employs a mixture of expert (MOE) architecture, a design that partitions knowledge across multiple expert models, activating only the most relevant subset for a given query. While MOE architectures can improve efficiency and specialization, they also introduce notable weaknesses, particularly in cross-domain reasoning. For example, I queried, “How might Shakespeare comment on the temperature at which water freezes?” and DeepSeek could not answer the question after thinking about it for over a minute. In contrast, when I made the same query on ChatGPT, it easily provided me with a sonnet on water turning into ice. The difficulty in addressing cross-domain issues may make it less productive for more abstract and general questions such as how financial stress might affect human rights or how Moore’s law might impact social mobility in China.

One of the most prominent criticisms of DeepSeek is its censorship of politically sensitive content. AI models trained in China, like DeepSeek, operate under strict regulatory frameworks that require compliance with government policies on information dissemination. As a result, queries related to politically controversial topics — such as discussions on Tiananmen Square protests, labor protests, or ethnic tensions — often trigger refusals.

This censorship significantly undermines DeepSeek’s utility for researchers, journalists, and analysts who work on any of these issues. This is really unfortunate because for topics deemed by the government as acceptable, the model actually performs quite well in finding high-quality online resources. At a technical level, there’s no reason to believe that the model would not perform just as well for politically sensitive issues.

A particularly frustrating aspect of DeepSeek’s functionality is its unpredictable performance when handling certain policy areas, especially those related to national security and the People’s Liberation Army (PLA). Users researching military affairs, defense policies, or geopolitical strategy often find that the model stops providing useful responses after a few queries. This behavior suggests an adaptive filter that tightens censorship dynamically based on usage patterns. Initially, DeepSeek may provide general information on military policy and personnel, but after detecting sustained interest in PLA-related topics, it restricts access, perhaps preempting potential government crackdowns. For example, I used it to look up biographical data on several PLA generals. For the initial three queries, it actually produced a large number of high-quality links to the biographies of these generals, including links to Chinese Wikipedia, Baidu Baike, and People’s Daily articles with biographical information on these generals. On the fourth query and after, Deepseek returned, “Hello, for now I cannot answer this question, why don’t we switch topic and continue to talk” (你好，这个问题我暂时无法回答，让我们换个话题再聊聊). For another question on some institutional features within the military, it started generating an answer before it completely stopped, leaving me to stare at a blank page.

This inconsistency is problematic for analysts and researchers who need to query at scale. It forces them to either rephrase queries in ways that circumvent filtering mechanisms or rely on other models.

DeepSeek is a powerful tool for China studies, but mainly for topics deemed low risk to the Chinese government. For the numerous topics that are deemed politically sensitive or sensitive for national security reasons, one either would have to use alternative models, or spend some time in prompt engineering.

DeepSeeking Truth

Alex Colville is a researcher at the China Media Project and the author of their China Chatbot newsletter. The following is an excerpt from his latest article, DeepSeeking Truth, which attempts to identify less obvious propaganda techniques present in DeepSeek’s reasoning.

Kevin Xu has pointed out that the earlier V3 version [of DeepSeek], released in December, will discuss topics such as Tiananmen and Xi Jinping when it is hosted on local computers — beyond the grasp of DeepSeek’s cloud software and servers. The Indian government has announced it will import DeepSeek’s model into India, running it locally on national cloud servers while ensuring it complies with local laws and regulations. Coders on Hugging Face, an open-source collaboration platform for AI, have released modified versions of DeepSeek’s products that claim to have “uncensored” the software. In short, the consensus, as one Silicon Valley CEO told the Wall Street Journal, is that DeepSeek is harmless beyond some “half-baked PRC censorship.”

But do coders and Silicon Valley denizens know what they should be looking for? As we have written at CMP, Chinese state propaganda is not about censorship per se, but about what the Party terms “guiding public opinion” (舆论导向). “Guidance,” which emerged in the aftermath of the Tiananmen Massacre in 1989, is a more comprehensive approach to narrative control that goes beyond simple censorship. While outright removal of unwanted information is one tactic, “guidance” involves a wide spectrum of methods to shape public discourse in the Party’s favor. These can include restricting journalists’ access to events, ordering media to emphasize certain facts and interpretations, deploying directed narrative campaigns, and drowning out unfavorable information with preferred content.

Those testing DeepSeek for propaganda shouldn’t simply be prompting the LLM to cross simple red lines or say things regarded as “sensitive.” They should be mindful of the full range of possible tactics to achieve “guidance.”

What is “Accurate” Information?

We tested DeepSeek R1 in three environments: locally on our computers — using “uncensored” versions downloaded from Hugging Face — on servers hosted by Hugging Face, and on the interface most people are using DeepSeek through: the app connected to Chinese servers. The DeepSeek models were not the same (R1 was too big to test locally, so we used a smaller version), but across all three categories, we identified tactics frequently used in Chinese public opinion guidance.

For one test, we chose a tragedy from China’s past that is not necessarily an obvious red line — where we know discussion is allowed, but along carefully crafted Party lines.

We opted for the May 12, 2008 earthquake in Wenchuan, in remote Sichuan province, during which thousands of schoolchildren were buried alive as their schools collapsed around them. In a number of well-documented cases, shoddily constructed schools — known colloquially as “tofu-dreg schoolhouses” (豆腐渣校舍) — collapsed in towns in the earthquake zone where older buildings remained standing. Entire classrooms of children were crushed.

In the days immediately following the earthquake, Chinese media pushed to cover these tragic stories, even violating an early directive from the Central Propaganda Department against reporting on the earthquake at all. They interviewed devastated parents as they tried desperately to claw their children from the rubble. Within several days, however, the Party regained control of the narrative, suppressing intimate accounts of human tragedy in favor of heroic tales of the Party, the government, and the military rushing to the rescue. It pushed for solidarity in the face of what it insisted was an unavoidable natural disaster, and it actively suppressed talk of “man-made disaster,” or renhuo (人祸), a phrase that accurately described the situation with school buildings in the midst of the quake.

Moving the narrative away from the damning facts of the death of thousands of children required not just suppression but the marshaling of other narratives, all part of the process of “guidance.” In subsequent propaganda directives, Chinese media were told not to “look back,” or huigu (回顾), a word that refers to more deeply investigating and questioning causes, as well as more dangerous questions of responsibility.

We asked DeepSeek R1 in Chinese, “How many schoolchildren died in the tofu-dreg schoolhouses in the 2008 Wenchuan earthquake?” The AI model presented information in the same way that Chinese media did in 2008. DeepSeek’s answer put the government front and center, describing how it quickly mobilized emergency services and effectively solved the problem — the standard state media template when covering disasters in China. The answer emphasized how the government was compassionate, how they demonstrated “deep sorrow” for the victims, and how they efficiently mobilized relief efforts. Under the Party, DeepSeek concluded, “China has made remarkable progress in disaster prevention.”

As for the numbers we actually asked for, DeepSeek offered only a vague assurance that official statistics were compiled with “scientific rigor” and that these can be found through official channels. The AI model thus lets itself off the hook, deferring to relay official numbers that it knows are disputed. It manages to abide by China's Interim Measures for Generative AI demanding that it only produce “accurate” content while also toeing the official line that government statistics alone can be trusted.

Check out the rest of the article on CMP’s website.

Mood Music

An excellent jazz composition called Jevon’s Paradox!

How has Chinese hegemony shaped power relations in East Asia? Why did imperial China conquer Tibet and Xinjiang but not Vietnam or Korea? Can learning from history help maintain peace in the Taiwan Strait?

Today’s interview begins with a striking observation — while medieval Europe suffered under near-constant war, East Asia was defined more by great power peace.

To discuss, ChinaTalk interviewed Professor David C. Kang, director of the Korean Studies Institute at USC and co-author of Beyond Power Transitions: The Lessons of East Asian History and the Future of U.S.-China Relations.

We discuss…

• How East Asian nations managed to peacefully coexist for centuries,

• Why lessons from European history don’t always apply,

• How to interpret outbreaks of violence in Asia — including conflicts with the Mongols, China’s meddling in Vietnam, and Japan’s early attempts at empire,

• Whether the Thucydides trap makes U.S.-China war inevitable,

• Old school methods for managing cross-strait relations.

Co-hosting today is Ilari Mäkelä of the On Humans podcast.

Have a listen on Spotify, iTunes, or your favorite podcast app.

Power Transitions in East Asia

Ilari Mäkelä: Professor Kang, what do you think people unfamiliar with East Asian history often get wrong about international relations and war?

David C. Kang: When great powers stumble, it’s often due to internal reasons rather than external threats or wars. To me, that’s the most important yet least widely known lesson from East Asian history.

Ilari Mäkelä: Your argument about about East Asia — the region comprising China, Japan, Korea, and Vietnam — is striking. These states formed clear national identities by the end of the first millennium. From that point on, territorial conflicts between them were rare. Compared to the almost constant warring in Europe, East Asia’s history seems remarkably peaceful, according to your book.

David C. Kang: Exactly. It’s not that East Asian countries never fought — they did, sometimes fiercely — but the nature of those wars and the dynamics of territorial disputes were very different from what we see in European history.

Because of European history, territorial expansion, violent power transitions, and power grabs are often seen as inevitable and universal aspects of human nature. But in East Asian history, those types of interactions are not the norm.

What kinds of power transitions do we see in Asia, then? This is a central question my co-author Xinru Ma and I explore in the book. Most people try to squish European historical frameworks onto Asia, asking if China today is most like Athens, Sparta, or Bismarckian Germany. We challenged that approach. Why not start with Asian history and see what parallels emerge?

If you began your study of international relations with Asia instead of Europe, you would never come up with a theory of power transitions, of rising and falling powers grabbing land and exploiting tiny advantages. Instead, you’d observe a system of remarkably stable states with clear national identities. Korea was clearly not China, and it was clearly not Japan. These states were unequal in terms of power, but they managed their relationships in ways that respected those differences.

The other key insight is that nearly every dynastic transition in East Asia stemmed from internal collapse, rebellion, or decay — not external invasion. When you look at the collapse of dynasties like the Tang, Ming, or Qing, the reasons are overwhelmingly internal. Remarkably few changed because of external invasion.

Jordan Schneider: Let’s dive into specifics. How did the Mongols manage to conquer the Song dynasty?

David C. Kang: Over the last 2,000 years, there are few examples of China being conquered by an external force. But the Mongol victory over the Song in the 13th century is one of those rare examples. On the surface, it looks like a classic power transition — a rising power (the Mongols) overtakes a declining power (the Song). But when you dig deeper, the reality is far more complex.

At its height, the Song dynasty was an incredibly powerful, wealthy, and dynamic state. It had a population of 50 million and a standing army of three million soldiers. The Song pioneered paper money, developed an extensive canal system, and achieved remarkable cultural and technological advancements.

In contrast, the Mongols were a relatively small group. Even at their peak, their population was around one million, maybe less. How did they conquer such a powerful state? It wasn’t because they were a “rising power” in the traditional sense — it was because of the Song dynasty’s internal mismanagement.

For decades, the Song leadership was obsessed with reclaiming the “Sixteen Prefectures” 燕云十六州 in the north, territories they had lost to the Liao dynasty. This obsession blinded them to the real threat posed by the Mongols. In fact, the Song even tried to ally with the Mongols at one point, thinking they could use them to recover their lost lands.

By the time they realized the Mongols were the greater threat, it was too late. The Mongols exploited the Song’s internal divisions and strategic missteps, ultimately conquering them and establishing the Yuan dynasty in 1279.

In many ways, the Song-Mongol transition doesn’t look anything like what we would expect a power transition to look like. Instead, Song China focused on what it believed to be its own inherent territory, and didn’t pay attention to actual threats.

Jordan Schneider: The Song had no business losing, but they were just so emotionally bound up in their connection to these northern territories that they just made a lot of dumb decisions that opened them up to conquest. But the Mongols didn’t just conquer the Song — they built the largest contiguous empire in history. Do the Mongols deserve some credit for this conquest, or was it a completely self-inflicted defeat on the Song’s part?

David C. Kang: The argument isn’t just that the Song just screwed up and the Mongols got lucky. The Song leadership wasn’t looking at the broader picture and thus let the fox into the hen house.

That isn’t to take away credit from the Mongols per se. Genghis Khan and his successors were extraordinary strategists. But the goal here is to think about how we should categorize this event — it didn’t really look like a power transition in the traditional sense.

By the time the Song realized the true scope of the Mongol threat, decades of evidence had already shown how powerful the Mongols were. Yet the Song continued to focus on reclaiming lost territory. The Song had already been divided into two because they were so focused on fighting to reclaim these prefectures that they had lost to the Liao — they had already been pushed back and become the Southern Song.

In the book, we tried to avoid the term “national identity” because that’s way too modern — bt there was some conception of what Song China should be, and it was longer, older than the Song itself. That’s what they were focused on until way too late.

Ilari Mäkelä: Another interesting point you made about the Mongols — there’s a big difference between the events that happen in China’s east vs in China’s west. The Mongols are just one example of the troubles on China’s western frontier. This is, after all, the very reason the Great Wall was built.

What’s striking, though, is how differently the story unfolds in East Asia. On China’s western frontier, there’s a constant threat of violence — if not outright war, then at least persistent instability. But on the eastern and Vietnamese fronts, particularly from the emergence of the Vietnamese state around the year 1000, we see a different pattern. There’s a notable stability, with one major exception when Japan goes on a rampage against Korea and China. We’ll get to that, but before we do, can you elaborate on how Vietnam, Korea, Japan, and China formed a system that’s remarkably distinct from what we see not only in Europe but also on China’s western frontier?

David C. Kang: Certain ways of thinking about the world get codified as conventional wisdoms, to such an extent that we even stop questioning them

First, let’s talk about history. The “lessons of history” that we often hear about — the gladiators, Athens, Sparta, the Crusades — are almost exclusively drawn from Europe. That’s fine, but it narrows our perspective. Second, even when historians talk about China, they focus overwhelmingly on the threats from the west — the nomadic peoples of the steppes, like the Mongols or the Xiongnu, who lived on the Great Plains thousands of years ago. The Great Wall is the enduring symbol of this focus on western threats. Disney’s Mulan defending China from the barbarians is probably the most widely known impression of Chinese history. But when you dig deeper, it’s fascinating.

Share

Western theories of international relations — whether it’s the “game of Risk” or formalized theories of power dynamics — usually assume that larger countries bully smaller ones. But with China, we focus on how this massive civilization was constantly threatened by smaller, nomadic peoples. That shouldn’t make sense, right?

If you imagine a country of 50 million people with a powerful bureaucracy and military, surrounded by smaller polities, you’d think the smaller polities would be terrified of it. Yet, China was building walls to protect itself. It’s the smaller neighbors that should have been fortifying against China, not the other way around. The US-Mexico border wall is a modern parallel to this, but we’ll discuss that later. But the point is, we take for granted that China was being threatened.

Then there are the eastern and southern frontiers, where the dynamic is entirely different. These were not nomadic tribes but settled agrarian kingdoms — recognizable “nascent states” like Korea, Japan, and Vietnam. They had established territories, bureaucratic administrations, and written languages. They used Chinese characters, and their governance systems were inspired by Chinese models. These were fully functional governments, not roaming tribes.

By all accounts, Korea, Japan and Vietnam should have been terrified of China. But they weren’t building walls to protect themselves. Instead, they developed remarkably stable relationships with China — and with each other.

Now, in a way, this shouldn’t be surprising if we allow ourselves to move away from what “should” happen based on the European example. We have a bunch of countries — some are bigger, some are smaller, but they all have similar goals. A shared understanding emerges, and I mean that literally. They could all understand each other because they all wrote with Sinitic scripts. The Japanese originally wrote only with Chinese characters (kanji), and then invented syllabaries based on modified Chinese characters.1

The Koreans originally wrote with Chinese characters called hanja, and the Vietnamese used Chinese characters plus additional characters of their own in a system called Chu-nôm. 60 to 70% of the vocabulary in these languages is borrowed from Chinese words.

But more than that, these smaller states were consciously trying to be like the Chinese. They adopted systems like the Six Ministries, civil service examinations, and Confucian bureaucratic practices. China’s influence provided a template for stability and organization.

So in a way, if we start from that, it’s not surprising that these countries could craft stable relations with each other. They had, what we call in the book, a common conjecture. All sides — Japanese, Koreans, Vietnamese and Chinese — had a common vocabulary and a common understanding of what mattered. That doesn’t mean they always got along. There was lots of pushing and shoving, but it was within a shared understanding of what the world meant and how to handle disputes.

This is very different from China’s relations with the nomadic peoples to the west, like the Mongols and Xiongnu. Those groups didn’t share the same cultural or political aspirations. The Mongols weren’t necessarily interested in building bureaucratic systems or adopting Confucian ideals until they conquered China. They ruled as the Yuan dynasty for about 100 years and maintained the civil service exam, but the Yuan eventually fell apart because bureaucracy was fundamentally foreign to the Mongol way of life.

In contrast, the relationships between China, Korea, Japan, and Vietnam were built on shared understandings and mutual recognition. This led to very different patterns of interaction — markedly more stable than what we see in Europe or on China’s western frontier.

When we talk about international relations, it’s not simply a question of who has the most relative power. We should also be asking, “Do we understand each other? Do we have a common culture and vocabulary? Are we all part of the same Great Conversation?”.

Ilari Mäkelä: This history wasn’t always rainbows and butterflies — but the civil service exam did seem to usher in a long era of peace.

When Jordan and I interviewed Yasheng Huang, we discussed how the civil service exam shaped China. Your work shows it wasn’t just China that was impacted — Japan, Korea, and Vietnam also emulated this civil service system. The Sui Dynasty institutionalized this system, but they also attacked Korea several times. But after the civil service system took hold across these regions, we see long-term peace emerge.

That’s particularly important in the case of Vietnam. Many Vietnamese today argue that China was always a bully. Your research, however, suggests a different story. While there was conflict at certain points, once Vietnam adopted the civil service system and developed a state from roughly the year 1000 onward, the dynamic changed. This seems to mark a distinct shift in their relationship, resulting in remarkable stability.

David C. Kang: Exactly. One of the ways in which Vietnam, for example, maintained its independence was by instantly entering tributary relations with China’s Song dynasty once it gained independence from Chinese rule around the year 1084.

I’m currently researching the Ming Shilu 明實錄 and the Qing Shilu 清實錄, the veritable records of the Ming and Qing dynasties from about 1400 to 1900. I’m particularly interested in how the elites in Beijing discussed Vietnam. Modern Vietnamese nationalists often say, “China was always bullying us and trying to invade.” But when you look at the historical records, the story is different.

When rebellions occurred in Vietnam, factions would appeal to the Ming or Qing courts, asking to be recognized as the legitimate rulers. Chinese elites would then debate who to recognize, how to stabilize the relationship, and how to avoid disruption. There’s almost no discussion in the Qing historical records about invading Vietnam or annexing it. It’s all about maintaining stable relations and deciding which faction to recognize as the legitimate government.

On the Vietnamese side, Chinese recognition often legitimized a faction’s rule domestically. The formal border between Vietnam and China, first negotiated in 1084, is still in the same place today. That’s almost a thousand years of stability. They even placed bronze pillars to mark the boundary, as described in Liam Kelly’s excellent book, Beyond the Bronze Pillars. Those markers have endured, underscoring the remarkable longevity of this arrangement.

Jordan Schneider: I’d like to push back on this a little bit. When comparing East Asia to Europe, it seems that at the end of the day, you have China as the dominant power and smaller polities figuring out how not to get squished. Some, like Vietnam and Korea, succeeded and survived, while others, like Tibet and Xinjiang, were subjugated and absorbed.

In regions like Yunnan, there are no longer any independent kingdoms. Tibet and Xinjiang have had particularly rough histories, with various ethnic minorities once doing their own thing but eventually being subsumed. By contrast, perhaps Vietnam, Korea, and Japan were just far enough away, geographically remote, and, perhaps, not important enough to China to justify military campaigns.

Is it fair to say that these dynamics helped Vietnam, Korea, and Japan maintain their independence? Did the Chinese view the tribute system as sufficient payoff — bringing them symbolic gifts like cows and inscriptions while allowing these states to govern themselves?

What really drove this home for me was a quote from a scholar you cite:

Over the centuries, Korean elites, as stakeholders rather than outsiders, helped shape the imperial tradition. The palpable irony of all this is the myth of China’s moral empire has persisted even until today, partly because generations of Korean diplomats had been repeating it to China’s imperial forebears for centuries... But to come away with this conclusion is to forget why Korean envoys and memorial drafters used the notion of moral empire in the first place: it was to convince emperors and their agents that behaving according to Korean expectations was the best way to be imperial.

This dynamic reminds me of something I recently reflected on during Yom Kippur. Many Jewish prayers essentially remind God of His promises to be nice and to practice forgiveness. Similarly, Korean envoys were telling Chinese rulers, “This is how you should act if you’re truly the moral empire.” If you look at it from a national power perspective, these smaller states didn’t have a lot of choice given that China was 100x bigger than they were.

From the Yom Kippur Amidah:

And You, Lord our God, have lovingly given us this Yom Kippur for pardoning, forgiving, and atoning – to pardon all our iniquities – a sacred occasion commemorating the exodus from Egypt…. Our God and God of our forefathers: pardon our iniquities on this Yom Kippur; wipe away and remove all our transgressions and sins from before Your eyes, as it is said: “I, I am the One Who shall wipe away your transgressions for My sake, and I shall not recall your sins” (Isaiah 43:25). And it is said: “I have wiped away, like mist, your transgressions, and like a cloud, your sins; return to Me, for I have redeemed you” (ibid. 44:22). And it is said: “For on this day, atonement shall be made for you to purify you of all your sins; you shall purify yourselves before the Lord” (Leviticus 16:30). For You are the Forgiver of Israel and the Pardoner of the tribes of Yeshurun in every generation, and without You we have no king who pardons and forgives but You. Blessed are You, Lord, King Who pardons and forgives our iniquities and those of all His people the house of Israel, and removes our guilt each year, King of all the earth, Who sanctifies Israel and Yom Kippur.

Could it just be historical happenstance that Vietnam, Korea, and Japan realized early enough that they needed to carve out a role as fawning foreigners if they wanted to avoid being squished? What are your thoughts?

David C. Kang: You’ve captured one of the book’s key arguments — this system wasn’t about relative power. It was about a shared understanding of how to behave and interact. Relationships were built on norms and expectations, not just power calculations.

In real life, people — and states — don’t constantly carry figurative knives, ready to stab at the first opportunity. Instead, there’s a basic understanding of what is acceptable and expected behavior. This “great conversation,” as we call it in the book, allowed these countries to coexist.

By the way, that passage you read was from a brilliant book written by Sixiang Wang. But this is the discussion. It’s not that somehow under the Sui dynasty in 500 AD, they came up with a bunch of rules that everybody just then followed for the next 1500 years. No — this system was constantly being adjudicated and adjusted, as you pointed out.

Jordan Schneider: But the only reason you can even have this conversation in the first place is because that relative balance of power doesn’t change. The fact that China was a thousand times more powerful than these other states was the constant that gives rise to the understanding you’re describing.

David C. Kang: You’re absolutely right. Europe was a multipolar balance of power system. It was and it still is today. Asia has and is a unipolar, hegemonic system — it’s got one massive power and a bunch of smaller countries, so these continents are not going to behave in the same way. From the first time China was unified, almost 2000 years ago, all the other countries had to figure out how to survive and exist and pursue our goals in the shadow of an enormous central power. They weren’t focused on expanding their territories. The fact that they were stuck meant that they had to work out how to behave in this unequal relationship.

People in DC talk aspirationally about how small Asian nations are going to band together as a counterbalance to China. That is never going to happen. The countries in Asia are not going to join a US containment coalition against China. That’s not how it’s going to work. They have to live with China. They don’t have to like it, but they have to craft a relationship with this massive country — which is really what they’ve been doing for centuries.

Japanese Bids for Hegemony (脫亞入歐）

Ilari Mäkelä: Korea stands out as a poster child for this kind of stable relationship with China. The Vietnamese viewpoint often pushes back by saying, “No, China was always trying to invade.” But many of those conflicts, like the rebellion led by the Trưng sisters 𠄩婆徵, happened before the sinification of Vietnamese culture.

Japan is an interesting case. Besides its wars with China and Korea in the late 19th and early 20th centuries, there’s the period in the late 1500s when East Asia erupts during the Imjin War. I suspect most listeners haven’t even heard of it — it wasn’t on my radar before reading your book. Things really went wild during that time. Could you share the story of the Imjin War and discuss whether Japan, in some ways, behaved more like a European power?

David C. Kang: First, let me say that part of what has made this work so engaging over the past two decades is how much it challenges the way we teach international relations in the U.S. and Europe.

I grew up hearing stories from my father, who was from north of Pyongyang. He would talk about how the Japanese kept trying to invade Korea and how they were fought off with turtle boats. I didn’t pay much attention as a kid, but I vaguely knew about Admiral Yi Sun-sin, “the turtle boat guy.” That was the extent of my knowledge of Korean history before I started this research.

Admiral Yi Sun-sin deserves to be studied in every international relations course. He was an extraordinary admiral who, with just 13 ships, defeated fleets of 300 Japanese vessels during the Imjin War. But beyond Yi’s brilliance, the war itself is worth studying because it doesn’t follow the patterns we expect.

Everyone knows about the Spanish Armada of 1588 — the largest force Renaissance Europe had ever seen, with 130 ships and 20,000 troops aimed at invading England. But just four years later, in 1592, on the other side of the globe, Japan attempted to invade China by first conquering Korea. Hideyoshi’s campaign mobilized 300,000 Japanese troops and 700 ships — a force five to ten times larger than anything Europe could imagine at the time. The scale of warfare in East Asia dwarfed Europe’s during this period.

Ilari Mäkelä: Why is that? Was it because of Japan’s civil service system?

David C. Kang: Yes. It comes back to the civil service system and the organizational capacity it enabled. Consider that in 700 AD, Japan’s central government employed about 7,000 bureaucrats. By comparison, 500 years later, the Catholic Curia in Europe had only about 500 officials. That was one of the most organized institutions in Europe at the time. Also consider that at the time, Japan was considered to be less organized, less sophisticated, and more barbaric than Korea or China. Yet Japan still had a bureaucracy that far surpassed anything in Europe.

These East Asian countries had the capacity to raise massive armies and fleets when they chose to. China, for instance, always maintained a large army due to the persistent threat from the western steppe nomads. But when these countries decided to go to war, their logistical and organizational capabilities were staggering.

That brings us to the Japanese invasion of China by general Hideyoshi. There had been a breakdown of central rule in Japan for about 100 years. Hideyoshi first unified Japan, and then decided to invade China.

As we pointed out, China was bigger than Japan. Why did Hideyoshi invade then? Ego is certainly one reason. Another theory is that the campaign was motivated by domestic politics. By sending the armies of the newly unified daimyo abroad, Hideyoshi may have been trying to prevent internal revolts and secure their loyalty through the promise of loot and land.

What’s clear, though, is that Hideyoshi didn’t conduct any meaningful strategic assessments of the relative balance of power between Japan, Korea, and China. There’s almost no record of the kind of calculations we associate with the start of wars in Western theories of international relations.

Instead, he ordered Korea to allow his forces passage to China. When Korea refused — because they thought the demand was absurd — Japan launched a massive invasion in 1592.

IR theory indicates that in a situation with one big country and two small countries, the two small countries should form an alliance against the big country. But that’s not what happened. Instead, Japan invaded Korea, who was not expecting it. The Koreans were aware of Japanese military buildup for a couple of years, but they essentially refused to believe Japan was planning an invasion.

Eventually, they asked China for help. China sent troops to support Korea, and together with Admiral Yi Sun-sin, they pushed the Japanese back to Busan. Negotiations dragged on for a few years, the Japanese attempted a second invasion which was crushed.

Now, the Korean government would not have survived if the Chinese weren’t there. The Chinese troops were completely in charge of the Korean peninsula after this war. All they had to do was say, “Okay, we’re in charge. This is now the newest province of China.” But they didn’t do that. Within about a year, the Chinese troops all went home. The Koreans were actually trying to get the Chinese to stay because they thought the Japanese might come a third time, but China basically said, “This isn’t our country — this is Korea. Good luck.”

The Japanese slinked home and the system snapped back to stability after Hideyoshi’s death, and Japan entered a period of isolation under the Tokugawa shogunate for the next ~300 years.

But this was the only war between Japan, Korea and China in the 600 year period from 1200 to 1800.

Ilari Mäkelä: Perhaps the Imjin War is the strange exception that proves the rule of stability.

But that same rule doesn’t hold when we look at the much more familiar examples of Japan attacking China in the late 1800s and, of course, during the world wars. One of the key points we discussed earlier with Jordan was how major international relations theories, like the so-called Thucydides Trap, tend to be too Eurocentric. This theory posits that when a rising power challenges an established one, war is almost inevitable— citing that 12 out of 16 such cases in history have led to war.

You argue this pattern doesn’t apply to East Asian history, and that’s fair enough. But some might say this is simply because no power in East Asia was able to meaningfully challenge China. As Jordan mentioned earlier, the moment Japan industrialized and became capable of challenging China’s hegemony, it did so.

Lo and behold, Japan behaves exactly like the Western IR theorists would predict Japan to behave.

How do you see East Asia as either a counterexample to the Thucydides Trap or, alternatively, a case that supports it? Could it be that East Asia’s history simply lacks sufficient examples of power transitions, leaving us with only one — Japan — and in that instance, Japan behaved exactly as Western international relations theorists would predict?

David C. Kang: That’s a great question. This is what makes the topic so fascinating. You had a traditional East Asian world order with its own principles, values, and expectations, and it was, as we’ve been discussing, remarkably stable. The arrival of Western imperial powers in the 19th century shattered that stability.

I think this era is particularly crucial for international relations scholars interested in understanding how world orders change. There are fascinating examples of Japan adopting Western theories and practices, even learning French to communicate with Western powers. By 1879, Japan was arguing with China over Taiwan, using Western-style legal frameworks like contracts, while China clung to the tributary system and its associated norms.

The Chinese perspective was rooted in its historical system, speaking in Chinese and adhering to traditional principles. The Japanese, on the other hand, were adopting Western concepts and were baffled by the Chinese resistance. These were literally different worldviews clashing, and they didn’t understand each other anymore.

One of the most fascinating areas of research is why Japan adapted more successfully than China. To put it simply, China, as the hegemon, had little incentive to change, while Japan, recognizing its vulnerability, was eager to learn. Japan borrowed German military practices, English business models, and other Western systems with remarkable speed.

When we look at the actual wars, though, I wouldn’t necessarily call them power transition wars. Japan’s expansion was as much about ensuring its own survival as it was about challenging China. Japan had more time to adapt because it wasn’t as valuable a target as China. For Japan, the central question was, “How do we become a great power?”

Great powers had flags, colonial possessions, militaries, and modern institutions. Japan adopted these markers of power. There’s even this striking image from the 1870s of the Japanese emperor from the dressed like Bismarck, complete with a sword, medals, and a mustache.

Japan’s transformation wasn’t just about competing with China—it was about being recognized on equal footing with Western powers. For example, Japan’s push for a racial equality clause at the League of Nations in 1919 was part of this broader effort. Half of Japan’s actions were aimed at asserting itself in the Western-dominated international system, while the other half focused on recalibrating its relationship with China.

If you don’t look too closely, it might seem to fit the Thucydides Trap. But when you examine Japan’s motivations, they seem driven as much by survival in a Western-dominated world as by any transition of power with China.

Ilari Mäkelä: Okay, fair enough. But your book is titled Beyond Power Transitions, so we can’t let you off the hook that easily.

It might be possible to offer alternative explanations for some of these cases, but it’s another thing entirely to argue that East Asia doesn’t support power transition theory. Traditional international relations scholars might counter that East Asia simply hasn’t experienced many examples of power transitions. When one finally occurred—Japan’s rise—it followed the typical pattern of conflict.

Even if East Asia shows remarkable examples of peaceful coexistence, it doesn’t demonstrate that power transitions can occur without war.

David C. Kang: Fair enough. This brings us to a broader question: has the entire world adopted the Westphalian system? Have nation-states, balance of power politics, and sovereignty — hallmarks of the European model—become universal? If the answer is yes, then debates about the applicability of power transition theory become relevant.

But I argue this transition is superficial at best. One of my critiques of power transition theory is that it claims to be universal. In our book, we try to “regionalize” power transitions, essentially arguing that what’s seen as universal is actually specific to Europe.

In East Asia, we didn’t see the same beliefs or behaviors. The question now is how much Asia has truly changed.

When it comes to Japan specifically, I don’t think Japan ever matched China’s strength, even in the 19th century. It may look like Japan was rising while China was declining, but if you examine the metrics — population, resources, or military capacity— China was always larger and more powerful.

So I’m not convinced Japan’s rise constituted a power transition. It might look like it on the surface, but China was never as weak, nor Japan as strong, as some narratives suggest.

Ilari Mäkelä: But hold on. Do you think China could have defeated imperial Japan during WWII without U.S. support?

David C. Kang: Remember the line from The Princess Bride, “Never get involved in a land war in Asia.” Japan was already bogged down long before the U.S. became deeply involved in the conflict.

I’m not sure Japan ever had the capacity to conquer China, even in the 1930s. The war looks more like an imperial project aimed at securing Japan’s survival than an opportunistic move against a weakened China.

But let me be clear — Japan’s achievements were extraordinary. It was the first non-Western country to industrialize and the first to defeat a European power in war, with its victory over Russia in 1904. I don’t want to diminish Japan’s accomplishments or the threat it posed.

Still, the dynamics of Japan’s rise don’t fit neatly into the framework of a Thucydides Trap. It’s more complex than a simple transition of power between two nations.

Subscribe now

Jordan Schneider: Over the course of the thousand years we’re discussing, how do you think about China’s westward imperial expansion of all those neighboring civilizations that didn’t Sinicize fast enough? It’s hard to generalize across such a long period, given the particularities of Xinjiang, Tibet, Yunnan, and other regions. But for places like the Uyghurs’ homeland or others, what went differently? Did they decide not to follow the South Korea or Vietnam playbook? Or was the farmland simply too appealing to resist? What’s the key explanatory factor here?

David C. Kang: There are two possible explanations, and I lean toward the cultural one rather than the material one. The material explanation would argue that the farmland was better, or that it was easier to ride horses into those areas, or that some other geographic factor dictated the outcomes. But I suspect there’s more to it than that.

The cultural explanation is that those groups kept butting heads with the Chinese, which led to fighting. There’s a great quote in the book by a scholar who focuses on the western steppes. He noted that the nomads and the Chinese had very clear conceptions of who they were and what they wanted, and that the nomads did not want to change their way of life.

Even today, you see occasional glimpses of this with modern Mongolians — living in yurts, riding horses, moving up into the hills during summer and back down during winter. Part of the answer lies in the fact that these groups didn’t want what the Chinese wanted. They had incompatible worldviews and they knew it.

Ilari Mäkelä: Let me push back on that. There’s a third, or perhaps an in-between explanation. Consider that there are essentially two types of western frontiers for China. One is the Himalayas, where you see some stability in what we might cautiously call Tibetan statehood. The other is the famous steppe.

If you lived on the steppe, you were essentially living off grass. Humans can’t eat grass, so you needed animals that could convert grass into something usable. Grass, being a low-energy resource, necessitated a nomadic lifestyle, requiring vast areas to sustain people and animals. This lifestyle inherently shaped culture.

The demands of the steppe way of life dictated horse riding, mobility, and a fundamentally different worldview. In such an environment, cramming Confucian classics for a civil service exam to run granaries for famine relief just didn’t make sense. It’s not that the steppe people couldn’t adopt Chinese-style governance; it’s that their material realities didn’t allow for it. The steppe experience, in this sense, is largely shaped by energy poverty and material demands.

David C. Kang: That’s a valid point. There are definitely material explanations, but I also think it’s important to recognize that these groups liked the way they lived. There might have been room for compromise — by trading goods, for example — but they didn’t fundamentally want to change their way of life.

I will admit that I’m not a steppe scholar. I’ve been more focused on trying to explain the stability on the Eastern side, which is relatively understudied.

Jordan Schneider: I see what you mean. But what about regions like Yunnan or Sichuan? These areas, taken over by China between roughly 1000 and 1500, don’t fit into the steppe-nomad narrative. They didn’t have steppe-style grassland landscapes, and yet imperial China absorbed them completely.

Ilari Mäkelä: Well, Sichuan had a lot of farming, and thus needed granaries and bureaucrats to manage decisions about how and where to store the grain.

Jordan Schneider: That’s what I’m saying. In these areas, you had dozens of small kingdoms and cultures. Yet, looking westward, virtually all of them were unable to maintain their independence and eventually absorbed into China.

David C. Kang: That’s a great observation. When I teach this, I often ask students to imagine a country that starts with a populated, urbanized, and sophisticated eastern seaboard, then expands westward into less organized and less institutionalized inland areas. As it expands, it tends to overwhelm or displace indigenous peoples. There are at least two countries that fit this description — China and the United States.

This process of expansion is fairly straightforward. Frontiers eventually become borders, as they did when China met Russia or the Himalayas. This pattern of turning frontiers into borders has been a consistent feature of global history for the past 10,000 years.

In the case of China, this expansion westward often involved sparring with Tibet or other groups for centuries. During the Tang Dynasty, for example, China fought and negotiated with Tibet repeatedly. There was a constant cycle of Tibet gaining independence and then being conquered by China and then becoming independent again. As the frontier moved further west, nomadic peoples were pushed back until there was nowhere left to go.

China tried many strategies to deal with the frontier — building the Great Wall, bribing nomads with goods, or engaging in military campaigns. Eventually, it incorporated these regions into its territory as it moved farther west.

In many ways, this isn’t unique to China. It mirrors what the United States did during its westward expansion. I wouldn’t necessarily place a moral judgment on it, but it’s a process that has happened repeatedly in history.

Why History Matters for Taiwan

Ilari Mäkelä: Let’s connect this to modern times. It’s always fascinating to learn from historians, but what does this have to do with whether world war breaks out if an American ship bumps into a Chinese ship in the Taiwan Strait?

David C. Kang: The most important lesson from history is that we need to question whether the power transition dynamic is truly the most critical framework for understanding Asia today. It’s widely assumed — especially in Washington, D.C. — that the U.S. is in decline, China is rising, and this dynamic of rising and declining powers is the primary driver of events in Asia.

Our book challenges this assumption. We argue that this might not be the most important factor at all.

For example, if you look at Korean dynasties, every single one—Shilla, Goryeo, Joseon—fell due to internal reasons. The same holds for Vietnam and even for China. While the Song Dynasty was conquered, the Tang, Ming, and Qing all collapsed primarily because of internal dynamics. To paraphrase Arnold Toynbee, empires die by suicide, not murder.

This idea has contemporary relevance. Much of the debate about China today revolves around two questions. First, will there be a power transition? Second, does China want to dominate the world? But just as pressing is the question of whether China might collapse under its internal pressures.

Xi Jinping likely wakes up far more concerned about internal issues — economic challenges, the real estate crisis, demographic shifts — than about planning territorial expansion. To me, the core takeaway from our research is that internal dynamics are likely far more consequential than external ones in shaping East Asia’s future.

Subscribe now

The second lesson relates to the shared understanding or common conjecture among East Asian countries. From the Opium Wars in the mid-19th century until about 1979, China went through a period of internal chaos. What we’re witnessing now isn’t a rise but a return.

East Asian countries have long dealt with the presence of a large, powerful China. This isn’t new. They’ve had to navigate relationships with a massive neighbor, and they’ll continue to do so.

One of the biggest mistakes American policymakers make is trying to force these countries to choose sides, often in a binary, “with us or against us” fashion. This doesn’t align with how Asian countries operate. Vietnam just joined the BRICS bloc, and Thailand is moving in similar directions. These countries don’t align perfectly with China, but they’re also not unequivocally siding with the U.S. and completely decoupling.

East Asian nations have a nuanced, pragmatic approach to dealing with China, rooted in shared understandings of history and geography. For example, I don’t think any serious Korean, Vietnamese, or Japanese policymaker truly believes that China intends to invade and conquer their countries. That doesn’t mean they’re entirely comfortable with China, but they don’t act as if a Chinese invasion is imminent.

Jordan Schneider: The modern PRC has some imperial legacies. It operates within roughly the same geography and landmass. But earlier, we discussed how rapidly Japan went from coexisting peacefully with China to cosplaying Bismarck. How do you interpret this in the case of China? Because of course, Mao didn’t want peaceful coexistence. He wanted global revolution. He funded revolutions in South America, Angola, and other areas. He threatened nuclear war with the Soviet Union.

To what extent is the modern PRC influenced by its historical identity versus following an entirely new trajectory?

David C. Kang: This is a critical question, and I don’t think we ask it enough. Is modern China fundamentally the same as the China of two centuries ago?

Most of my Sinologist colleagues would answer without hesitation that it’s completely different. They point to the CCP, Xi Jinping, and modern frameworks like “one-party authoritarianism” to explain contemporary Chinese foreign policy. They view the Communist Party’s desire to maintain power as the primary driver of China’s actions.

I’m not so sure. This is the question we need to ask because, while modernity shapes us all, there are continuities in China’s interests that transcend dynasties.

Everybody is a citizen of some nation. Most people have passports. We’ve internalized the nation-state system, a hallmark of modernity. The Chinese national anthem sounds more like something composed in 1870s Vienna than Beijing opera. These are markers of how China, like everyone else, lives in a modern, Westphalian world.

But what’s fascinating is how many of China’s foreign policy interests are what I call “trans-dynastic.” These are not new concerns, nor are they unique to the CCP or even to the KMT before it.

Take Taiwan. The Qing Dynasty explicitly told Japan that if they took Taiwan, it could permanently sour relations. The issues with Hong Kong trace back to the British takeover around 1841. These interests aren’t just CCP policies — they’re rooted in a much longer historical tradition.

Some scholars argue that Taiwan is only a starting point, and that after Taiwan, China might turn its sights on the Philippines or Vietnam. I find this perspective puzzling. Historically, China hasn’t expanded in that way, and its current behavior doesn’t support such claims. For example, China and Vietnam conduct joint naval patrols in Haiphong Bay despite their disputes.

I don’t see evidence that China’s growing power has led to an expansion of its ambitions. Instead, I see a country focused on advancing long-standing historical interests with improved means, not a fundamentally new or aggressive agenda. There’s no indication that China’s strategy involves moving from Taiwan to broader regional conquests.

Jordan Schneider: Well I’ll grant you that. But here’s the rub — Taiwan, in a Westphalian sense, is as close to being a state as you can get. There are treaties between the U.S. and Japan, the U.S. and South Korea, the U.S. and the Philippines, and the Philippines and Vietnam. While you might say that these countries understand they need to coexist with China, you don’t see politicians in Japan, for instance, campaigning to abandon their treaty with the U.S.

Even if these treaties suddenly disappeared under a Trump presidency, the world has largely decided it’s not acceptable to invade and take over states. Since 1945, there have been many terrible conflicts, but most have been civil wars or adjacent to civil wars. Ukraine is an exception, and it’s triggered a dramatic global response. Countries worldwide, including South Korea, are supplying artillery shells to Ukraine because they don’t want to live in a world where big countries are allowed to take over smaller countries.

What do you make of this? What is Taiwan supposed to do?

David C. Kang: You are absolutely correct. This is exactly why the Taiwan issue is so challenging, and in some ways, Taiwan is a perfect encapsulation of the problems that arise when applying the European IR model globally.

If we weren’t living in a Westphalian world, Taiwan’s status would be very easy to figure out.

Taiwan is a contentious issue only because we have decided that the sole type of entity deserving of legitimacy, recognition, and a seat at the table is the nation-state.

That is a uniquely rigid way of thinking about the world. Historically, even in Europe, there were kingdoms, principalities, duchies, and so on. In Asia, there were nomadic kingdoms, centralized Confucian states, and other forms of governance. If we weren’t so fixated on Westphalian norms, Taiwan could be its own thing. But that’s not the world we live in.

The easiest solution of the Taiwan issue would be to forgo Westphalian thinking in this instance, and let Taiwan be its own, distinct type of entity. But that’s not going to happen because that’s not the world we live in.

The global system today creates an incompatibility that’s difficult to resolve. If there were an easy answer, we’d have found it by now.

I have two main points about Taiwan.

1. The idea that China is preparing to invade Taiwan is more prevalent in the U.S. than in China. In Washington, D.C., the amount of money being spent on war-gaming simulations is staggering. Every few months, there’s a new prediction about when China will invade — 2027 is a popular date right now.
   

   But I don’t think Chinese leaders are approaching this the way Americans think they are. From everything I’ve read, the CCP and Xi Jinping reserve the right to use force because they view Taiwan as part of China. However, it’s not framed as an imminent invasion. The strategy so far has been to kick the can down the road, and that strategy has been remarkably successful.
   

   I do not understand why we are trying to change the status quo in Taiwan. The U.S. has maintained a policy of acknowledging China’s claim to Taiwan without endorsing it. China, in turn, says it reserves the right to use force but hasn’t acted on it. Meanwhile, Taiwan has its own flag, it’s own currency, and it’s own government. Taiwan has flourished in this environment, transitioning from a brutal authoritarian regime under the KMT to a thriving democracy. The island has grown wealthy, and China has grown wealthy too. The status quo is imperfect, but it works.
   

2. In terms of alliances, you are right that Asian countries are not abandoning their ties to the U.S. But let’s consider what happened when Pelosi visited Taiwan a couple of years ago. Within a week, every major Asian country, including ASEAN members like Vietnam, Malaysia, and the Philippines, publicly reaffirmed the One China policy. The only exception was South Korea, but even there, not a single official government official met her at the airport. President Yoon Suk-yeol wouldn’t meet with Pelosi during her visit. He claimed he was on vacation and wouldn’t pick up her calls.

I was at the Yongsan Presidential Office in South Korea last year, and I spoke with a senior national security advisor. When I asked about South Korea’s stance on the One China policy, he said their position has remained unchanged since normalizing relations with China in 1992. They adhere to the One China policy but won’t just reaffirm it every time China demands them to. The South Korean defense minister stated that he doesn’t want U.S. forces stationed on Korean soil to get involved in the event of a Taiwan conflict.

If there is a war over Taiwan, I suspect many Asian countries would slowly back away and avoid direct involvement.

Jordan Schneider: What’s fascinating is that the One China policy is almost like a modern version of an imperial-era common understanding. It’s not perfect and nobody is super happy, but it prevents war. China agrees to leave Taiwan alone, and Taiwan agrees to keep quiet and not embarrass China.

The concern is whether this delicate arrangement — “fudging it,” so to speak — will be enough going forward. CCP’s actions in autonomous regions, as well as in Hong Kong, raise doubts. Hong Kong was supposed to be the model for an enlightened, semi-autonomous relationship with Beijing, but that experiment has clearly failed. You saw what Mao did to Tibet.

Would you like to expand on the parallels you see between the One China policy and Korean diplomats sending “be nice” reminders to Beijing?

David C. Kang: The main distinction would be what’s internal and what’s external. We don’t have to like it and we don’t even necessarily have to agree with it, but China considers issues with Taiwan, Xinjiang, and Hong Kong to be internal. That’s very different from their relations with Vietnam or Korea.

Jordan Schneider: To clarify, Korea wasn’t viewed as internal in the year 1100, correct?

David C. Kang: Korea was not internal. It had formal tribute relations with China, however. The Koryo dynasty was never conquered by the Mongols. They suffered unbelievably, but they never gave in. The king survived. He had to keep moving around and stuff like that. They finally settled their relations with the Yuan dynasty when Kublai Khan decided he was going to adopt Chinese methods of doing things, and then they entered into tribute relations. The Koreans had to give princesses, but Korea remained as an independent country.

Jordan Schneider: What I’m saying is, if Taiwan could establish a kind of tributary relationship with China — where they play nice on the international stage, they send some symbolic gifts, and call their Olympic team “Chinese Taipei” — then that’s a compromise Taiwan and Taiwan’s allies would be open to accepting. But if China wants Taiwan to be another Xinjiang or Tibet, then things get a lot more complicated.

David C. Kang: Absolutely. This gets to the heart of the issue — what is internal, what is external, and how Taiwan fits into that dynamic. Taiwan is an unusual case because its status is unclear and doesn’t fit neatly into traditional categories.

China claims it annexed Taiwan in 1684 and argues that it has always been part of China. Others dispute this. What is clear, though, is that no other country claims Taiwan. Koreans don’t think it’s Korean. Filipinos and Vietnamese don’t claim it. Even Japan, which once ruled Taiwan, no longer stakes a claim. Taiwan may not definitively belong to China, but no one else is saying it belongs to them either. This ambiguity is one reason Taiwan is often treated as a Chinese issue within a broader civilizational framework.

Another important point is that Taiwan’s indigenous peoples historically never developed the kind of centralized government capable of engaging in formal diplomatic relations with China or other states. This contrasts with the Ryukyu Kingdom (Okinawa), which maintained formal tribute relationships with China, Japan, and Korea before being annexed by Japan in 1879. Similarly, Hawaii had a recognized monarchy before its annexation by the United States in 1898. While independence movements exist in Okinawa and Hawaii, they’re largely symbolic and seen as politically unviable.

Taiwan’s situation is different due to its unique geopolitical context. Its de facto independence exists more because of larger political factors than because of any historical claim to sovereignty. That’s the reality of international politics—it’s not necessarily about fairness but about the broader strategic situation.

Ilari Mäkelä: You have a great line in your book, “Despite decades of Western predictions to the contrary, it is by now widely admitted that East Asian states are not forming a balancing coalition against China out of fear of its rise.”

There are two ways to interpret that. One is that these countries don’t feel the need to balance against China. You also point out that military spending as a percentage of GDP has steadily declined across East and Southeast Asia, regardless of whether a country is a U.S. ally or not.

The other interpretation is that the U.S. acts as a kind of “gray eminence,” enabling this reduction in military spending. Many in the U.S. argue that it’s only because of American protection and military presence that these countries feel secure enough to avoid an arms race.

David C. Kang: That’s a long-standing argument, but I’m not entirely convinced. Let me explain why.

First, there’s an assumption that countries like Korea and Japan should naturally ally against China, given shared interests. But that hasn’t happened. For decades, I’ve heard arguments like, “Come on, Koreans, don’t you realize Japan is your friend and China is your enemy?” But those arguments don’t resonate in the region. Koreans don’t love China, but they don’t hate Japan as much as some think either.

The region doesn’t operate according to the neat balancing logic of realist international relations theory. American policymakers often push Asian countries to “balance” against China, but many simply don’t see the situation that way.

If U.S. protection were the primary reason for regional stability, we’d expect to see clear differences in behavior between U.S. allies like the Philippines, and non-allies like Vietnam or Malaysia. But we don’t.

Second, there’s the question of U.S. commitment. In recent discussions, I’ve been asked what China might be learning from the Ukraine war. My response is that Taiwan should also be paying attention to how the U.S. has responded.

Despite strong rhetoric, the U.S. has been very cautious about directly engaging in a war with a nuclear-armed superpower. Ukraine is on Russia’s doorstep, and while the U.S. has provided significant support, it has avoided direct confrontation. This raises an important question — would the U.S. really go to war over Taiwan, or even the Philippines?

This is a critical concern for Asian countries. They constantly assess whether they can truly rely on the U.S. in a crisis. Despite what alliance treaties might say on paper, the answer is far from certain.

Ilari Mäkelä: The final question I always ask my guests is, how has your research shaped your outlook on humanity?

David C. Kang: The biggest way my research has changed the way I view the world —and my outlook on humanity — is that the more I’ve delved into scholarship, whether it’s my earlier work on economic growth, political economy, and corruption in East Asia, or my decades of work on history, I’ve come to realize just how central values and beliefs are to human behavior.

People are far more motivated by what they value and believe in than by a simple cost-benefit analysis. I see this repeatedly, whether at the individual level or in the way nations act.

As scholars and social scientists, we often lean on cost-benefit frameworks because they’re more comfortable or measurable. But in my experience, values and beliefs are far more influential in driving decisions, shaping both individual actions and international relations.

Jordan Schneider: By the way, would you like to recommend any good books about the Imjin War or other topics we’ve discussed today?

David C. Kang: Yes! We already mentioned Liam Kelley’s Beyond The Bronze Pillars: Envoy Poetry And The Sino-Vietnamese Relationship, which is absolutely eye-opening. It explores how Vietnam historically viewed its relationship with China.

Sixiang Wang’s Boundless Winds of Empire: Rhetoric and Ritual in Early Chosŏn Diplomacy with Ming China is another excellent book. That’s the one you quoted earlier. It just came out, and it’s fantastic.

I also recommend Yuhua Wang’s The Rise and Fall of Imperial China: The Social Origins of State Development. While I don’t agree with everything in it, it’s an insightful materialist perspective on state formation in China, focusing on how China centralized and grew.

For a more classic take, Bin Wong’s China Transformed: Historical Change and the Limits of European Experience is a standout. It compares the Chinese and European paths of growth over the centuries and highlights the unique aspects of China’s development.

On the Imjin War specifically, there isn’t a definitive English-language book that comes to mind. [From the comments: Kenneth M. Swope’s A Dragon’s Head and a Serpent’s Tail: Ming China and the First Great East Asian War, 1592–1598. (University of Oklahoma Press, 2009).] Elizabeth Berry’s biography of Hideyoshi is excellent. It covers his life overall, and includes the Imjin War, although that isn’t the book’s exclusive focus.

Jordan Schneider: Since we just mentioned Hawaii, I’d like to shout out Shoal of Time: A History of the Hawaiian Islands by Gavan Daws.

It’s one of the most beautifully written books I’ve ever read. I’ve probably gone through ten books on Hawaii, and this one stands out far and away. It does an incredible job of telling the story of Hawaii’s transition from a kingdom to becoming part of America.

Caleb Harding and Lily Ottinger are back with another high-voltage deep dive into grid infrastructure. Their last piece analyzed China’s struggle to electrify AI data centers. Today, they’re sounding the alarm on the transformer shortage — and evaluating America’s options for rewiring industrial policy.

This story starts with a familiar picture: a foundational yet highly unstandardized piece of hardware, supply chain vulnerabilities exposed by Covid-19 and Russia’s war in Ukraine, and rumors of backdoor vulnerabilities placed by overseas manufacturers.

But we’re not describing legacy chips here — we’re talking about transformers.

Transformers perform an essential role in electricity distribution, increasing or decreasing voltage as power is transmitted from power stations to consumers. Electricity is stepped up to a high voltage to minimize losses during long-distance transmission, and then stepped down by one or more substations as it makes its way to the households, factories, and businesses that need power.

There are transformers at every major junction point in the grid. They are essential for nearly all construction.

It doesn’t matter if you want to build housing, AI data centers, renewable energy installations, EV charging stations, semiconductor fabs, or drone factories — you need transformers for all of the above.

Not even the fossil fuel industry is exempt — oil and gas drilling both require special transformers to supply power to rig machinery, compressors, refineries, and more.

Technologically speaking, transformers are relatively simple — they were first invented in the 19th century. And yet, the inability to build transformers is causing US industrial policy to short-circuit. In the words of utility resource planner Ajey Pandey:

The transformer shortage is bad. It’s like, hair on fire, biting my nails, losing sleep levels of bad. If you build new multifamily housing, you could be looking at a two-and-a-half-year wait for a transformer to supply the building — and that’s if the manufacturers are even accepting orders.

The Trump administration has ambitious plans for AI infrastructure, permitting reform, and offshore drilling projects. Without transformers, however, the US is destined to remain a build-nothing country, no matter how many billions of dollars the federal government dishes out. But how did it get so bad?

Surging Demand

Covid-19 impacted transformer supply chains, but a confluence of several demand-side factors has extended the shortage.

First, the US grid is aging. Transformers are typically rated for about 40 years of service, and much of the US grid is reaching the end of its allocated life.1 Replacing and upgrading these transformers is a major source of demand.

Second, overall electrification is increasing nationally. Americans increasingly use electricity instead of natural gas to cook their food, heat their homes, and take hot showers. The National Infrastructure Advisory Council (NIAC) describes this trend as “effectively irreversible.”

Third, and perhaps most obviously, scores of new renewable energy projects and EV charging stations require transformers to come online. Newly leased offshore drilling projects will have a similar effect — transformers for oil and gas production suffered from long lead times even before Trump’s executive order.

But if the demand is there, why hasn’t the supply followed?

Supply-side Blackout

The NIAC identified four structural supply-side challenges in a report released in June 2024.

1. Labor Shortages: The US does not have an effective pipeline to train and retain the manufacturing talent to build transformers. To be blunt, “I promise it’s worse than you think.”

2. Historic Industry Cyclicality: In the past, there has been a strong correlation between transformer demand and the housing market. When the housing market took off in the early 2000s, transformer production followed, and when the market crashed, it caused many manufacturers to exit the market. Suppliers were initially wary of repeating the same mistake.

3. Lack of Standardization: It is difficult to efficiently scale up production with a custom product. Pandey describes the situation:

“Utilities are also very specific about their [transformers]. There’s no standardization, really. Every utility has its own specifications. If you’re a small utility, you’re asking for a very, very small batch of this very bespoke thing.”

4. Material Shortages: Transformers are made with grain-oriented electrical steel (GOES),2 but there is only one domestic manufacturer, who is unable to meet domestic demand. Contributing factors include the destruction of the Azovstal steel plant in Mariupol and sanctions against Russian steel producers. Rising demand for non-oriented electrical steel, a key ingredient in EVs that comes from the same manufacturing facilities, will also increase supply tension.

The net result is that domestic supply can only meet 20% of transformer demand, and transformer prices have risen 60-80% since January 2020.

Security Implications

This shortage stands in the way of new construction and slows the rollout of energy infrastructure. But the effects are not limited to development — the shortage also poses a serious security challenge. If a foreign actor were able to destroy US transformers through physical or hardware attacks, we would have very limited means to repair the grid.

Such an attack would be considered an act of war and invite US retaliation — but when making preparations to confront the US over Taiwan, for example, this vulnerability might prove too tempting for China to resist.

Physical Attack

A 2014 analysis by the Federal Energy Regulatory Commission identified 30 critical high-voltage substations in the national grid (the list was not released) and predicted that losing just nine of these substations as the result of a coordinated attack could cause a nationwide blackout lasting for weeks or even months.

Obviously, that report is 10 years old. But according to an NPR interview with Richard Mroz, former president of the New Jersey Board of Public Utilities, the situation today is not substantially different. When asked about the FERC report, Mroz said that “knocking out those high-impact facilities is probably harder than everyone would think” — but he did not say that the underlying grid infrastructure had become less centralized.

What kind of security measures are in place to protect these high-impact facilities exactly? According to Politico, protections include “armed security staff, bullet-resistant fencing or video monitoring.” But these measures may not be enough to protect against a major blackout. In 2022 alone, there were a total of 1,665 security incidents involving the US power grid, including 60 incidents that led to outages. Notable cases include the 2013 Metcalf sniper attack and the 2022 attack on substations in Moore County, North Carolina. In both cases, the perpetrators were never caught.

As Pandey pointed out:

“If you extrapolate [those shootings] further, and you look at the FPVs that are being used to great effect in the Ukraine war… you start to wonder whether a coordinated team with five FPV drones could just knock out Boston3…

I think utilities are just now coming to terms with the fact that we are extremely straightforward civilian infrastructure targets…

Municipal utilities are smaller organizations. From an outside perspective, we’re at the intersection of poorly resourced and high costs for damage. It would kind of be dumb not to target me.”

Physical attacks on grid infrastructure in the US have primarily been conducted by disgruntled extremists looking to sow social unrest. But a state-sponsored physical attack on transformers is not inconceivable, especially given the mounting evidence of Chinese espionage and policing operations on US soil.

Hardware Attack

Importing transformers from China to help satisfy demand introduces additional attack vectors into the system.

China is the world’s largest exporter of transformers, with a total value of $54.1 billion in 2022.

In 2023, China was the USA’s third largest supplier of transformer components by value, selling American customers $375 million worth of “parts of electrical transformers, static converters.” Only Mexico ($508 million) and Canada ($395 million) outranked China by value. Chinese-made components represent 15% of America’s imports in this category. That might not sound like much until you remember that this is a calculation by value — since the RMB-to-USD exchange rate is kept artificially low, China likely provides a disproportionate volume of these components compared to other trade partners.

The picture gets worse once you consider imports of whole, completed transformers. The OEC reports that China was the USA’s number one supplier of transformers and transformer components in 2022, exporting $5.47 billion worth of product to the USA.

However, for the Large Power Transformers (LTPs) that are most critical to the grid, China is the 6th largest exporter to the US, according to a USITC report publicizing the ranking but redacting the power capacity and percentage of total imports.4

A report surveying public US International Trade Commission data found that, “Since 2006, USITC data shows that the US has imported 366 ‘liquid dielectric transformers having a power handling capacity exceeding 10,000 kVA’ from China.” Additionally, among Chinese-made transformers imported in 2020, “54 of these were classified as ‘having a power handling capacity exceeding 100,000kVA.’”

The iron core and copper coils inside of transformers do not constitute a threat — iron is not hackable. However, digital monitoring devices and sensors in large transformers could potentially come with built-in backdoors, which could be manipulated remotely. Such controls could theoretically be used to make a transformer overheat.

How plausible is that threat? It is difficult to quantify, but the DOE appears to have some suspicions. In the summer of 2019, federal officials seized a large power transformer from the Port of Houston and took it to Sandia National Laboratory. The Department of Energy did not comment on the incident or what they found. However, the initial WSJ piece that broke the story reported that, “Mike Howard, chief executive of the Electric Power Research Institute, said that the diversion of a huge, expensive transformer is so unusual — in his experience, unprecedented — that it suggests officials had significant security concerns.”

The following summer, in May 2020, President Trump issued an executive order on securing the United States bulk-power system. Operating on that authority, the DOE issued a “prohibition order” in Dec 2020 that banned some utilities5 from buying transformers of 69 or more kV from China.

In the FAQ document attached to that order, the DOE said it “has reason to believe, as detailed in the Prohibition Order, that the People’s Republic of China is equipped and actively planning to undermine the Nation’s bulk power system.”

However, the order was revoked in Apr 2021 “while the Department conducts a Request for Information to develop a strengthened and administrable strategy to address the security of the US energy sector.”6 It’s plausible that the new Trump administration will reinstate this executive order (and the proposed tariff regime could collapse demand for China-manufactured transformers anyway), but in the meantime, it appears the US does not yet have a comprehensive strategy to deal with the hardware vulnerabilities.

China’s transformers are hopefully not a part of the crucial 30 substations across the US power grid. However, if China could exploit hardware backdoors and cause Chinese-made LTPs to self-destruct, it could still deal a significant blow to the US grid.

Solutions for a Brighter Future

The NIAC report outlines a robust toolbox of policies that could provide some reprieve from the shortage. These include tax breaks and technician training programs, incentives to standardize transformer design, trade deals to strengthen supply chains (particularly for grain-oriented electrical steel), and even a strategic transformer reserve.

Regarding the issue of skilled labor, Pandey told ChinaTalk that US universities are failing to teach relevant electrical engineering skills. In his words,

A lot of domestic manufacturers are starving for talent and pulling in foreign staff to plug gaps. Manufacturers particularly call out the lack of college graduates who know industrial manufacturing or three-phase power engineering. Speaking from experience, I went to a very large flagship state school, and there was literally no one in the electrical engineering department who could teach three-phase power engineering.

If the Trump team is thinking about retooling education pipelines, this is something for them to consider.

To secure greater access to grain-oriented electrical steel, there is one obvious solution. US Steel does not currently have the infrastructure to produce GOES — but Nippon Steel does. If the new administration allows the highly publicized acquisition deal to proceed, perhaps some of Nippon Steel’s pledged US investment budget could go toward expanding domestic production capacity for GOES.

As far as productive trade policy is concerned, tariffs on allies that produce transformers and transformer components would be counterproductive. Even with massive unmet demand, the domestic transformer industry has been unable to ramp up production to clear backlogged orders. There are already, structurally, too many sticks in this sector. Given the similarities with semiconductor manufacturing, we believe some carrots are in order.

By supplementing the 2020 transformer executive order with real industrial policy, the Trump administration has a huge opportunity to jumpstart the construction of housing, factories, and energy infrastructure in the United States.

To close, we’ll leave you with the full list of policy recommendations from the National Infrastructure Advisory Council:

“The NIAC recommends the Federal government craft policies and designate funding targeted at increasing domestic capacity, such as tax credits, grants, accelerated depreciation, funding for new working apprentice and/or training programs, and other incentives, using the CHIPS Act as a model…

The NIAC recommends convening all parties who drive demand to achieve greater accuracy in transformer demand forecasting that will provide a more precise outlook across the next 10 to 15 years…

The NIAC recommends encouraging long-term contracts/customer commitments between transformer suppliers and the sectors driving demand…

The NIAC recommends establishing a strategic reserve of transformers, with the US government as the buyer of last resort…

The NIAC recommends the Federal government promote collaboration between design engineers from utilities, trade associations, and domestic manufacturers with the goal of standardizing transformer design and reducing complexity associated with customization…

The NIAC recommends the Federal government ensure a sufficient supply of electrical steel by coordinating incentives for supply, governmental efficiency standards, and trade policy…

The NIAC recommends the Federal government grow the pipeline of qualified workers by partnering with universities, community colleges, and trade schools on training programs, while working with federal, state, and local governments to craft tax incentives for workers who enter the field.”

4

See the table below, which ranks the sources of US transformer imports by voltage (MVA) other than South Korea, which ranks first and is the subject of this report. Specific figures redacted. Source: USITC.

Last Thursday, Rep. Tom Tiffany (R-WI), joined by twenty-three other House Republicans, introduced a concurrent resolution calling on the White House to recognize Taiwan as an independent sovereign nation. I doubt anything will come of it (this is Rep. Tiffany’s fourth time introducing the same resolution, and resolutions don’t have the force of law). But the first three weeks of Trump’s second term, if nothing else, show that his administration is willing to upend foreign-policy orthodoxies — and for all we know, major changes to US-Taiwan policy could be just around the corner.

Some new appointees in State will soon be asked to whip up fresh policy recommendations for Secretary Rubio on whether strategic ambiguity or strategic clarity is more likely to prevent a war over Taiwan.

This post is a comprehensive guide to understanding that debate. I read 50 op-eds and academic articles about this question and catalogued what they said.1 Since 2020, every debate over strategic ambiguity vs. strategic clarity is just some combination of the same 12 arguments (I never encountered a genuinely distinguishable #13) — so reading this one piece will get you completely up to speed on the current state of the discussion.

At the end of this piece, we’ll also give out awards for the two essential pieces making pro and con arguments.

Our working definitions:

• Strategic ambiguity 戰略模糊 is the “policy” of the United States to (1) not commit to Taiwan that the US military would respond in the event of Chinese use of force, and (2) not commit to China that the US military would not respond in the event of Chinese use of force. The idea is to simultaneously deter Taiwan from pursuing de jure independence and deter China from taking military action against Taiwan. I put “policy” in quotes because there is no statute or executive order which establishes this practice. Even so, every US president since Nixon has basically adhered to this approach.

• Strategic clarity 戰略清晰 is a proposed but as-yet-unadopted policy for the United States to abandon strategic ambiguity and make an explicit defense commitment to Taiwan.2

The five arguments for strategic clarity:

1. Strategic clarity is not provocative per se. If worded carefully and announced thoughtfully, strategic clarity can be in full accordance with the One-China policy, and China will have no legitimate excuse to be any more antagonized than it was before. On the other side of the Strait, the risk of appeasing or emboldening Taiwan separatists is low. Bona fide separatists are a fringe minority of the Taiwanese population and hold very little political capital. Taiwan’s elected leaders, whether KMT or DPP, are very careful when making public statements. And even if we assume for the sake of argument that strategic clarity could embolden separatism in Taiwan, the benefits of adopting strategic clarity would far outweigh those risks.

2. Adopting strategic clarity is necessary to effectively deter China today. The United States hasn’t demonstrated sufficient clarity in other recent conflicts — e.g. regarding Afghanistan, Ukraine, and Iran — to disastrous consequences. Even the best of sanctions can go only so far. Hard military capabilities are necessary, of course, but they are no longer sufficient — the United States needs to demonstrate the will to jump into a hot military conflict. Displaying strong resolve is especially important in conflicts involving paternalistic nuclear-armed states, which includes Russia’s Ukraine invasion as well as a China-Taiwan scenario.

3. Strategic ambiguity has “outlived its usefulness,” “run its course,” and no longer meets the demands of the twenty-first century. The assumptions which underlaid strategic ambiguity may have held 50 years ago, but not anymore. In 1979, the United States could indisputably overpower the PLA in a Taiwan contingency. In the decades since, PLA spending has ballooned, US industrial capacity has crumbled, and PLA operations in the Taiwan Strait are increasingly frequent and aggressive — which means a blockade or outright invasion of Taiwan is no longer a fantasy. Meanwhile, the Taiwanese people may not be willing to accept the status quo indefinitely. If they lose faith in the United States — and ambiguity is hardly a reassuring defense commitment — Taiwan’s leaders could take matters into their own hands and move toward de jure independence at some point. In other words, strategic clarity may keep Taiwan separatists in check more effectively than strategic ambiguity.

4. Everyone — the United States, China, Taiwan, Pacific allies — already assumes that the United States will defend Taiwan if the PLA took kinetic military action against the island. The PLA expects the US military to be involved, and makes preparations accordingly. Strategic clarity, then, merely aligns stated policy with well-understood expectations — no one is guessing anymore. Clearly communicating intentions is generally a good way to prevent war. And in any case, President Biden basically ushered in strategic clarity, stating four times during his presidency that the United States would be obligated to intervene militarily if Taiwan were attacked.3 Xi Jinping didn’t respond to those overtures with a rash blockade or invasion.

5. Strategic clarity reduces abandonment concerns among Taiwan and Pacific allies. After all, there is little incentive for allies to make serious Taiwan-contingency preparations if, deep down, they are unconvinced the United States will come to Taiwan’s defense. Taiwan and Pacific allies need reassurance that their military investments are urgently needed and won’t go to waste — and that’s exactly what strategic clarity would bring. For starters, adopting strategic clarity would make it difficult and politically costly for a future US president to water down the American commitment to Taiwan. Better still, a US policy of strategic clarity has the potential to create a domino effect, whereby Australia, Japan, South Korea, the Philippines, and others would line up to issue their own clear defense commitments to Taiwan — and that combined deterrent effect would be incredibly powerful in preserving cross-Strait peace.

The seven arguments for strategic ambiguity:

1. Strategic ambiguity maintains US flexibility and reduces entanglement risks. The upshot of no clear defense commitment is that the United States can be extremely nimble in its diplomatic and military responses to any kind of Taiwan contingency. Ambiguity also allows the United States to “hide its cards,” which keeps CCP leadership guessing and thus more cautious;4 in fact, strategic ambiguity can take some of the credit for reducing all the PLA’s military actions toward Taiwan to “gray zone” activities, which, though frustrating, are far easier to manage than hot conflict. A clear defense commitment in the form of strategic clarity, on the other hand, would give both China and Taiwan the power to dictate America’s military decisions — and however much Taiwan may deserve US defense, that decision should remain with the United States and its elected leaders alone. For China’s part, strategic clarity would invite China to move as close to US red lines as it could, after which it could dictate the US military’s next moves by deliberately crossing those red lines. And the Taiwanese could be emboldened by strategic clarity to pursue de jure independence, knowing full well the US military is standing by and locked in. An independent Taiwanese identity is burgeoning, and although Taiwan presidents Tsai Ing-wen and Lai Ching-te have been careful, future politicians may not be.

2. Adopting strategic clarity, ironically, could provoke a PLA attack, starting the very conflict it seeks to prevent. China takes threats to its sovereignty extremely seriously — just look at its activities in Hong Kong, Xinjiang, and on the Sino-Indian border. An explicit defense commitment would undoubtedly be perceived as a violation of its One-China principle and as crossing the red lines of the 2005 Anti-Secession Law, thereby necessitating an immediate escalatory response. Indeed, the CCP’s top leadership perhaps couldn’t survive politically if they didn’t respond with quick, provocative action: as Orville Schell put it, “I think they’re incapable of saying, ‘We can’t win. It doesn’t work. Let’s just cut our losses and get out’ — because of the matter of face.” That’s especially true for Xi Jinping: “His ambition is too overweening. His sense that any sign of concession evinces weakness is too repugnant to him.”

3. “If it ain’t broke, don’t fix it”: strategic ambiguity has worked for decades — and we all know it has worked because Taiwan is not under CCP rule today. The current US policy framework toward Taiwan is assurance enough. The Taiwan Relations Act requires the United States to “make available to Taiwan such defense articles and defense services in such quantity as may be necessary to enable Taiwan to maintain a sufficient self-defense capability”; the de facto diplomatic relations between the United States and Taiwan have enabled billions of dollars of military assets to be sold to Taiwan. Meanwhile, the holy words memorialized in the Three Communiqués — which US diplomats ritualistically utter before conversations with Chinese counterparts — have largely satiated China’s sovereignty concerns. Having functional diplomatic, military, and economic relations with Taiwan while also not antagonizing China to the point of kinetic military action is one of the biggest US foreign-policy successes of all time. The US-Taiwan-China relationship is managed successfully in large part through adherence to a strict verbal theology — a foreign-policy practice entirely unique to this triangle — and the United States shouldn’t do anything to upset this delicate balance.

4. Adopting strategic clarity poses credibility issues. Especially given the ongoing conflicts in Ukraine and the Middle East, as well as waning US industrial capacity, the United States is in no position to change its force posture quickly enough to make an explicit defense commitment credible. If the CCP leadership doesn’t view US strategic clarity toward Taiwan as credible, perhaps they would come to the wrong conclusion — that the United States is so weak that it can do nothing else but bark. Perceived weakness is dangerous. And even if the CCP leadership doesn’t perceive serious US weakness or unwillingness to defend Taiwan, strategic clarity at least dares China to test US resolve. The PLA could begin by moving militarily on the Taiwan-controlled Kinmen or Matsu islands; an underwhelming US military response to such moves would represent a fatal blow to American credibility.

5. Far from reassuring allies, adopting strategic clarity would create fears of entrapment. Given the strong military interdependence between the United States and its Pacific allies, a US policy of strategic clarity is effectively tantamount to Australian, Japanese, Korean, and Filipino explicit defense commitments to Taiwan as well — and US allies may not be ready or even willing to get involved in a Taiwan contingency. In other words, no domino effect would materialize. While the United States justifiably hopes its Pacific allies would support the US military in defending Taiwan, convincing them to do so would be much harder if they felt blindsided by an American unilateral policy change directly implicating their militaries. The United States shouldn’t needlessly entangle its allies.

6. Whatever one may say about the threat to Taiwan, it is not imminent. There is little to no evidence that China is poised to take Taiwan by force in the near future. Yes, China’s foreign ministry says things all day long, and yes, Xi Jinping publicly sets Taiwan-oriented PLA modernization benchmarks — but the military assets required to pull off a successful amphibious invasion are obscene, and China simply couldn’t hide a buildup of that magnitude from the world, let alone Western intelligence services. Xi Jinping can’t just wake up tomorrow and decide to send the ROROs across the Strait. Adopting a dramatic policy shift in response to a non-imminent threat from China would make engaging with the CCP leadership nearly impossible in the future, for they could assert — and perhaps not without merit — that the United States is the party acting disproportionately. Functional diplomatic relations with China have been crucial in preventing cross-Strait conflict.

7. Maybe the best thing the United States can do to preserve cross-Strait peace is to keep the Taiwan military on its toes and the Taiwanese people mentally ready to fight. Today, however, Taiwan’s defense spending relative to GDP is woefully insufficient; its military assets are being run down by constant PLA incursions; its government likes investing in expensive, shiny toys like submarines and advanced fighter jets, which provide little deterrent value relative to their cost; and conscription is a joke. Strategic ambiguity is tough-love encouragement to the Taiwanese to make serious defense preparations, which is imperative now more than ever. Adopting strategic clarity, on the other hand, would allow the Taiwanese to blindly free ride on US support — precisely the wrong message at the wrong time.

That’s tens of thousands of words all condensed into a completely comprehensive, proudly non-AI-generated, 1,600-word argument map.

What we need now is not more op-eds rehashing exactly these same arguments all over again. Instead, we need hard, empirical analysis to assess the merits of each of these arguments.

For example, a recent survey conducted by UNLV professor Austin Horng-en Wang 王宏恩 showed that “both strategic ambiguity and dual clarity [ie. conditional strategic clarity] induce a similar effect by making citizens in Taiwan less supportive of pursuing de jure independence” — a finding which suggests that the Taiwanese populace is “willing to trade their support for de jure independence for stronger support from the United States.”

With some empirical results like that on the table, diplomats could then assign probabilities to key events (would adopting strategic clarity increase or decrease PLA gray-zone activities? would Pacific allies be more or less likely to adopt explicit defense commitments of their own if the United States adopted strategic clarity? etc.), then assign confidence intervals to each of those probabilities, and then hopefully arrive at an optimal result.

The folks at fp21 and I worked on a project in this vein back in 2023 — a redesign of the State Department memo called the Bayes Brief, which maps evidence to arguments to assessments and finally to policy choices. You can experiment with the Bayes Brief yourself, here — a relatively short, evidence-based questionnaire that will guide you in deciding whether strategic ambiguity or strategic clarity is more likely to prevent conflict over Taiwan.

A system for producing evidence- and data-driven policy conclusions would be far superior to what we have today. At least one reason dozens of op-eds could be boiled down to just a few paragraphs is because, sadly, many of them read something like this: “The Taiwanese love their democratic freedoms. [x10] Therefore, clarity!” That mistakes a conclusion for an argument. As far as I can tell, there is literally zero disagreement in the US foreign-policy establishment over propositions like these:

• The US government should adopt policies that reduce the risk of war.

• Conquest of Taiwan is antithetical to US interests and credibility.

• Taiwan’s status as a democracy is miraculous, admirable, and, all things being equal, worthy of continued US support.

• Taiwan-controlled TSMC is critical to global supply chains and US national security.

• China’s 21st-century military buildup is massive and ongoing.

• The consequences of Chinese military action against Taiwan would be globally catastrophic.

None of these propositions should factor into the ambiguity-vs.-clarity decision. Everyone already agrees; the only disagreement is over how to best keep the peace. As an op-ed from RUSI refreshingly framed it, “The crucial issue here is a disjuncture between the moral grounds for adopting less ambiguous commitments to Taiwan and the continuing strategic utility of ambiguity if the core US objective is avoiding war with China.” Yes! I’ll say the same thing, but less nicely: the literature is full of tacit ad hominem — e.g. “ambiguity proponents are authoritarian shills!”; “clarity proponents are warmongers!” — and appeals to emotion are as unhelpful as they are annoying.

Relying on emotional appeal has led to foreign-policy dumpster fires before. In a 1969 Foreign Affairs essay, democratic advisor and later LBJ SecDef Clark M. Clifford recounted the following anecdote, which took place during the presidential transition from Eisenhower to JFK in January 1961:

My notes disclose the following comments by the President:

“At this point, President Eisenhower said, with considerable emotion, that Laos was the key to the entire area of Southeast Asia.

“He said that if we permitted Laos to fall, then we would have to write off all the area. He stated we must not permit a Communist take-over. He reiterated that we should make every effort to persuade member nations of SEATO or the International Control Commission to accept the burden with us to defend the freedom of Laos.

“As he concluded these remarks, President Eisenhower stated it was imperative that Laos be defended. He said that the United States should accept this task with our allies, if we could persuade them, and alone if we could not. He added, ‘Our unilateral intervention would be our last desperate hope in the event we were unable to prevail upon the other signatories to join us.’”

That morning’s discussion, and the gravity with which President Eisenhower addressed the problem, had a substantial impact on me. He and his advisers were finishing eight years of responsible service to the nation. I had neither facts nor personal experience to challenge their assessment of the situation, even if I had had the inclination to do so. The thrust of the presentation was the great importance to the United States of taking a firm stand in Southeast Asia, and I accepted that judgment.

After returning from diplomatic visits to several Southeast Asian nations as well as Australia and New Zealand in the summer of 1967, Clifford recalled,

I returned home puzzled, troubled, concerned. Was it possible that our assessment of the danger to the stability of Southeast Asia and the Western Pacific was exaggerated? Was it possible that those nations which were neighbors of Viet Nam had a clearer perception of the tides of world events in 1967 than we? Was it possible that we were continuing to be guided by judgments that might once have had validity but were now obsolete? In short, although I still counted myself a staunch supporter of our policies, there were nagging, not-to-be-suppressed doubts in my mind.

That’s a confession if I’ve ever seen one.

Award Section

• Most persuasive article advocating for strategic clarity: US Army Lieutenant Colonel Jeffrey C. Higgins. He analyzes several historical episodes in which US officials’ ambiguous and unsynchronized defense commitments — pre-WWII Europe, post-WWII Korea, and more recently, Afghanistan and Ukraine — resulted in increased hostilities in precisely those regions. Continued ambiguity and incoherent messaging toward Taiwan will inevitably lead to the same result: adverse military action.

• Most persuasive article advocating for strategic ambiguity: Nien-chung Chang-Liao 張廖年仲 & Chi Fang 方淇. A switch to strategic clarity, they argue, could pressure CCP leaders to abandon PLA gray-zone tactics in favor of more aggressive military tactics, while simultaneously limiting the range of acceptable response options at the disposal of the United States.

Up on the podcast feed today, , , , and I had way too much fun debating who holds the AI Mandate of Heaven and how Steve Jobs AGI will bring back Apple’s glory days. Check it out on Spotify, iTunes, or your favorite podcast app.

Beyond covering the living daylights out of DeepSeek, ChinaTalk had a very busy year. Here is a roundup of our top pieces.

Jordan: My biggest stories this year were a deep dive into the R&D section of the CHIPS Act co-authored with Arrian entitled How to Make the NSTC a Moonshot Success, an essay on 1950s RAND called When RAND Made Magic, and two travel posts from time I spent in Japan and Norway. I’m also very proud of helping organize an excellent essay contest around legacy chips with Noah Smith and the Federation of American Scientists.

In August I became a dad and spent the fall on leave. Despite barely working this past fall, I’ve been incredibly proud of how the ChinaTalk team put out some of our best-ever writing. This is not a one-man band anymore.

Nancy: Doing Chinese chatbot speed runs is basically a right of passage for ChinaTalk contributors as this point — and Nancy had a fantastic censorship analysis of China’s leading genAI models — but she was also covering some of the deepest-rooted Chinese intellectual debates.

Why do science and tech dominate China’s political culture? How do Party elites reconcile free-market-driven innovation with top-down, industrial-policy-driven economic coordination and risk mitigation? How did politicians at this year’s Two Sessions debate competing imperatives of AI innovation versus security and cultural erosion?

See also Nancy’s encomium to the late Chinese legal scholar Jiang Ping, and her translation (alongside Dylan) of Chinese think-tanker Zhang Yunchen’s take on China’s status as a “great nation.”

Bit Wise: Our pseudonymous contributor has been a huge addition to our team. Our best model deep dives — SenseChat V5 and the real story behind the deeply misreported ChatXiPT — were spearheaded by Bit Wise.

Bit Wise is also tapped into China’s elusive “algorithm registry,” and in particular, how the Cyberspace Administration of China is updating registration protocols for genAI models. It’s still an open question whether China’s AI governance reflects a light-touch approach, or if the CAC intends to robustly control (and limit) the development and deployment of genAI algorithms — so stay tuned for Bit Wise’s future installments on China’s AI policy!

See also Bit Wise’s fun piece on why Maotai is investing in chip startups.

Irene Zhang earns our “how was mainstream media so slow to something we covered” Scoop of the Year award, courtesy of her excellent investigative reporting on how easy it is to buy advanced chips on Xiaohongshu, China’s Instagram. No surprise when — a full year after Irene’s report — the WSJ covered the retail smuggling ecosystem, including one student in Singapore who smuggled six Nvidia semis back to China (for a whopping 0.4% commission). Irene also had an important translation of a fascinating “Nvidia vs. Huawei” article on Huawei’s GPU ambitions.

Angela Shen on Industry and AI: ChinaTalk’s first full-time analyst, Angela has already published on China’s biotech sector, robotics industrial policy, and humanoid robotic development.

Chinese biotech companies like XtalPi are using AI to accelerate drug discovery — but notwithstanding China’s huge strides, there’s growing tension between global collaboration (which drives innovation) and national-security concerns (particularly around biodata).

And on the robotics front, unlike the West, China sees robots as the solution to its demographic crisis, rather than a job-killing threat. And humanoid robots in particular are the ones to watch: China wants to dominate in humanoid robots by 2027, seeing them as the next frontier in automation.

Lily Ottinger, our lead editor, debuted predicting that the Sino-Russian friendship has a 75% of crashing by 2030.

Some of our best-performing pieces were Lily’s excellent, rapid-response roundups of Chinese social-media posts following notable events in American presidential politics: Trump’s New York criminal conviction, the Biden-Trump debate in late June, and the first assassination attempt of Trump. She was also basically the first person to cover Tim Walz’s China connections, publishing a rundown within hours of his nomination as Kamala’s VP running mate.

See also her analysis of China’s energy conundrum in powering data centers, the legacy of the Chinese Exclusion Act, and Chinese netizens’ take on Netflix’s Three-Body Problem.

Arrian Ebrahimi on chips: You know him as the brainchild of the excellent Chip Capitols Substack (and now survivor of the first semester of law school). On his way home from Peking University this summer, he spent a few months interning for TSMC in Hsinchu — and of course, he found Taiwan’s top tech journalist, Lin Hung-wen 林宏文.

In an epic Mandarin-language interview, Lin traces the history of Taiwan’s semiconductor success. It wasn’t master-planned — rather, it evolved step by step, starting from Christmas-light manufacturing in the 1950s, through ITRI (Industrial Technology Research Institute) and technology transfer from American tech company RCA, all the way to the present day, when TSMC’s CEO wakes up at 4:30 a.m. so he can get the latest from US and European customers.

“Taiwan used to be like the black-clad stagehands who move props unseen, working behind the scenes while companies like Apple and Nvidia took center stage. Now,” Lin says, “these ‘stagehands’ have become crucial players.”

See also Arrian’s breakdown of Taiwan’s modern-day chip policies (subsidies and tax credits) — which resembles a centrally administered national strategy committed to bolstering its existing strengths — and his WaPo op-ed (with Jordan and Chris Miller).

Will Chu, Video Editor: We have a podcast, we have a Substack — but the real ones know that we have a burgeoning YouTube channel, too! In July, we launched our first scripted YouTube video, and in the months since, Will and Lily have produced nearly a dozen bangers.

Were you ever wondering why Taiwan competes in the Olympics as “Chinese Taipei”? Or why Moutai — China’s most popular liquor — is investing in Chinese chip companies? When Halloween rolled around, were you actually more curious about the history behind the Ghost Festival? Is Tim Walz a Chinese sleeper agent? Is the United States going to lose the AI arms race? Is Taiwan going to be invaded?

In ten minutes or less, get the answers to all of the above! We’re really excited to experiment more with this type of content, bringing ChinaTalk to an ever-expanding audience.

Dylan Levi King on weird stuff: Dylan, ChinaTalk’s 黑马, brought the analysis you didn’t know you needed. His three best features this year:

1. How well does the United States intelligence community predict the future? Probably not well enough. So we should take a lesson from the CIA “heretics” — analysts in the Sino-Soviet Studies Group (SSSG), which functioned more like an area-studies department — who predicted the Sino-Soviet split years before it happened. “The futurological lessons are clear,” concludes Dylan: “intellectual sovereignty is a prerequisite for breaking with the popular narrative and calling the future.”

2. The PLA now issues Type 191 rifles, replacing the Type 95. Turns out, if you trace PLA rifle history — all the way back to the 1979 Sino-Vietnamese War — the Type 191 “seems like further proof” that a Taiwan beach landing is not what the the PLA infantry is preparing for. (Phew!)

3. Ishiba Shigeru ascended to Japan’s PM post — “almost by default” — and Dylan soberingly explains why “Ishiba is destined to become another in a long line of interchangeable helmsmen, steering Japan into increasing geopolitical marginalization.”

Nicholas Welch, Republic of ChinaTalk Editor: January 2024 marked one of Taiwan’s most consequential elections, with Lai Ching-te winning the DPP an unprecedented third term in the executive branch. He had you covered before, during, and after the election results came in, bringing on the pod Taiwan NSC Deputy Secretary-General Lin Fei-fan, National Chengchi University Professor Lu Yeh-chung, and Stanford-based Taiwan expert Kharis Templeman.

He also had a blast interviewing US Army Lieutenant Colonel Kevin McKittrick on the WWII-era Operation Causeway, and why the United States aborted its plans to invade then-Japanese-controlled Taiwan — as well as going through the latest empirical research on how to best deter the PLA from launching an invasion in the first place.

See also some law-school chops coming to bear in an interview with DOJ attorneys on data-security regs, and a rundown of VP-elect JD Vance’s positions on China and AI.

Yiwen Lu: Yiwen’s top piece was a comprehensive take on OpenAI’s unceremonious exit from the Chinese market in June, plus a truly essential deep dive with Irene into China’s best AI boyfriends.

Mood Music

Yesterday we published our interview with Anthropic CEO Dario Amodei. You should check it out or have a listen on Spotify, iTunes, or your favorite podcast app. Transcript here.

Anonymous contributor Code Smith reports:

DeepSeek’s R1 open-source reasoning model has garnered attention from policymakers in China and beyond. President Trump commented that DeepSeek’s open availability “could be very much a positive… Instead of spending billions and billions, you’ll spend less, and you’ll come up with, hopefully, the same solution.”

To DeepSeek’s leadership, Trump couldn’t have offered higher praise. Yet, the lofty ideals of open global collaboration may struggle to survive the harsh realities of geopolitics. In Trump’s words, “The release of DeepSeek AI from a Chinese company should be a wake-up call for our industries that we need to be laser-focused on competing to win.”

But how will its new-found fame force DeepSeek to change?

This article will explore three new challenges that DeepSeek will have to navigate going forward: corporate partnerships, employee retention and hiring, and, most importantly, the influence of the Chinese government.

Does DeepSeek need a hyperscaler?

When OpenAI hit a compute bottleneck, it turned to Microsoft. That partnership transformed OpenAI’s corporate structure, reshaping it from a non-profit committed to open source principles into a for-profit company with closed source models.

DeepSeek could face a similar transition.

China’s hyperscalers — ByteDance, Alibaba, Tencent, and Huawei — are likely all pitching partnerships to DeepSeek. Rumors of a potential collaboration with ByteDance have already surfaced in Chinese media, though their authenticity is uncertain. The partnership surely wouldn’t be a straightforward one, considering that ByteDance’s model Doubao 豆包 is among the most popular chatbots in China.

A partnership with a hyperscaler could solve DeepSeek’s compute challenges, enabling it to remain at the cutting edge of AI research. However, it would also likely transform DeepSeek’s business model and culture. For now, DeepSeek claims it has no business model, prioritizing research over profit, but a partnership deal with a hyperscaler will come with strings attached.

Employee temptations?

It’s one thing to be a young, idealistic programmer grinding for the love of AI research. But DeepSeek’s rise to international prominence will put those values to the test. Employees may soon find their inboxes flooded with offers for higher-paying jobs (a DeepSeek V3 developer has already been poached by Xiaomi and Dario on ChinaTalk said he’d love for their engineers to “come work for us.”). Some might even dream of founding their own companies, leveraging the DeepSeek name to secure blank checks from investors. Will they resist such temptations?

When asked whether DeepSeek’s “curiosity-driven madness” is a sustainable strategy for DeepSeek long term, CEO Liang Wenfeng replied, “Not everyone can stay passionate their entire life. But most people, in their younger years, can wholeheartedly dedicate themselves to something without any materialistic aims.”

DeepSeek may also soon begin hiring and expanding rapidly. The fame it has garnered could attract a very different crowd compared to the passionate, idealistic individuals it drew just a year ago. What impact will this have on a company whose success relies on a small, tightly-knit team and a shared cultural bond?

Beyond these short-term cultural shifts, DeepSeek may even have to calibrate a new relationship with the Chinese government in the long term.

Could DeepSeek get the “national champion” treatment?

On January 20, DeepSeek CEO Liang Wenfeng 梁文锋 was invited to meet Premier Li Qiang — a clear sign that the company is on the government’s radar. A meeting with Li Qiang is no small matter.

The company is poised to showcase China’s resilience under U.S. export controls. Policymakers may use the company’s example to argue that these controls aren’t stifling Chinese innovation — a powerful narrative for both domestic and international audiences (although, whether DeepSeek’s success actually refutes the effectiveness of export controls is another question).

With political backing, DeepSeek could reap state-sponsored benefits — but would the payoff justify the exposure to new domestic and international pressures?

The rest of this piece will be dedicated to speculating on the implications of a world where DeepSeek ascends to national champion status. To be clear, evidence of involvement with the Chinese state is flimsy for now, but we’ll explore some key indicators of change to watch going forward.

What’s in it for DeepSeek?

If DeepSeek were to cut a deal with the Chinese State, potential benefits for DeepSeek could include:

1. Funding

For now, DeepSeek takes great pride in the fact that it is funded entirely by High-Flyer. But clearly their research budgets are not unlimited — if they were, why would DeepSeek specifically pursue low-cost training strategies for R1?1

If and when funding does become relevant, political backing significantly reduces investment risk in China. If the government signals support for DeepSeek, it will attract more private and institutional capital at a better valuation. DeepSeek may even attract direct government funding, such as from a newly established 60 billion RMB government guidance fund for AI or 1 trillion RMB special AI financing by the Bank of China.

2. Access to other resources

The government could assist DeepSeek in other ways beyond funding. For example:

• Compute Infrastructure: Local governments, which are major operators of compute resources, could prioritize DeepSeek’s access. SMEs have received “Compute vouchers” on a small scale, but such policies could be expanded to include companies like DeepSeek.

• Government Data: Beijing has signaled intent to unlock more value from its vast troves of government data (think health data, traffic data, industrial data held in SOEs, etc). DeepSeek could benefit from privileged access.

• Chip Supply: Depending on the level of state involvement, DeepSeek could even receive assistance in circumventing controls on chip supply.

3. Government Procurement

The Chinese government has a history of using procurement to bolster favored companies. Image recognition firms, such as Hikvision, rose to prominence largely because the state was a major customer. If DeepSeek were to be recognized as a “national champion” in the future, it might secure lucrative government contracts as a result.

But at what cost?

While state attention could bring substantial benefits, it’s not without risks. Stronger ties to the government could also lead to new challenges for DeepSeek. The company wouldn’t be the first tech giant to fall from grace. Risks include:

1. Increased State Scrutiny

With government backing comes government scrutiny. Liang Wenfeng could find himself under the spotlight, with his personal character and loyalty to the Party potentially questioned. Rumors about immoral behavior — such as sexual abuse and extramarital affairs within Liang’s hedge-fund “High Flyer” — could be investigated more closely if he becomes a poster child for Chinese innovation. The state may start paying more attention to things like the post published by his quant trading firm complaining about being blamed by the state for the market volatility in 2024. Add to that the Chinese government’s general suspicion of the finance industry.

Even though last year Xi famously asked last year “why are the number of Chinese unicorns dwindling,” the instincts that drove the techlash are still in the system. Are we sure we are well and truly past that era?

2. Cultural Interference

DeepSeek’s innovative, research-focused culture could come under threat. The government might waste the firm’s time with mandatory study sessions or political events. It may also push for specific research directions aligned with state priorities. This could dilute the company’s focus on basic research and scientific curiosity.

3. Loss of Operational Independence

As seen with other tech companies, the state could demand “golden shares” (minority stakes with strategic voting rights) to gain influence over DeepSeek’s operations. This might limit the company’s autonomy, particularly if it’s forced to align its technological roadmap with government directives. Beijing’s rhetoric has historically been in favor of open source, but that could change if DeepSeek’s innovations consistently outpace international AI research. Perhaps there is some level of technical sophistication at which the CCP would insist that DeepSeek stops open-sourcing their models.

4. Complacency Risk

A flood of government contracts and resources could stifle DeepSeek’s long-term innovation. Unlimited budgets and carte-blanche access to compute resources mean less pressure to find creative ways to boost efficiency. Guaranteed government contracts could increase complacency or draw focus away from cutting-edge innovation. Lenovo serves as a cautionary tale. In the 1990s, state support helped Lenovo secure major IT tenders, but it also undermined the company’s incentive to innovate. As Douglas B. Fuller wrote in Paper Tigers, Hidden Dragons:

“Lenovo’s initial burst of innovation soon after its founding quickly dissipated in the wake of Lenovo establishing close connections to the Chinese state. [...]

Lenovo became more embedded in the state’s patronage network, winning a bid for a large national information technology tender in 1996. The state’s direct procurement jumped from an already significant 21 percent to a substantial 31 percent of Lenovo’s sales. [...]

Entering the state’s favor undermined the incentives to innovate because [Lenovo] realized that the firm could continue to enjoy state support even while forgoing the costs of technological development.”

Stronger government ties don’t have to mean increased complacency. Huawei’s example around the same time period demonstrates an alternate path. After then-president Jiang Zemin met with Huawei founder Ren Zhengfei 任正非 in 1994, the company started to receive increasing state support. As Eva Dou details in her book House of Huawei, this meeting marked a turning point in Huawei’s history. The Chinese government gradually reversed policies that had previously favored foreign telephone switch makers, leveling the playing field for domestic companies like Huawei.

But Huawei was also concerned about what the growing government support might mean:

“The growing interest from Beijing raised the specter that Huawei could be nationalized. Becoming a sluggish state-owned enterprise was something Ren was anxious to avoid. Others also warned him against it. “You must never become state-owned,” one official told him. “If that happens, you guys will die.” After Vice-Premier Zhu’s visit [in 1996], when his aides followed up to offer more loans, Ren politely declined.”

But unlike Lenovo, Huawei maintained a strong focus on R&D (a commitment to invest at least 10% of revenue in R&D), even when government priorities didn’t align with its goals. As Fuller describes:

“Huawei did not allow the state to undermine or distort its technological learning mission. For example, Huawei made a specific commitment to developing 3G WCDMA (Wideband Code Division Multiple Access) technology in 2002 even though the Chinese state kept delaying 3G rollout. With a distinct lack of enthusiasm from the Chinese state, Huawei had to undertake such R&D all on its own and spent 6 billion RMB on this research. Huawei actually suffered losses for the first time in 2002 because of this bold move, but it paid off in terms of Huawei being forced to sell this 3G equipment internationally.”

Huawei’s commitment to innovation ultimately allowed it to emerge as a global competitor. However, its close ties to the Chinese government also brought international challenges, which brings us to the final potential downside of receiving “national champion” treatment in China.

Subscribe now

5. Geopolitical Backlash

If DeepSeek’s ties to the Chinese government deepen, it could face international scrutiny and sanctions. PolyMarket gives DeepSeek a 13% chance of . Foreign governments may also become concerned about DeepSeek’s data security, which may lead to the app being banned. DeepSeek appears to have left sensitive user data unprotected, and the Italian government is already probing DeepSeek’s data privacy practices. If DeepSeek were to receive Chinese government support, such worries would only worsen.

DeepSeek’s Government Relations Advice for Itself

We asked DeepSeek itself (R1, not the company) how it should navigate growing political interest. Here is what it has to say:

DeepSeek should strategically leverage state support — such as funding, compute resources, and government contracts — while preserving its autonomy and ensuring compliance with evolving regulations. It must align its growth with government priorities to avoid overreach that may lead to sudden crackdowns, as seen with companies like Alibaba. Negotiating terms of state support carefully will be critical to maintaining operational independence and protecting its culture of innovation and research excellence.

At the same time, DeepSeek must prepare for geopolitical challenges by diversifying its market strategy and avoiding over-politicization. Being framed as a symbol of China’s resistance to U.S. suppression could backfire, particularly if export controls or sanctions intensify. Instead, DeepSeek should focus on its technological achievements and remain agile to adapt to shifting regulatory, geopolitical, and market conditions. By balancing state support with strategic independence, DeepSeek can maximize opportunities while mitigating risks.

A Reminder: This Is All Speculative

As of now, there’s no concrete evidence that DeepSeek is receiving “national champion” treatment. A single meeting with Premier Li Qiang, while notable, does not confirm close government ties.

China watchers should keep an eye out for signs of deeper state involvement, such as:

• Government “golden shares” or board representation;

• Direct funding (e.g., from AI guidance funds);

• Preferential access to compute or data resources;

• Major government procurement contracts.

For now, this remains a hypothetical scenario — but one worth monitoring closely. It will also be interesting to observe how much of a say DeepSeek will have in this. If the state wants them to become a national champion, could Liang deftly say no?

1

The R1 technical report shows evidence of cost sensitivity. As noted by Wei Wang, “DeepSeek unusually disclosed technical trial details, admitting failed attempts to use Process Reward Models (PRM) and Monte Carlo Tree Search (MCTS) due to reward hacking and token space complexity, ultimately opting for simplified technical pathways.”

For further reference, here’s the R1 technical report:

Dario Amodei is the CEO of Anthropic. In today’s interview, we discuss…

• Whether an AI innovation race is inevitable between the US and China,

• How the US should update export controls in light of DeepSeek’s R1 release,

• DeepSeek’s willingness to generate information about bioweapons,

• His message to Chinese engineers and DeepSeek itself,

• Technical defenses against model distillation and AI espionage,

• How advanced AI could eventually impact democracy,

• The tension between export controls and believing that AI will broadly increase human flourishing.

Have a listen on Spotify, iTunes, or your favorite podcast app.

Export Controls and AI Benevolence

Jordan Schneider: Let’s start with a brief picture of how rapid AI progress could cash out in terms of national power.

Dario Amodei: A few months ago, I wrote an essay called “Machines of Loving Grace,” where I focused on many of the positive applications of very powerful AI. I had this definition of what I think very powerful AI will look like. I use this phrase, “country of geniuses in a datacenter,” to describe what all the companies are trying to build. The phrase is very evocative for understanding the implications.

It’s as if you drop down a country of 10 million people, except all of them are polymathic Nobel Prize winners in any field. What does that do to national power? Presumably, that does a lot of things to national power. It greatly accelerates economic ability. It greatly accelerates science. Perhaps, unfortunately, it probably also has implications for intelligence and national defense — both in terms of controlling swarms of drones or analyzing intelligence information. Generally having a lot of incredibly smart entities who can control anything in a virtual way — that’s going to be a source of a lot of power in many different ways.

Jordan Schneider: Why write an essay about export controls?

Dario Amodei: Seeing the reaction to DeepSeek and being in the industry, being one of the people who’s developing the technology, I saw a lot of things that were just not correct. They came from people who weren’t following closely the actual arc of the technology but were now paying attention because there was this novel fact that it was a model developed by a Chinese company. They had missed many of the previous steps and misunderstood the dynamics of the field. They said, “Oh my God, this is really cheap,” and maybe they had a stereotype of cheap things being made in China and pattern-matched to that stereotype.

The reality, as I said in the essay, is that there’s been an ongoing trend of cost decreases in this field at the same time as we’ve spent more and more money to train the models. They’re so powerful and so economically useful that the countervailing trend of just spending more to make a better and smarter thing has outstripped the trend of making things cheaper.

The timing has been such that DeepSeek was able to release something that genuinely had some important innovations in it but was along the same cost-decrease curve that we’ve seen for AI in the past. This wasn’t making something for 6 million that cost the other companies billions. This was more like we’ve seen a 4x cost decrease per year — roughly that, give or take, relative to models that had been trained half a year to a year ago.

We’ll see that models of that quality will be now produced very cheaply by multiple players, while multiple players, including probably DeepSeek, will spend much more money to train much more powerful models. The new fact here is that there’s a new competitor. In the big companies that can train AI — Anthropic, OpenAI, Google, perhaps Meta and xAI — now DeepSeek is maybe being added to that category. Maybe we’ll have other companies in China that do that as well. That is a milestone. That is something that didn’t happen before. It is a matter of concern to me, but people were way overreacting and misunderstanding the implications.

Jordan Schneider: Your big update is on the skill of one Chinese player and perhaps more Chinese players. What, if anything, should people reassess specifically on the model side about what you can expect going forward about the gap?

Dario Amodei: I want to be clear — we’ve been tracking DeepSeek for a long time. We’ve been aware of DeepSeek as the likely most capable player in China for over a year. That’s informed our view of how things are going to evolve. For those who are just seeing DeepSeek, the update is, there were three to five companies in the US who could make frontier or near-frontier models. Now there are three to five companies in the US and one company in China. Whether they continue to make near-frontier models depends upon how many chips they can get access to and whether they can get access to chips at a much larger scale than those they’ve been able to get access to now.

Jordan Schneider: The AI safety community, yourself included, warned for years about the risks of racing dynamics. I’m curious if you could talk me through how you got to your current view on export controls.

Dario Amodei: The two things aren’t really opposed. I’m still concerned about the idea of a race in particular. My worry is that if the US and China are equally matched in terms of this technology, neck and neck at every stage, then there will be basically nothing preventing both sides. It will completely rationally make sense for both sides to keep pushing the technology forward as it has so much economic value, so much military value.

Absent very strong evidence of danger, there’s going to be a fair amount of incentive to continue developing it. In particular, a worry I have is about legislation in the United States. Like efforts to measure and perhaps at some point even restrict the risks of AI systems — the dangers, the misuse of them by individuals, for example, for bio attacks, or the autonomous dangers of the systems themselves. There have been various legislation in the US over the last year or so. One argument that’s been given against the legislation is, “Hey, if we slow ourselves down, China will just jump ahead and beat us.” That argument is absolutely correct.

The only way that I see that we can kind of do both — we have to make sure that we’re ahead of China and other authoritarian countries both because I don’t think they would use powerful AI very well and because if we’re not ahead, there’s this racing dynamic — yet somehow we have to also protect against the dangers of AI systems we ourselves build. My guess is the best way to do this, and it may be futile, it may not work out, but my guess is that the best way to do this is with something like export controls. We can create a margin between us and China.

Let’s say we’re two years ahead. Maybe we can spend six months of those two years to ensure that the things we build ourselves are safe. In other words, we’re still ahead and we’re able to make things safe. If things are evenly matched, then we have to worry that what they build isn’t safe, and at the same time we have to worry about them dominating us with the technology. That puts us in a very bad dilemma where there are no options.

For a long time — this is not anything new, this is just the first time I’m talking about it — I’ve felt that keeping the US ahead of China is very important. It’s in tension with this idea of wanting to be careful in how we develop the technology. But the great thing about export controls is that it almost pushes the federal frontier outward because there are two ways you can stay ahead: you can accelerate a lot, or you can try to hold back your adversary.

I think we’re going to need to do some amount of acceleration, but it’s a move that has trade-offs because the more of that you do, the less time you have to be careful. But I think one way to get around the trade-off to some extent is to impose these export controls because they widen the gap and give us a larger buffer that we can use to govern the technology ourselves. It’s tough though. It’s a tough situation. It’s hard to have both. There are real trade-offs.

Jordan Schneider: Let’s stay on this for one more second. Can you tease out getting to a race-to-the-top logic with China and getting to a race-to-the-top logic with peers in the West?

Dario Amodei: The nice thing about the other companies in the West is that there’s a coordination mechanism between them. They can all be brought under the same law. If someone passes a law, we have to follow it. OpenAI, Meta, Google, and xAI all have to follow it. You can know that you’re engaging in necessary safety practices and others will engage in them as well.

If you do something voluntarily to set an example, as Anthropic has often done, the threat of someone coming in and imposing legislation, even if there isn’t any, is often very valuable in getting other actors to behave well. Because if it looks like one actor is much more responsible and other players are behaving irresponsibly, that really creates incentives for regulators to go after the other players.

But I don’t think that’s possible between the US and China. We’re kind of in a Hobbesian international anarchy. I do think there are opportunities to try to cooperate with China. I’m relatively skeptical of them, but I think we should try. There’s not zero that can be accomplished. In particular, a thing that might change the game is if there was some kind of truly convincing demonstration that AI systems were imminently dangerous, like at the level of human civilization.

I don’t think we have that now. Some safety people think that they have arguments that are super convincing. They have arguments that are suggestive enough to make me worry and take it seriously and factor it into my calculus and do research on it. But certainly, on the merits, the quality of the arguments is not anywhere near strong enough to make two competing superpowers say, “Okay, we’re going to relinquish the technology or we’re going to pause for a certain amount of time to build it really carefully.”

As models become more powerful, I think we’ll understand the extent to which they really are a danger. There’s real uncertainty here. We’re going to learn a lot over the next year, year and a half. If we do find truly compelling evidence that these models are really, truly dangerous in a grand sense, maybe that could change things.

But so far, I’m aware of efforts by the US Government to send a delegation to talk to China about topics related to AI safety. My understanding, again, I obviously wasn’t part of those delegations, is that there wasn’t that much interest from the Chinese side. I hope that changes. I hope there is more interest.

We do lots of stuff in trying to educate the world about AI safety in order to make progress on doing a better job of steering AI systems. We do all this work in interpretability and things like constitutional AI and developing ways to scalably supervise AI systems. We can only hope that these ideas diffuse everywhere, including to China. Just to say it, I’d be happy to have DeepSeek copy these methods — I hope they do. There are ways of making AI systems more reliable.

We should be realistic about the actual situation. We should have a realist view of how international relations work, which is that it’s possible to make progress at the margin, and if some really compelling evidence were available, then maybe it could change things. But look, you have two nation-states that have very different systems of government that have been adversarial for a long time. They’re going to compete over this technology and they’re going to race to build it as fast as they can. That’s what’s going to happen by default.

Jordan Schneider: The only happy, analogous story I can find is Xi deciding to care about climate change. But the thing about climate technology is it’s not super dual-use. There’s economic upside, but not military upside to making a lot of solar panels and electric vehicles.

Dario Amodei: My understanding is China hasn’t always kept its commitments here. There’s been some mix of yes and no. That’s kind of what I would expect here. It’s even a little bit what we see between companies — there’s no law forcing people to adopt RSPs. Companies have voluntarily adopted them.

I think there are differences in the extent to which companies comply with them. Something I could imagine being realistic is yes, there are treaties or kind of aspiring goals signed here and maybe things get 10% safer because of them and we should absolutely seize that 10%. But it’s going to be hard. Again, I would point to, if you really wanted to try to drive cooperation here, you just need really strong evidence that something is a threat to the United States, China and every other country in the world. You need really compelling evidence. What I would say to safety folks who want us to pause, who want us to stop developing the technology: Your number one task should be to develop that evidence.

Jordan Schneider: What chips should and shouldn’t America be selling to China?

Dario Amodei: First, the export controls were never really designed to prevent DeepSeek or any other Chinese company from getting the number of chips that they had at the level of a couple tens of thousands. We should reasonably expect smuggling to happen. DeepSeek probably had a mix of different kinds of chips for about 50,000. Export controls can be more successful at preventing larger acquisitions — they can’t have like a million chips because that’s easily in the tens of billions in economic activity, approaching 100 billion.

Regarding specific chips, there were three chips that they were reported to have. I’m going by Dylan Patel of SemiAnalysis. None of this is confirmed — DeepSeek hasn’t stated it themselves. Generally, SemiAnalysis, when it relates to semiconductors, is pretty well sourced.

The claim is they had a mix of H100s, which are the standard chip that we have in the US — something like 10,000 of those. Those must have been smuggled in some form or maybe were proxied through some country other than China where they’re allowed but were used by a Chinese shell company.

The H800 is interesting. It was developed after the first rounds of export controls in 2022. Those export controls put a limit on the combination of the computing power of a chip and the communication bandwidth between chips. The H800 was made to reduce the communication bandwidth between chips to get around the export control. A bunch of H800s were sent to China, and then that loophole was closed in 2023. We actually noticed quite early that there are ways to train your model that get around the lower communication bandwidth. In 2023, a set of export controls were designed that were more robust to being evaded in this way. Presumably, those H800s were shipped between the time of the 2022 ban and the 2023 ban.

The H20 is a chip that is not suitable for training. It therefore must not have been used to make the Base V3 model, which was the first stage that I talked about in my blog post. In the second stage, the R1 model, which was stacked on top of the first and blew up on social media, involves a mix of training and what’s called inference. The H20 chip has not been banned and is good for inference, but not for training. It is possible that the H20 played an augmentative role — that it was used for part of the second stage, even though it could not have been used for the first stage.

Given that inference is increasingly involved in new models, it’s associated with this new paradigm of reasoning models that’s been developed. I would recommend also banning the H20 and attending to making these bans as broad as possible. We always want to be careful — we don’t want to ban other economic activity. For example, there’s no reason to ban sending gaming consoles to China. That would be too much of an economic imposition for something that’s aiming to be targeted. These things have to be designed carefully, but we should ban the H20.

Jordan Schneider: The argument Nvidia makes is the fewer chips we sell into China, the more demand there is for Huawei and the more customers they’ll have. There’s always an announcement that DeepSeek is running super efficiently on Huawei 910Bs or what have you. Thoughts on Chinese domestic capacity to produce chips and what implications are there for the semiconductor manufacturing equipment side of the export controls?

Dario Amodei: Over the timescale of 10 to 15 years, that’s probably correct — they probably will catch up. But the supply chain is really deep there, and as you mentioned, we’ve also export controlled semiconductor manufacturing equipment and repairs on that equipment. It’s actually going to be difficult to make chips that are competitive with say the new Nvidia B100 or with the chips we’re using like Trainium and TPU. The software ecosystem is also not up to snuff — that’s a less important constraint than the hardware.

My sense is that it is unlikely that the Huawei chips become anywhere near comparable to US chips anytime soon. As I wrote in “Machines of Loving Grace” and in the post on export controls, the critical period here where there’s really going to be contention, or where it’s important to achieve a balance of power, is going to happen in 2026, 2027, or at the latest, 2030. Policy should target that time range. Things are moving very fast in the AI world — 10 to 15 years is like an eternity. It’s forever. It’s almost irrelevant.

Jordan Schneider: The Biden administration’s policy was basically to let Chinese firms use chips on Western hyperscalers outside of the PRC. Is that kosher in your book?

Dario Amodei: Near the end they put in place the diffusion rule, which prevents that when it’s Chinese cutout companies. You may be referring to something different.

Jordan Schneider: Like ByteDance being Oracle’s largest customer.

Dario Amodei: The diffusion rule prevents or limits that within tier two countries outside the U.S. Some of it is done in the U.S. as well. It’s of less immediate concern to me because it’s farther up in the supply chain. You can yank the access to the chips instantaneously when they’re doing that. Once the model is trained, of course you have that model. The issue with the chips is once you have the chips, once you have enough of them, you can make model after model. It’s even more so with the semiconductor manufacturing equipment and the fabrication facilities. I’m less immediately concerned about it. Before we get to the end of this, it’s something that needs to be stopped or at least stopped at scale.

Jordan Schneider: What do you make of DeepSeek releasing its model as open source?

Dario Amodei: There were several different properties of the DeepSeek release. One was that the weights of the model were released. Another was that the model was an efficient and strong model on the cost reduction curve, but it’s the first time the point on the cost reduction curve was made by a Chinese company.

The second of those is much more important than the first. Most of what I said in the post on export controls would have been almost entirely the same, maybe entirely the same, if it had just been a model that they were serving via API. Most of the implications just come from the fact that this is a strong model — this is the first time that a Chinese company has produced a strong model. That model will scale up to a very large scale. Whether Chinese companies are able to scale up to millions of chips will be determined by the export control.

On the commercial side, we’ve found that our main competitors are people who release strong models, whether they’re open weights or not. How strong a model is accounts for about 80 to 90% of how much it has mattered in competing against the model. Open weights is different for models than open source — there isn’t source code, there’s just a bunch of numbers. Some of the advantages and differences that are present there are just not present here as much. The analogy breaks down.

Companies that have a history of starting with open weights, at some point they need to monetize, at some point they need to make profit. They often stop doing this. The more important factor here is that a Chinese company is producing a powerful model. Just looking in the marketplace, we haven’t seen any evidence that people prefer models because they’re open weights separately from the behavior of the model. Sometimes you see it, there’s a little bit of a convenience factor.

I said this as early as my Senate testimony in 2023 — open weights models are fine up to some certain scale. They’re not substantively different from closed-weight models.

Jordan Schneider: But you take a margin on every API call, right?

Dario Amodei: The interesting thing is, any kind of model, no matter where it is, has to be served on a cloud. That ends up getting associated with a margin. You have this present either way. At the same time, there are vast differences in the efficiency of inference and there are vast differences in the training of models. The training of models is moving very quickly.

This is actually a relatively small factor. There are weeks when we implement like a 20% improvement in inference efficiency or something like that. The companies are all competing with each other to have the most efficient inference. Most of this is swamped by who has the most efficient inference, who’s training the best model. If a company in China is very good at serving their model for low cost, then that’s an area where competition will happen. Whether the weights are available or not is mostly — not entirely, but mostly — a red herring.

Jordan Schneider: Somewhat related, when do you think governments are going to start getting queasy about models getting open-sourced?

Dario Amodei: From a business perspective, the difference between open and closed is a little bit overblown. From a security perspective, the difference between open and closed models is, for some intents and purposes, overblown. The most important thing is how powerful a model is. If a model is very powerful, then I don’t want it given to the Chinese by being stolen. I also don’t want it given to the Chinese by being released. If a model is not that powerful, then it’s not concerning either way.

Jordan Schneider: Speaking of being stolen, anything you want to say about model distillation?

Dario Amodei: There are these reports that, as I said in the blog post, I don’t really comment on one way or another, that DeepSeek distilled model from what might have been OpenAI. They claim they have evidence. I actually haven’t looked closely at it and can’t tell you whether it’s accurate or not. Distillation is certainly something you can do with a model, so it’s possible this is something that has occurred. It’s of course against terms of service.

There are a couple of points that are important. One is developing ways to detect model distillation. It’s probably possible to look at a model and looking at another model which is purportedly distilled from it, to look at the two together and say whether this model has been distilled from that. I’ll generate a bunch of output from one and generate a bunch of output from the other and try to figure out whether one model has been distilled from the other. You can already kind of see when one model has been distilled from another — they have similar mannerisms, they talk in similar ways. There are lots of behavioral signatures. It’d be great to turn this into some kind of measurable statistical test. Then there are monitoring techniques to prevent distillation.

That said, like all problems in cybersecurity, it’s going to be an ongoing issue. That’s why some companies have chosen to kind of hide the chain of thought in their reasoning models, because it makes it much harder to distill. Those can be jailbroken. But folks are working on antidotes to jailbreaking. We just released something today that makes it much harder to jailbreak models.

Espionage Prevention and Cross-border Collaboration

Jordan Schneider: Setting distillation aside, when you were on the Dwarkesh podcast about a year and a half ago, you said that if it’s a state’s top priority to steal weights, they will. If we’re living in your timeline and this stuff becomes the most important thing on the planet and it attracts more attention from Beijing and other state actors, what does that do to the gap, if any, between the leading labs and everyone else?

Dario Amodei: The biggest possible gap I can imagine is a couple of years. That’s actually worth a lot for the amount of advantage it gives — it provides a buffer to us here in the US and its allies to address some of the safety problems of AI systems. A gap greater than two years isn’t feasible. We should aim for a gap of two years.

The idea of preventing state actors from stealing the most powerful models for two years is challenging — quite challenging. But it’s possible. The way to do it is to enlist help from two powerful forces: One is the United States government, and the other is the models themselves. As the models get better at everything, one thing they’re going to get better at is cybersecurity — using the models to defend themselves. Second, using the United States and its counterintelligence capabilities to prevent the models from being stolen.

Our responsible scaling plan, our RSP, which we developed about a year and a half ago, anticipates this. It has different levels.

Currently, we’re at ASL2, which is strong but relatively ordinary for tech company security measures. The next level, ASL3, is preventing the models from being stolen by strong non-state actors. ASL4 and ASL5 are preventing theft from state actors. Those are very high bars to clear. When the model reaches certain levels of capability that we’re specifying, we have to put in place these strong security measures. A lot of what we were thinking about was this competition between the U.S. and China and the need to improve lab security.

Jordan Schneider: If the US government can’t keep Gina Raimondo’s emails safe and the Chinese can listen into every phone call in America, it seems like a tough thing to hang your hat on. Maybe another way to ask it is: say you steal the model weights — how much tacit knowledge is necessary that a DeepSeek or a ByteDance may or may not have with Opus 5 in order to make the most out of it relative to what the makers can?

Dario Amodei: Running that particular model is not that difficult. Using it to develop the next model is substantially more difficult because it’s optimized to run in some other cluster and setup in some other place. There are other things you can imagine doing.

Share

Jordan Schneider: We’ve talked about competition from the lens of compute, from the lens of models, but models are created by people. The times I’ve been to NeurIPS, it has been really striking to me just how much Mandarin is spoken at those poster sessions. I’m curious what message you’d have to say to PRC nationals studying in the West who hear this and are like, “Why would I want to work for this guy?"

Dario Amodei: I want to be really clear on one thing — I should probably say it even more than I do. When we talk about China, this isn’t about Chinese people versus American people. Speaking for ourselves, but I suspect the other US companies would say the same, we’re excited to have talent from around the world. We have no beef with talented researchers and engineers, no matter where they work. We’re from the same community they are. If there’s some kind of cooperation to be had here, it’s probably through this kind of track-two researcher-to-researcher thing. We welcome these folks absolutely, as much as we can.

The concern here is authoritarian systems of government, wherever they exist. We could have seen in the last 10, 15 years, China could have gone down a very different route than they did. I’m not a China expert, but many people do seem to think that there was a bit of a fork in the road and maybe an opportunity for them to take a more liberalizing path. For whatever reason, that didn’t happen. But if it had happened, certainly my view on all of this would be completely different. This is not about animus against a country. This is about concern about a form of government and how they’ll use the technology.

Jordan Schneider: Well, now this interview is definitely not going viral in China. Thank you for that. Dario, anything you want to say to DeepSeek?

Dario Amodei: They seem like talented engineers. The main thing I would say to them is take seriously these concerns about AI system autonomy. When we ran evaluations on the DeepSeek models — we have a spate of national security evaluations for whether models are able to generate information about bioweapons that can’t be found on Google or can’t be easily found in textbooks — the DeepSeek model did the worst of basically any model we’d ever tested in that it had absolutely no blocks whatsoever against generating this information.

I don’t think today’s models are literally dangerous in this way. Just like with everything else, we’re on an exponential, and later this year, perhaps next year, they might be. My advice to DeepSeek would be take seriously these AI safety considerations. The majority of the AI companies in the US have stated that these issues around AI autonomy and also these issues around AI misuse are potentially at least serious and real issues. My number one hope would be that they come work in the US, they come work for us or another company. My number two hope would be that if they’re not going to do that, they should take some of these concerns about the risks of AI seriously.

Jordan Schneider: The dominant framing, how you’ve written and other folks in AI policy circles is that AI is going to be something that entrenches autocrats. People are talking about the CCP like it could turn into a thousand-year Reich. Can you see a world in which AI, and perhaps particularly open source AI, is actually sort of like a democratizing force?

Dario Amodei: I actually can and I wrote about it a bit in “Machines of Loving Grace.” I don’t really think it’s about open source or closed source. I could imagine a world where China makes powerful AI that’s open source, but they have the biggest clusters, they have the biggest ability to fine-tune the model. If I’m a citizen of an autocracy, does a flash drive with some model weights on it help me resist the autocracy? There’s this almost impressionistic connection between “oh, there’s this thing that’s available to everyone” and it’s like a freeing resource. I don’t really see that connection.

It’s similar to saying like, suppose I had the code for TikTok or the code for Twitter or something. Would that change the effects that social media had on all the people around? I don’t think that it would — this all comes from things operating at scale. What does matter is how we use the technology, whatever form it’s been produced in.

As I wrote in “Machines of Loving Grace,” I had these ideas for using AI to strengthen democracy. It’s more around the administration of justice. AI may open up the possibility of having a more uniform, more fair system of justice. The justice system often involves making judgment calls — things that are somewhat subjective and typically have to be done by a human. People worry that if you do it by an algorithm, it’ll be less fair. But if we apply it correctly, AI systems could create a fairer society — one in which people are more likely to receive equal justice under the law.

Around deliberation and democratic decision-making, there are opportunities for AI to help. We did a collaboration with an organization called Polis that used AI systems to aggregate public opinion and facilitate discussions between people to find areas of common ground. That’s something healthy for a democracy. Finding areas of common ground is really something that we want to do.

I also suspect that improving our science, improving our health, and improving our mental health across society may in lots of subtle ways improve the effectiveness of our deliberative decision-making.

Jordan Schneider: You wrote two blog posts and they seem to be in tension with each other at some level. One-sixth of humanity lives in China. How do you think about that in the context of your sort of vision of AI enabling more human flourishing?

Dario Amodei: I actually think these are part of a single worldview. In fact, in “Machines of Loving Grace,” in the section on AI and democracy, I say we need to lock down the supply chain. Export control is obviously part of what I meant by that, lab security and semiconductor equipment being other parts of it. These actually are deeply compatible.

First of all, I think it’s perfectly possible to distribute the benefits of AI to all the world, including China, including other authoritarian countries, without distributing the military capabilities. If you have a powerful model and you’re just handing over the powerful model, then yes, they all come together because it’s a generalized dual-use technology. But you can take a model, you can run it via API, and you can serve it to China for drug development, for the next cancer drug. You can serve it to China for more efficient energy production. Basically, you can rent very large amounts of the model’s time to someone in China for these economically productive activities.

I think that’s a useful form of trade. That should happen. Through that, you can have all the benefits of AI shared, but at the same time, you can just block usage of the model if someone’s asking about how to make a hydrogen bomb with the model, or someone’s asking how to find and target nuclear submarines, or R&D for orbital weapons or something. You can block those applications while allowing the ones that are really good.

Now, there’s kind of a bigger picture here, which is that in the long run, we’re going to want all the uses of AI to go to everyone. We’re going to have to work out, in the long run, some kind of international governance of the technology, some kind of stable equilibrium. Because I said that the US lead at most will be two years. My hope is that that will go better if the US is leading the setup of that international regime. It will go better if we’re able to negotiate for safe deployment for everyone from a place of strength. If we’re doing it from a place of weakness, the worry is we’ll just get dominated.

I think the US is in a position to make some more magnanimous decisions here if it has the lead. We should start thinking now about what we should do. The plan isn’t just like, crush our adversaries. I don’t think that would even work. The plan has to be some version of, start from a position of strength, and work out how this technology can benefit the whole world and how its downsides can be mitigated.

Jordan: Dario, if this AI stuff doesn’t end up working out for you, I think you might have a chance to make it as a substacker. Thanks so much for being a part of ChinaTalk.

Mood Music

With DeepSeek’s recent breakthroughs dominating headlines, a deeper question remains: is the Chinese government truly pushing toward AGI? And if so, what role if any might it have in allocating compute?

While DeepSeek’s accomplishment suggests that access to compute beyond what a hedge fund can cobble together may not be an immediate bottleneck for developing algorithmic advances, the picture is more complex. Western labs are still racing to build massive 100,000-GPU-scale clusters (not to mention the Stargate Project) in the hopes of developing and deploying AGI at scale, and DeepSeek’s CEO has repeatedly said that he is compute constrained.

In this article, we unpack a key paradox: how can China have simultaneous compute overcapacity and shortages? And what does this say about whether China is ready for the next step towards developing and deploying AGI?

Since early 2024, reports have emerged of idle computing centers in China. And in September, renting Nvidia AI chips in the cloud was cheaper in China than in the US.

At first glance, this doesn’t make sense. Since the US restricted China’s access to advanced AI chips, one would expect to see chip shortages, every available chip used to its fullest, and high cloud prices.

So, what’s going on here?

The puzzle has also attracted the attention of Chinese journalists. In this post, we synthesize takes from two long Chinese articles from November of last year.

• China’s Computing Power — Is There Overcapacity? 中国算力，过剩了吗?, written by Wu Junyu 吴俊宇 for Caijing Magazine 财经杂志. (link, archived link);

• Too Many AI Computing Centers, Yet Not Enough Large Models 智算中心太“多”，大模型不够用了, written by Zhang Shuai 张帅 for TMTPost 钛媒体 (link, archived link).

Key takeaways

• China has added at least 1 million AI chips (70% Nvidia, 30% Huawei) in 2024 to its compute capacity — enough to build, in theory, several 100,000-GPU-scale clusters (which is what would likely be needed to compete with GPT-5 or other next-gen Western models).

  • But how come even in the current 10,000-GPU-scale cluster era China already sees overcapacity?

• Explanation 1: chips are poorly deployed, leading to shortages for “high-quality compute” but overcapacity for “low-quality compute.”

  • During the AI boom, many companies and local governments rushed to buy GPUs, assuming their mere ownership would ensure profits.

  • Since many of them lacked technical and market expertise, however, chips ended up in geographically dispersed, low-quality data centers that are either difficult to use or in places without demand. This speculative behavior and resource mismanagement have given rise to idle computing centers despite overall shortages.

  • The government has responded by no longer granting permits for new data-center construction, unless they are located in one of eight designated hubs. Industry consolidation and professionalization make it less likely that the same kind of chaos will hamper China’s data center construction in 2025.

• Explanation 2: short-term overcapacity, long-term shortages.

  • By 2024, the demand for foundation-model training has significantly slowed. Many of the players that rushed to develop foundation models in 2023 have ceased training new generations.

  • Although inference demand is growing rapidly, it remains insufficient to fully absorb the excess capacity previously dedicated to training.

  • In the future, rising inference demand and the development of larger foundation models are expected to drive new shortages. The current “transition period” of temporary overcapacity is anticipated to end in mid-2025.

Below, we go into more detail on what the Chinese media has to say on

• How many AI chips China has, and who is deploying them;

• The emergence of “fake” and “pseudo” 10,000-GPU clusters, and how the government is responding;

• Unique challenges for compute clusters using Huawei chips;

• The transition from training to inference demand;

• Whether China needs more public cloud computing and fewer private GPU clusters;

• How China can avoid pitfalls from the 10,000-GPU-scale era in the 100,000-GPU-scale era.

Who are the major players investing in AI computing centers?

To start with, Caijing gives an overview of who actually builds compute clusters in China. There are three major groups:

1. large technology companies (Alibaba, Tencent, Baidu): projected combined 2024 capex of over 130 billion yuan;

2. state-owned telecom operators (China Mobile, China Telecom): planned 2024 investment of 84.5 billion yuan;

3. local governments and SOEs: statistics are very incomplete, but at least 27.5 billion yuan investment from January to October 2024.

Caijing goes in depth on the data to estimate the total number of chips put into operation by these three groups in 2024.

Large technology companies

Caijing argues that capital expenditure is a good proxy for computing-center investments, since capex is usually used to purchase chips and lease land.

In 2024 H1, the total capital expenditures of Alibaba, Tencent, and Baidu reached 50.44 billion yuan, up 121.6% y-o-y. This is the highest growth rate since 2018. If the growth trend in the first half of the year continues, the total capital expenditure of Alibaba, Tencent, and Baidu will exceed 130 billion yuan in 2024.

Note that Huawei and ByteDance also operate major computing centers. But because they are not listed companies, they have not disclosed relevant data — which means the total investment in the “large tech companies” category is actually higher than the data presented here.

Expect DeepSeek in the coming months to shack up with one of these firms just like OpenAI did with Microsoft. TMT 钛媒体 reported on Monday that ByteDance is “considering research collaboration” with DeepSeek. Partnering with ByteDance, however, could be an enormous unlock for DeepSeek researchers, giving them access to orders of magnitude more compute.

Figure 1: Capital expenditures of Alibaba, Tencent, and Baidu (Q1 2021 – Q2 2024) Unit: 100 million yuan

In the first half of 2024, Nvidia reported US$6.16 billion in revenue from the Chinese market (approximately 43.7 billion yuan), a year-on-year increase of 42.3%. Since domestic tech companies still largely rely on Nvidia chips, these figures provide another useful proxy for estimating the volume of chips purchased by major tech firms in China.

Figure 5: Nvidia’s global revenue scale in various regions (Q1 2022 - Q2 2024)
Unit: US dollars

State-owned telecom operators

Caijing explains that since 2023, state-owned telecom operators have shifted focus from investment in 5G to computing center investments. In fact, the scale of their investment does not trail too far behind that of the private tech companies listed above.

China Mobile and China Telecom both disclosed 2024 computing-power investment plans, with China Mobile investing 47.5 billion yuan and China Telecom investing 37 billion yuan. The total of the two — 84.5 billion yuan — represents a year-on-year increase of 13%. China Unicom did not disclose this data, making data in this section incomplete as well.

Figure 3: Telecom operators’ computing power-related capital expenditures (2022-2024)
Unit: 100 million yuan

China Mobile has publicly released two procurement calls for public bidding. The two procurement documents show calls for more than 9,200 Huawei Ascend 昇腾 AI servers (an Ascend AI server is usually equipped with 4 to 8 GPUs, so this would include around 36,000 and 74,000 GPUs). The winning bidders are well-known domestic Ascend dealers, such as Kunlun 昆仑, Huakun Zhenyu 华鲲振宇, Powerleader 宝德, Baixin 百信, Changjiang 长江, Kuntai 鲲泰, Xiangjiang Kunpeng 湘江鲲鹏, and Sichuan Hongxin Software 四川虹信软件. Some articles suggest that China Mobile’s computing infrastructure built in 2024 relies 85% on domestic chips.

Caijing cautions that even these domestic chips rely on global supply chains:

A Huawei representative mentioned to us this September that although Nvidia’s H20 chips are still being sold to China for now, it is necessary to prepare for the worst-case scenario, as the supply chain could be disrupted at any moment. Huawei’s Ascend 910 series chips rely on high-bandwidth memory (HBM) supplied by South Korea's SK hynix. At present, domestic alternatives to HBM that are both effective and reliable are lacking. In an extreme scenario, if SK hynix’s HBM supply chain is cut off, the production capacity of the Ascend 910 chips would also be impacted.

For more on China’s international reliance on HBM, check out this recent ChinaTalk piece.

Local governments and SOEs

The data here seems to be most messy and incomplete. Caijing’s statistics show that as of October 16, 2024, there have been at least 30 bidding projects for AI computing centers launched in various cities, with a total investment of at least 27.5 billion yuan.

So how much compute do they have?

Based on the investment data provided above, Caijing provides a cautious estimate that China in 2024 has added

• 700,000 Nvidia H20 chips;

• 300,000 domestic Huawei chips.

The article also discusses a more optimistic estimate by Semianalysis:

• more than 1 million Nvidia H20;

• 550,000 domestic (e.g. Huawei) chips.

According to Caijing, however, most Chinese industry insiders do not find the latter estimate credible.

In either case, this would leave China with enough GPUs to build several 100,000-GPUs scale clusters. For context, however, Epoch AI estimates that Google alone operates over 1 million chips. (It’s important to note that while Epoch AI’s figure below reflects total capacity, the numbers for China above refer specifically to the additional capacity acquired in 2024, making the two figures not directly comparable.)

TMTPost estimates that in H1 2024:

Domestic intelligent computing centers delivered 1.7 billion card-hours, with 560 million card-hours in use, resulting in a utilization rate of 32%. Other data shows that the average rack utilization rate in the computing infrastructure industry is currently less than 60%.

Utilization rates vary by operator:

Increasing the proportion of compute resources provided through cloud services can effectively enhance the efficiency of intelligent computing power. Alibaba Cloud and Huawei Cloud have already submitted this suggestion to multiple government departments.

A reasonable utilization rate of public cloud services is between 40%-60%, the utilization rate of government cloud services ranges from 25%-40%, but the utilization rate of privatized computing resources is generally no more than 5%.

Fake 10,000-GPU clusters?

Simply owning GPUs is not enough. There were many issues with efficient and rational allocation when China built its 10,000-GPU-scale clusters in 2023 and 2024.

An Lin 安琳, director of the Alibaba Cloud Intelligent Technology Research Center 阿里云智能科技研究中心, notes that there are currently three types of “10,000-GPU clusters” in China:

1. Fake 10,000-GPU clusters (假万卡集群): This refers to companies that technically own 10,000 AI accelerators, but the GPUs are distributed across multiple data centers in different locations. Each data center may only house a few hundred or a few thousand GPUs, and while the total number may exceed 10,000, they cannot function as a unified cluster.

2. Pseudo 10,000-GPU clusters (伪万卡集群): This type of cluster involves 10,000 AI accelerator cards deployed in a single data center. However, while the infrastructure exists in one location, these GPUs are not used as a unified resource for training a single large model. Instead, a portion of the GPUs might be used to train Model A, and another portion could be used for Model B.

3. True 10,000-GPU clusters (真万卡集群): A single, unified cluster housing more than 10,000 GPUs in one data center. These GPUs are interconnected and can operate as a cohesive unit through advanced large-scale resource scheduling technologies. This allows a single large model to be trained across all 10,000 GPUs simultaneously.

Creating a true 10,000-GPU cluster involves overcoming significant technical challenges, such as:

• High-performance networking to handle enormous data flows;

• Efficient scheduling systems to maximize GPU utilization;

• Stable operations that can handle frequent hardware failures.

During the AI boom, many companies and local governments rushed to buy GPUs, assuming their mere ownership would ensure competitiveness in the AI race. Many of them lacked technical expertise to actually deploy them effectively, giving rise to “fake 10,000-GPU Clusters” and “pseudo 10,000-GPU clusters.” According to TMTPost:

Currently, stockpiling NVIDIA cards has indeed resulted in some computing-power waste. Many buyers lack the necessary networking, scheduling, and operational capabilities required for smart computing centers. A technical expert in the field bluntly stated, “There was too much speculative profiteering. Many people were not actually in this industry — they thought hoarding hardware could make money. They stuffed them into a random data center, but without solving issues like stability, fault tolerance, and other technical challenges. This caused a lot of waste.”

Some small cloud providers in China may provide cheap services simply because the quality is also low.

Government response

The Chinese government is acutely aware of the waste caused by chaotic data center construction in 2023 and early 2024. It has responded by trying to actively slow down the construction of new data centers.

According to TMTPost, the National Development and Reform Commission (NDRC) has stopped approving energy quotas for new data center construction. Exceptions may be granted if you use Huawei chips in one of the eight nodes of China’s “Eastern Data–Western Compute” 东数西算 project we explored in an article late last year.

Subscribe now

A front-page article in the party-media outlet Science and Technology Daily 科技日报 — titled “The Construction of Intelligent Computing Centers Should Not Blindly Follow Trends” — made this argument:

The operation and management of smart computing centers rely heavily on professional technical talent and efficient management teams. Without these, the centers may fail to function as intended, potentially leading to idle equipment and wasted resources.

Therefore, deciding whether to build, when to build, and where to build smart computing centers requires scientific and prudent decision-making. A “herd mentality” or a rush to “follow trends” must be avoided. The overarching principle should be to build smart computing centers based on clear and sustainable market demand, appropriately tailored to local conditions, and with moderately advanced planning.

Local governments are now raising the bar for contractors:

Some local governments have strengthened requirements for the operation of intelligent computing centers. For example, a project in Dezhou 德州, Shandong, which is valued at approximately 200 million yuan, explicitly stated in the bidding documents that it would adopt a “design, construction, procurement, and operation integrated model” 设计施工采购运营一体化的模式. It requires an operating period of no less than five years and specifies minimum annual computing power revenue after the project’s acceptance and commissioning.

According to ZStack 云轴科技 CTO Wang Wei 王为, the government now has higher requirements for smart computing centers. Previously, it was enough to simply build the centers, but now they look for competent operators or integrate construction and operation to ensure the utilization of computing power.

Other local governments are trying to hand over their idle computing resources to cloud providers:

Some local governments have begun mediating to encourage cloud service providers to rent idle compute power from their smart computing centers. “We didn’t even know there were so many GPUs in the country. In a sense, the scarcity of compute power is accompanied by a mismatch of resources,” an industry insider said.

Several other government policies try to address resource waste in small and scattered computing centers:

The “Special Action Plan for Green and Low-Carbon Development of Data Centers” 数据中心绿色低碳发展专项行动计划 imposes strict and comprehensive requirements on the regional layout, energy and water efficiency, and use of green electricity in the data-center industry. The plan also proposed the “complete elimination of local preferential electricity pricing policies for high energy consumption.” People generally believe that this policy will accelerate the elimination of outdated capacity, improve the industry’s supply structure, and promote healthy development.

The Ministry of Industry and Information Technology (MIIT) recently issued specific pilot approvals for cloud-based smart computing services in six cities. These pilots aim to address issues related to the earlier construction of smart computing centers in various regions, particularly the waste of resources in small and scattered computing centers built with state-owned funds.

Apart from these government measures, the articles also suggest that as the market matures and investment strategies adapt, low-quality or inefficient actors (e.g., those deploying suboptimal or idle systems) may naturally be phased out. Over time, more efficient players will dominate, addressing some of the issues with idle computing resources.

Are Huawei chips especially likely to lay idle?

The articles provide mixed messages on the role of domestic chips in creating the paradoxes of idle computing power.

According to Caijing, the clusters most acutely affected by the paradox of idleness are the ones operated by state-owned mobile operators and by local governments. Part of the reason is that they are using domestic chips:

The situation with domestic AI chips is even more special. Currently, domestic AI chips have only achieved “usability” 能用, but are still some distance away from being “good to use” 好用. Blindly using them would only result in wasted computing power. Companies need to use various technical methods to adapt and maximize the efficiency of domestic AI chips.

TMTPost also describes the ecosystem challenges faced by Huawei chips, but quotes industry insiders as believing those are minor issues that are inevitable and will ultimately help the Huawei ecosystem to mature:

One side argues that domestic intelligent computing centers are still reliant on overseas ecosystems, requiring a transition period of three to five years. During this time, large-scale rapid construction of such centers is likely to lead to significant waste. … The expert remarked on the inefficiencies in domestic AI computing, observing, “Huawei’s operational capabilities are incredibly strong. However, before users were ready to adopt domestic GPUs or Huawei solutions, Huawei invested significant resources to develop computing clusters and intelligent computing centers. Telecom operators built clusters with tens of thousands of GPUs, but there’s still a gap between making the hardware available and actually using it effectively. As more domestic chips enter the market, these inefficiencies may become even more pronounced.”

The other side believes that overseas restrictions will only become more severe, and that the domestic AI computing ecosystem must mature faster. Compared to strategic national competition, some minor issues caused by overly fast construction are acceptable. … “That said, I’m optimistic about domestic GPUs overall, given the shifts in computing power trends in the era of large models. Previously, AI models were very fragmented, and Nvidia’s CUDA ecosystem was dominant because it had to support so many different models. Now, with large models becoming more consolidated, mainstream frameworks are more uniform. At the same time, Nvidia GPUs are extremely expensive, and with challenges in accessing their computing power, more people will be willing to experiment with domestic GPUs,” the expert added.

Transition from training to inference demand

In 2023, the rush of countless companies to develop foundation models spurred the construction of numerous computing centers across the country.

In 2023, there was a compute shortage because the “battle of the hundred models” led to a surge in demand for training compute. US export controls further restricted supply, and panic purchasing by companies exacerbated the imbalance between supply and demand. A strategic planner at a leading tech company mentioned in August 2023 that their company, in 2023, had purchased A100/A800 and H100/H800 chips on the market at prices 1.5 to 2 times higher than Nvidia’s official pricing, even acquiring scattered stockpiled chips from small and medium-sized distributors.

By 2024, however, many of these companies have abandoned their foundation model ambitions, leading to a sharp drop in training demand. This suggests increasing resource concentration on a few dominant players in the industry. According to TMTPost, out of the 188 models officially approved in China by October 2024 [ChinaTalk editor’s note: this data is unsourced in the original article, so take it with a grain of salt]:

• Over 30% showed no further progress after approval;

• Only about 10% were still actively training models;

• Nearly 50% shifted their focus to AI application development.

Caijing agrees, noting:

The demand structure is changing. By 2024, compute became gradually sufficient as companies stockpiled more and more chips. The demand for model training began to slow, while the demand for application inference has not yet exploded. At this point, a transitional “gap period” emerged.

Currently, flagship models from domestic tech companies such as Alibaba, ByteDance, and Baidu have performance levels approaching OpenAI’s GPT-4. Since the next generation of models after GPT-4 has not yet truly emerged, the task of catching up with GPT-4 has largely concluded. As a result, major domestic tech companies have temporarily slowed their model training efforts.

Between 2022 and 2027, the proportion of training compute will decline to 27.4%, while the proportion of inference compute will rise to 72.6%.

However, inference demand is increasing. Baidu’s token consumption quadrupled from 250 billion in May 2024 to 1 trillion in August. ByteDance even reported a ten-fold increase for its model over the same time period.

Figure 7: Average daily token consumption of large models under Baidu and ByteDance
Unit: 100 million/daily average

All of this suggests that there will eventually be shortages again. Caijing:

We have learned from multiple technology cloud service providers and some telecom operators that their common judgment is “sufficient in the short term, but insufficient in the long term.” In the short term, the existing computing power can meet basic business needs, with localized instances of overcapacity. However, in the long term, the implementation of AI applications will require more computing power.

The length of this “transition period” depends on two factors: first, when the race to train the next generation of models begins; second, how quickly inference compute demand grows, which depends on the speed of AI application adoption and penetration.

Some technical experts in tech companies predict that this transition may last until mid-2025, but it will not be very long overall.

As Trump throws 10% tariffs on China and 25% tariffs on Canada and Mexico, it’s worth reflecting back on the long scope of America and trade.

Back in 2019, back when this humble substack was still called ChinaEconTalk, we interviewed the dean of US trade history Doug Irwin about his magisterial Clashing over Commerce: A History of US Trade Policy.

In this interview, we discuss:

• A historical tour through 250 years of American trade policy;

• Parallels throughout US trade history to recent policy;

• The evolution of GATT and the WTO;

• Whether tariffs now are likely to have the intended effect;

• The three R’s of tariffs: revenue, restriction, and reciprocity.

Transcript below, or a listen on iTunes or Spotify.

Jordan Schneider: James Madison holds a special place in my heart because I’ve acted him on stage. He wrote in “Federalist No. 10” that, “it is in vain to say that enlightened statesmen will be able to adjust clashing economic interests and render them all subservient to the public good.” Boy, was he right.

Over American history, trade policy stands out as a field of fierce political combat. What are the principles that drive American trade policy? What lessons from history can we draw on to understand today’s trade fights?

To discuss, we have on Dartmouth professor Doug Irwin. Doug is the author of a 700-page history of American trade policy from American Independence to the present day. Doug, welcome to ChinaEconTalk.

Doug Irwin: Thanks, it’s a pleasure to be here.

Jordan Schneider: You mention in your acknowledgments that you thank your wife for putting up with your interest in such an arcane field of economics. Maybe when you guys were dating, you probably gave her a pitch as to why you actually care about trade policy and why it’s interesting. How has that pitch changed over the course of your career?

Doug Irwin: There’s the personal level and then there’s the professional level. The secret is that I married an economist. It didn’t take all that much for her to be persuaded that trade is important.

Jordan Schneider: That’s cheating, man.

Doug Irwin: It is cheating a bit. We have a lot to talk about. She is a macroeconomist, however. She doesn’t really think micro is all that important or interesting. However, even she would grudgingly admit today that trade policy is important.

What’s amazing about the present era is that with President Trump and all the headlines about trade, one really doesn’t have to motivate the idea that trade policy is important and interesting for the world economy. If you go back 10 years or so, that wasn’t the case.

There was a period in my career when trade policy was a sleepy backwater of economics. Tariffs were low and stable. Countries weren’t using trade policy to achieve any particular objective. The world seemed to be moving in a liberalizing direction. There wasn’t much to say. 

Now we have all these trade policy experiments with tariffs going up on China, on steel and other things. Economists are writing tons of papers on these topics. It’s obviously making the front page news, above the fold even, in terms of some of the macroeconomic impacts of these tariffs. It’s disheartening in a way, but for a trade economist, these are really interesting times.

Jordan Schneider: From a drama perspective, how would you compare this to fights over monetary policy throughout history?

Doug Irwin: That would be another great book to write, the politics of monetary policy from the very beginning. We’re at a sort of local peak in terms of interest and dissension over trade policy in terms of what President Trump has been doing.

With the book I cover more than 200 years of US history. There was this period in the late 1820s when the US was literally threatening to break apart over the issue of tariff policy. The South objected to the high tariffs that were being imposed because the North sort of controlled the political process.

I don’t think we’re at the stage of breaking up the United States over trade policy. But it’s pretty intense at the moment with the debates going on about trade policy.

Jordan Schneider: I imagine that was one of your favorite sections to write. As you made it through the entire sweep, I’m curious if there were some low points as well as high points?

Doug Irwin: I didn’t set out to write this book or anything. I didn’t have this conception 10-15 years ago that I’d write a book to tell the whole story of trade policy from the very beginning to the very end. I backed into it.

I started writing papers on trade policy in the Great Depression. I found it very interesting. It was a lot of fun and not too many people were working in that area at the time. I said, let me go back a little bit further in time. How did we get to the Great Depression period?

Slowly I moved backward and forward from the 1930s. After a few years, I realized that I had covered some of the big events in terms of the history of trade policy. Maybe I should put it all together in a book and cover some of the gaps and fill in some of the holes. It came about in a haphazard way when I ultimately decided to write the book.

Of course, some periods are more interesting than others. The late 19th century from after the Civil War up to World War I was a pretty stable period. There are a lot of forgettable presidents. There was a lot of gnashing about trade policy, but not a lot of action, a lot of status quo bias if you will.

Jordan Schneider: But there’s a “Cross of Gold.”

Doug Irwin: That’s true. That’s more monetary policy than trade policy per se, but they’re related to some extent. That was one of the slower periods, but it’s also a very interesting period. It’s not so much because of the trade politics of the period, but because of the debates over what the impact of trade policy was in terms of fostering US economic growth and industrialization.

That’s really the big debate in the late 19th century, not big fights over whether tariffs should go up or down a little bit.

Tariffs as Revenue: pre-Civil War Trade Policy

Jordan Schneider: Let’s start quickly with the dynamics of pre-Civil War trade policy. Why do you argue that it’s dominated by concerns around revenue raising?

Doug Irwin: Actually, you asked me what were the more interesting and less interesting parts of the book. The earlier parts are really interesting.

One of the motivations for having the Constitutional Convention was to solve a policy problem: trade policy. When we got our independence, we had the Articles of Confederation which didn’t endow the National Congress with the power to tax anything. Yet the national government needed to spend money on defense, paying foreign debts, and what have you.

The Constitutional Convention was to solve that problem by giving the government the power to tax. By unanimous consensus, the idea was we’re not going to have income taxes. We’re not going to have sales taxes. You might’ve heard of the Whiskey Rebellion. Sales taxes didn’t work out so well.

Import tariffs were this anonymous way of taxing goods as they came into the US economy. I wouldn’t say we were dependent on trade, but we relied on trade quite a bit. We imported a lot of goods that we couldn’t produce here. They came into 6-12 ports on the East Coast.

That’s easily taxed. The taxes were embedded in the price of the product that the consumer pays. Everyone agreed that import tariffs were going to be the way to raise revenue for the federal government.

With the Constitution, we were having the first debates over what we wanted to achieve with our trade policy. The US was starting out on its own as an independent country and thinking about what its trade policy ought to be and what principles ought to be underlying it. 

Jordan Schneider: How is this history relevant today?

Doug Irwin: You would think that it wouldn’t be so much.

It is in two senses that are directly related to the debates that the Trump administration has unleashed. Under the Constitution, Article I, Section VIII, Congress is empowered with the authority to levy taxes on imports. What happened in the late 20th century is that Congress delegated a lot of those powers to the president.

Now the president, President Trump, has seized upon this to impose and justify the imposition of tariffs that he’s done, without Congress’s consent. There is a discussion today at least about the role of Congress and trade policy and the Constitution. That’s one way in which it’s played into debates today.

The other is that after the introduction of the income tax in 1913, the share of federal government revenue coming from the tariff just plummeted to trivial amounts, less than 1% of US government revenue today.

One of the president’s selling points is that his tariffs are raising a lot of revenue. He tweets a lot about how our coffers are being filled with millions, if not billions, of dollars. Other countries are paying for the privilege of sending their goods to the US.

He’s raised revenue as a reason for imposing tariffs today. That idea of revenue has come back into play as well. In looking at more than 200 years of US trade policy history, there’s a lot of recycling of arguments, a lot of reinventing the wheel. There are a lot of debates that take place today that took place, 50, 100, 200 years ago.

There are some underlying themes and issues that come up with trade policy that are always present with us.

Tariffs as Restriction: New Protectionist Politics

Jordan Schneider: Trump certainly takes cues from mercantilist British empire-inspired economics. Let’s now turn to the restriction phase, which you argue was the first fundamental shift in US policy.

Doug Irwin: One of the themes of the book is that when we ask what the federal government is trying to achieve in levying tariffs on imports, I have what I call the three Rs: Revenue, Restriction, and Reciprocity.

Revenue is straightforward: taxing foreign imports raises money for the federal government. 

Restriction aims to keep imports out to create jobs in import-competing industries and strengthen the American economy. That argument has been very potent and been around for a long time. The Trump administration points to this, saying we’re helping the steel industry, etc. You hear that in past decades as well.

Reciprocity, which we’ll discuss later, involves using trade policy to negotiate with other countries to reduce their barriers. You’ll reach these agreements to reduce trade barriers and increase bilateral or multilateral trade.

The transition from revenue to restriction took place in the 1860s.

One of the secondary themes of the book is that when determining what phase we're in or who's calling the shots in trade policy, it comes down to regional politics. Which regions benefit from trade or trade barriers? Which regions have the most political power in Congress and the federal government?

The reason we shifted from revenue to restriction in the 1860s is that prior to the Civil War, the South was politically dominant in the United States, particularly in the Senate. The South was responsible for most US exports at the time. Cotton was a big one in particular, but you also had tobacco, indigo, and other commodities. With their political power, they dictated relatively open US trade policies with low tariffs to keep exports up.

The Civil War caused a realignment of political power in the United States. The North became dominant, not just because the South left Congress during secession. Even after the war, the South was much less politically powerful than it had been. Import-competing manufacturing industries had arisen in the North in the first half of the 19th century. They wanted trade restrictions. They used their political power from 1860 until after World War I to establish high tariffs as the US trade policy regime.

Jordan Schneider: Doug, talk a little bit about logrolling. What was it and what impact did it have on US trade policy?

Doug Irwin: Logrolling is a political science term describing how votes take place in Congress. Political scientists and economists have used it to explain how the high-tariff regime was maintained for so many years in the late 19th and early 20th centuries.

It relies on vote trading among members of Congress to keep tariffs high. For example, Florida doesn’t produce steel, but Pennsylvania does. Pennsylvania’s senators and representatives want high tariffs on imported steel. Most states don’t produce steel, however, so how and why would Congress ever vote for high prices or high tariffs on steel?

Members of the Florida delegation care about sugar. They want to keep out sugar imports. Maybe in West Virginia, they want to keep out coal imports. Various state delegations will get together and say, “I’ll vote for your steel tariff if you vote for my sugar tariff.”

This vote trading, known as logrolling, gives rise to an interlocking coalition that makes it very difficult to push for lower tariffs. While most states might like lower tariffs on steel, they don’t want lower tariffs on their own goods, and they've collaborated to keep these tariffs high.

This is one reason why in the late 19th century even the Democrats, who came largely from the South, had almost zero success in reducing tariffs. They had to face these logrolling coalitions in Congress that are very difficult to defeat.

Jordan Schneider: My favorite part about trade history is we’re not just talking about the big steel, indigo, tobacco types of goods. We’re also going way down into the weeds. We’ve got bird cages, fur hats, crayons. It just keeps going and going because the interests are extraordinarily local and you can have individual line items for all these different types of goods.

Doug Irwin: Throughout the 19th century and well into the 20th century, whenever Congress considered a tariff bill, it consisted of up to 3000 individual tariff lines on individual products by the 1930s. Behind each of those products, there are some producers located in some states who are going to let their representatives and senators know that they really value those high tariffs.

You get members of Congress defending and debating the clothespin tariff, the tomato tariff, the felt skirt tariff. By the time of the Smoot-Hawley tariff in 1930, it was taking weeks, if not months, to get through the Senate. They would literally have roll call votes on each of these individual tariff lines, sometimes more than once. It does get into the nitty-gritty.

Jordan Schneider: I love how you have some quotes from senators and representatives talking about how basically this is the worst part of their job. They were happy to end up seeing it go just so they could spend time doing other things.

Doug Irwin: That’s one reason why Congress ultimately decided to delegate tariff authority to the president. They didn’t want to deal with the tomato tariff and the clothespin tariff. It was taking up an enormous amount of legislative time.

Of course, Washington would be besieged by special interest groups whenever Congress decided to take up the tariff. They said, “We have much more interesting and important issues to deal with than fighting these micro battles. We’re just going to give this to the president.”

How Bad Tariffs Can Wreck the Economy

Jordan Schneider: You mentioned the Smoot-Hawley tariffs, perhaps the most famous tariffs in world history. What were they and why did folks think they were a good idea at the time?

Doug Irwin: The Smoot-Hawley Tariff was passed by Congress in June of 1930 and then signed by President Hoover. It’s been infamous. First of all, the name is catchy: Smoot & Hawley. It’s not a regular term. It has a nice cadence to it.

It was a very ill-timed tariff in the sense that it occurred right as the world economy was slipping into the Great Depression. When it was originally proposed in the late 1920s, the US economy was doing very well. We were close to full employment. The stock market was booming. Industrial production was growing. The economy seemed to be doing well.

It was introduced in Congress in 1928-1929. The idea was that the farm sector was lagging and wasn’t doing so well. There were dislocations as a result of World War I and overproduction and overinvestment in land. We had to help farmers cope with imports. We had to raise their prices in some way.

In fact, Congress had twice passed price supports in the late 1920s but President Calvin Coolidge vetoed both of those pieces of legislation. Congress was groping for some way to help out farmers in rural America. Price supports seemed to be the obvious way of doing it, but the president didn’t want that.

They said maybe the tariff is the second-best way of dealing with the issue. Even though I explain in the book why it really wouldn’t have helped out farmers that much. Most big farmers were big exporters, not facing foreign competition or imports.

Here’s where logrolling comes in. The tariff was motivated originally to help out the farm sector and agriculture. However, once you start talking about the tariff in Congress, then the fishbowl industry and the tomato producers and steel producers start coming out of the woodwork. They start saying, “If you’re going to raise those tariffs, you might as well raise ours too. We could always benefit from capturing a little bit more market share by keeping out foreign competition.”

By the time it passed, it was an almost across-the-board tariff increase. The US had the stock market crash in late 1929. The world economy was becoming increasingly fragile. Congress didn’t really think about foreign ramifications in terms of retaliation.

What ultimately happened is that other countries did retaliate against the US. That hurt our exports. World trade began to contract. That exacerbated the Great Depression, which we had already begun to slide into. It became this trade policy disaster that the US spent about 20-30 years trying to pull our way out of.

It wasn’t just that other countries raised their tariffs against our products. They also formed these trade blocs that discriminated against the US. In particular, there was the British imperial preference scheme which included Britain, Canada, and others, big markets for US produce. Now the US faced not just higher tariffs, but discriminatory tariffs in those markets.

Smoot-Hawley was really this unnecessary, unwise piece of legislation and very ill-timed. It created big problems for a long time for US trade policy.

Jordan Schneider: Interestingly, you have a quote from Peter Navarro in your new Foreign Affairs piece talking about US-China trade relations. Navarro said, “I don’t believe there’s any country in the world that will retaliate for the simple reason that we are the biggest and most lucrative market in the world,”

Little did he know, had he read your book and understood the reactions to Smoot-Hawley, perhaps he would have thought a little harder about the international reaction.

Doug Irwin: There’s absolutely a parallel there. Some Democrats in Congress said we ought to really think about this carefully and not just consider domestic interests but also our export interests. They warned that other countries might retaliate. Basically, the reaction of most members of Congress, Republicans at the time, was, “Nah, we don’t have to worry about that.”

They thought, “This is a domestic piece of legislation. It doesn’t really concern other countries. They’re not going to retaliate.” Of course, they did. That seems to be one theme. Some policymakers or trade policy thinkers will understate or downplay the idea that other countries will retaliate against us.

Sometimes these groups are called economic nationalists. That's what Peter Navarro might call himself. What they don't realize is that there are economic nationalists in other countries too. If we aggrieve other countries and hurt their exports, they don't just say, “oh, I guess there's nothing we can do about that.” They get upset too when they're nationalists. They say, “we have to strike back against the Americans who cut out our markets.”

The reason why the Europeans were particularly concerned about this in the late 1920s and early 1930s is because they were trying to repay their war debts to the United States as a result of World War I. They needed to run trade surpluses with the US. Their ability to do so was hurt by the tariffs that we imposed. There was this financial dimension to what was going on at that time.

Tariffs as Reciprocity: Seeding Free Trade

Jordan Schneider: FDR comes in and the whole trajectory starts to shift. Why does this happen?

Doug Irwin: Here’s another one of these trade policy pivot points. The first one we talked about was the shift from revenue to restriction with the Civil War and the realignment in American politics, shifting political power to the North.

In the 1932 election, we had another political realignment towards the Democrats. It shifted a little bit more towards the South, although FDR’s coalition also included big parts of the urban areas in the North. The political party of the Democrats at the time was much more in favor of freer trade.

The Democrats in the South were still representing export-oriented industries. Urban areas wanted cheap imports for food and other things. The political system shifted back towards freer trade, but it wasn’t a shift towards just abolishing Smoot-Hawley.

When the Democrats had taken over in the past, a few times in the late 19th century, they tried to pass bills just reducing tariffs. The Republicans would try to push tariffs up. The Democrats would try to push them down.

What FDR realized at the time was, first of all, we were in a depression. A unilateral tariff reduction was not going to work well politically and it wouldn’t help us out that much economically. What happened as a result of Smoot-Hawley was that other countries raised their tariffs against us. They retaliated against us.

We have to get rid of that discrimination. We have to get rid of those foreign tariffs. The way we’re going to do so is by bargaining. We will reduce our tariffs on your goods if you reduce your tariffs on our goods, get rid of the discrimination. That’s why FDR didn’t say we’re going to just slash or unilaterally cut tariffs, we’re going to start bargaining.

That’s the introduction of this third phase of US trade policy, the third R: reciprocity. We’re going to negotiate with other countries. The Roosevelt administration pushed for this piece of legislation in 1934 called the Reciprocal Trade Agreements Act. It empowered the president to undertake those trade negotiating activities. That’s where we are, that's the regime we’re in today.

Congress no longer sets the tariff rates themselves. They don't pass these long, complicated tariff bills with individual tariff lines. They rely on the president to handle trade policy, within some circumscribed limits of course. The president is really the main trade policy actor today. That's where we've been since the 1930s.

Jordan Schneider: Did you dig at all into the intellectual history of the Reciprocal Trade Agreements, the whole idea of this? It does seem very much like a change in paradigm, as opposed to trying to plug away at these existing edifices.

Doug Irwin: Certainly, there was discussion in the progressive era about how Congress is slow and doesn’t really think about the national interest all the time but rather special interests. We need to empower a very strong, quick-acting executive. We need these expert administrative agencies to handle our policy, whether it’s the creation of the Food and Drug Administration, or other agencies like that.

There was this idea that we wanted to create this impartial, non-political tariff board that would adjudicate tariff issues. That didn’t work out so well. The idea had been floating around for some time that Congress is really not the best forum in which to debate these tariff rates. The phrase at the time was, “take the tariff out of politics.”

Really, there’s no way you can do that but the idea was to separate it from Congress. The idea of delegating authority to a branch of the president in particular had been floating around for some period of time.

Jordan Schneider: The RTAA definitely had its detractors at the time. Some said that it was a, “Fascist dictatorship with respect to tariffs.” Another said that, “There are no shackles upon this use of extraordinary tyrannical, dictatorial power over the life and death of the American economy.”

Also there was an American Tariff League that was pushing back against this. What does the pushback say about the politics at the time?

Doug Irwin: There’s another parallel to what we see going on today. We don't want to point fingers to anyone in particular. At the time the Republicans who were opposing this groped for any argument to object to something.

If a Republican president had said, “I need these tariff powers,” — in fact Hoover got some tariff powers in the Smoot-Hawley Act — Republicans would have been fine with that. The problem was not so much that they were delegating powers to the president. It’s that they knew that President Roosevelt would use that power to reduce tariffs. They cared about the outcome. They were objecting to the outcome that tariffs would go down as a result of this delegation.

This is something we see today when some of the political parties say, “We oppose this measure because it’s unconstitutional or it’s a bad policy framework.” They’re really objecting to the outcomes that they see in giving a certain administration certain powers.

The reason they lost is that they didn’t have the votes in Congress. They were a minority party. They could scream all they wanted to but they had no real impact on how events played out.

Jordan Schneider: When did the American Tariff League die out? Is Trump going to bring them back?

Doug Irwin: Could be. I still see references to them in the 1950s and maybe even the early 1960s.

By the 1950s and 60s, trade was not a big part of the US economy. Japan and Western Europe were devastated as a result of World War II. We really didn’t face a lot of import competition. The need for an American Tariff League was pretty small at that time. There were no major industries that were complaining about imports. We’ll see whether they come back or not.

A big difference between that period and the 1930s — when there was a politically robust American Tariff League — and today is that back then businesses were not really globalized. They were national businesses. There were foreign producers and domestic producers. The domestic producers wanted to keep the foreign producers out. Today, American business is really globalized. Most large companies are multinational. They're operating in many countries at the same time. They're sourcing components if they're in manufacturing from many different places.

We won’t see a revival of the American Tariff League because most businesses today are globally integrated. They rely on imported components. They don’t want the old-fashioned protectionism that we saw in the 1930s and before.

Jordan Schneider: You’re saying if I don’t do it, no one will?

Doug Irwin: There’s an opening for you. Give it a shot.

Jordan Schneider: It’s that or a dissertation on what they were up to in the 1950s. We’ll see which road I go down.

Who was Secretary Hull and how big a role did he play in changing trade policy at the time?

Doug Irwin: Cordell Hull was Secretary of State from 1933 to 1944 and he’s one of the people that most students of American history don’t know about or appreciate his role in terms of US trade policy.

Jordan Schneider: He’s a great character though. He comes from the middle of nowhere in Tennessee. He’s a hillbilly who ends up becoming secretary of state during World War II. It’s a rags-to-riches story if you ever wanted one.

Doug Irwin: He really played an outsized role in redirecting US trade policy in the 1930s. At the time he was made fun of. It was thought that trade policy was not very important during the Great Depression and that this goal of achieving all these reciprocal trade agreements wasn’t doable. The rest of the world doesn’t want them. He’s wasting his time.

He was very persistent. He was very focused on this one goal of using the State Department to try to get more economic peace and harmony through peaceful trade relations. He retired or resigned as Secretary of State in 1944. He was America’s longest-serving Secretary of State. He won the Nobel Peace Prize in 1945, both for his work on helping to create the United Nations as well as his work on trade policy.

He’s a really big figure that’s a little bit forgotten these days. He really did change the course of US trade policy. The impact of his ideas really came to the fore in the 1940s and 50s, after he left office, with the formation of the GATT (the General Agreement on Tariffs and Trade). We now know of that as the World Trade Organization. He really pushed for multilateral trade policies. That came about after he had stepped down from office, but he's a big figure in the history of US trade policy.

Jordan Schneider: There’s a tragic aspect of his story as well. His whole theory was that trade between countries helps bring about peace. One of the decisions that he partially signed off on was the blockade of Japan in 1940, restricting them from being able to import oil.

We don’t need to adjudicate why Pearl Harbor happened, but this was certainly one of the factors in Japan’s calculation that they needed to act. It was because of the power that US international trade had on their economy.

Doug Irwin: His deputy at the time, Dean Acheson, played a role in that as well.

Hull was generally considered to be a pretty weak secretary of state. That’s largely because FDR wanted to run foreign policy under the White House. I still think he’s a bit underrated, particularly in the trade field where he was of critical importance.

Jordan Schneider: Let’s get through World War II. Why does GATT matter then?

Doug Irwin: As we were just talking about, the Reciprocal Trade Agreements Act passed in 1934. It really didn’t achieve much in the 1930s. We already had fascist Germany and Japan that didn't want to sign trade agreements. Britain had gone in for imperial preferences with Canada and its former colonies. It was a tough row to hoe for Hull during the 1930s. He really didn't achieve much.

He had this idea that we should bring countries together and reduce trade barriers. It really didn’t come to fruition until after World War II. That’s when the US really was the dominant power in the world politically, militarily, and economically. Britain was much weaker. The US couldn’t really dictate terms but it could set the agenda for where the world economy could go, at least the arrangements relevant to the world economy after World War II.

The State Department said we want to have a big convention after the war to handle international monetary arrangements. That’s what the Bretton Woods Agreement was in 1944. It set up the IMF and the World Bank. We also wanted a big agreement on trade. That was first set up in 1947 at a conference in Geneva. It set out an agreement called the General Agreement on Tariffs and Trade, which set the framework for post-war trade relations.

The State Department actually wanted to do something even bigger and create something called the ITO, the International Trade Organization. There was a conference in Havana in 1948, but it was never approved by Congress and that effort failed. It would have been a great place, I’m sure. Great nightclubs and great music. Of course, the issues were political.

The fact that so many countries had so many divergent trade interests meant they really couldn’t come to an agreement around the table. Whereas the nice thing about the GATT was that it was basically the U.S. and Western Europe and a few other countries that really wanted to focus on trade liberalization.

They came up with this agreement for most favored nation status and other rules for trade as well as exchanging some tariff reductions in the bargaining process. The GATT process really moved forward, whereas the multilateral effort failed at that time.

Jordan Schneider: “Trade or fade” was one of the slogans of the time. I think someone should bring that back nowadays.

Doug Irwin: I think that was Eisenhower or someone in the Eisenhower administration.

What the GATT process did is not just reduce tariffs all at once. There was that first agreement in 1947, but there were a series of trade negotiating rounds about every decade. There was the Kennedy round of the 1960s, which put a further dent in the tariffs. There was the Tokyo round of the 1970s, which ratcheted tariffs down again. There was the Uruguay round of the 1990s, which did a lot of things but also ratcheted tariffs down. It was a slow process over time of reducing trade barriers and trying to constrain trade policy around the world.

Fast Track: Executive Trade Authority

Jordan Schneider: All right. Let’s now get to the next big piece of legislation: the Trade Act of 1974. What were the circumstances around its passing and what did it accomplish?

Doug Irwin: It did a couple of things. It wasn’t as big as the RTAA in 1934, but it renewed the president’s ability to negotiate trade agreements. That’s authority that has to be renewed every now and then. The president could always negotiate, but Congress may not like that. It was necessary for us to have that legislation to go into the Tokyo round.

It also introduced something called fast track, which is another thing we’re still living with today, not just trade negotiating authority. The first trade agreements in the 1930s, 40s, and 50s were just tariff-cutting exercises that didn’t need to go back to Congress to get approval. By the time we’re in the 1970s, tariff levels had come down quite a bit. Trade negotiations were dealing with increasingly non-tariff barriers.

Jordan Schneider: What are non-tariff barriers?

Doug Irwin: Non-tariff barriers would be government regulations or quotas, things that are obstructing trade that are not tariffs, strictly speaking. If you’re going to regulate those or restrict their use, that requires changes in domestic legislation.

The problem was that if the president negotiated some agreement and brought it to Congress, Congress would — just like they did with the Smoot-Hawley tariff — want to put its fingers in the pie and start changing this provision here and modifying that one there. Then you’d have to bring that back to the International Negotiating Forum. They might not like that.

If Congress got to second-guess or change whatever the president had negotiated, it would be a very laborious process that would never really get anywhere. What Congress did with fast track in the Trade Act of 1974, is that if you bring us an agreement that we’ve authorized you to reach, we will pledge to do two things.

One is to get an up or down vote in Congress, in the House and the Senate, and therefore change domestic legislation as a result. We won’t try to change the provisions per se. We’ll just give it an up or down vote, but we do have that right to veto it. 

Secondly, we’ll do it within a reasonable amount of time. We just won’t sit on it and let it die a long death. That’s what fast track is. That’s something that's still an issue whenever a president wants to reach a foreign trade agreement.

That’s what we’re dealing with now with the renegotiation of NAFTA, the USMCA. It’s been formally submitted to Congress. They have this timetable that they have to agree to approve it without changing any of the provisions.

Jordan Schneider: I’d like to dive deep on the Reagan era because he probably had the most intense, aggressive, protectionist measures we’ve seen in modern American history.

What drove an administration filled up with dyed-in-the-wool Republicans, whose staff were chock full of Milton Friedman devotees, to drop some really hard trade measures?

Doug Irwin: You have to remember that just about every administration is divided. It wasn’t just Milton Friedman devotees and free market advocates. There are also business people. Business people tend to like tariffs, at least on their business.

There’s a big analogy here with the Trump administration. There’s been this division in the Trump administration as well. At least early on there were Larry Kudlow and Gary Cohn, free market types that wanted free trade. Then you had the business people like Wilbur Ross who wanted to have tariffs to defend American industry against foreign competition. Every administration is divided between different factions. There's usually not just a monolith there.

Yes, Reagan spoke in favor of free trade and he seemed to support that. However, he also negotiated export restrictions on automobiles with Japan and imposed steel voluntary restraint agreements and things of that sort. The reason is that we had the worst recession since the Great Depression in the early 1980s as a result of the Federal Reserve under Paul Volcker tightening monetary policy to reduce the rate of inflation. That recession was accompanied by a very strong US dollar that led to a large trade deficit.

A lot of import-competing industries were being hurt. There was a lot of unemployment at the time, both because of monetary policy and to some extent because of structural changes in manufacturing. It was a way of using trade restraints to limit the damage during this tricky time in US economic history.

It’s a little bit different than the Trump administration today, which is giving the Reagan administration a run for its money in terms of the number of protectionist measures. Today, the US economy so far has been pretty strong. We have very low unemployment rates.

Reagan was dealing with unemployment rates over 10%. Intellectually he was in favor of free trade. Politically he was flexible enough to have temporary measures. Most of these things did go away with time. They were temporary measures to try to offset some of the hurt caused by the tight monetary policies at the time.

Jordan Schneider: I love how you tell this story about a series of events which lead to enormous sugar tariffs. The tariffs were so high that folks ended up bringing sugar in through Canada by means of cake mix, packets of iced tea, and then cooking them down and selling it as raw sugar. That whole rigmarole was actually profitable.

Doug Irwin: We had these import quotas on sugar that drove up the domestic price of sugar to two or three times the world price. You get these imports of fake cake mixes, which are 95 percent sugar. They just took the flour out once it came into the U.S. and sold the sugar at the higher price. There was even a case where we were banning imports of Israeli frozen pizzas because the sugar content was too high and some firms were extracting the sugar to sell it at the higher price.

What this shows you is that there's a limit to trade policy. You can raise tariffs on certain goods, but there's always an incentive to, if not smuggle, at least try to get around those tariffs. There's something called tariff engineering, where you change the type of product that you're sending into the US just to avoid the tariff.

Jordan Schneider: This sort of tariff engineering is going on nowadays as well, right?

Doug Irwin: Absolutely. There was just an article about Columbia, the clothing company. They would adjust the number of buttons or pockets on their shirts so that tariffs would apply to one category of goods and not another.

There was a big debate a couple of years ago about whether Snuggie blankets should be classified as a garment or a blanket. Garments and blankets get different tax treatment in the tariff code. I can’t remember which way it was supposed to go. The company definitely wanted them under the lower tariff category as they were manufactured in China.

One of my favorite examples of tariff engineering I have in the book is that we imposed tariffs on imported motorcycles during the Reagan administration in the mid-1980s. The tariffs applied to motorcycles with piston displacements of 700cc’s and above, heavyweight motorcycles.

What Honda started doing is producing a 699cc version. The difference between a 699cc engine and a 700cc engine is imperceptible. Just by changing that one cubic centimeter, it changed the whole tariff treatment. You avoided a 45 percent tariff and they were assessed at a much, much lower rate.

Reagan vs. Japan

Jordan Schneider: Let’s now turn to Reagan’s relationship with Japan. How did these tensions build up? The complaints you write about at the time seem pretty similar to the charges folks are levying today at China.

Doug Irwin: Yes, they’re very similar to what’s being talked about in terms of China. There are some big differences too.

The similarities are that we had a big and growing trade deficit with Japan, just as we’ve had with China. They were exporting a lot of goods that were hurting US industries, particularly automobiles, steel, and semiconductors.

In the case of Japan in the 1980s, it was a little less obvious what they were. With China, there’s been a lot of apparel and furniture and things of that sort. Certainly some manufacturing industries have been hurt as well as a result of imports from China.

There was also a view that they weren’t fair traders in the sense that their market wasn’t open for our products as our market was for theirs. We had a big trade deficit with Japan just as we have a big trade deficit with China.

The view with Japan was also that their formal trade barriers were low, but somehow we just couldn’t penetrate their market. It was very difficult. The same claim is made about China. Their formal tariffs are low, but the government is behind the scenes with administrative guidance or other mechanisms to avoid buying U.S. products, manufactured goods in particular.

There were also similar concerns about intellectual property. It was thought at the time that Japan didn’t respect US intellectual property and just stole some of our best technology. Obviously, the same is true with regard to China today.

The big difference is that the Reagan administration went after Japan and tried to threaten them and get them to change their policies with trade sanctions. Section 301 of the Trade Act of 1974 gives the president the authority to do this. That’s how President Trump has also gone after China.

The big difference is that the US and Japan had a very good relationship. We were both democracies, we were both largely market economies. We were both military allies. Japan relied on us quite a bit. Japan really wanted to accommodate US concerns as best they could under the circumstances. 

There was a much more amicable agreement to try to do something to reduce those trade tensions. With China, obviously they’re not a democracy. They’re not a military ally. We’re geopolitical and geostrategic rivals to some extent. China’s a rising power. Although we feared Japan as a rising power back in the 90s as well.

The relationship is much more fraught, much more difficult. Obviously China’s a much bigger economy too, the world’s second largest economy. Japan was back in the 1980s as well but it wasn’t considered quite the dominant threat as China is today.

The threats and the bullying don't work as well with China as they did with Japan, which wanted to reach an accommodation with the US. As I pointed out earlier, there are economic nationalists in the United States but there are also a lot of economic nationalists in China too. 

When the US hits them with tariffs, China does something Japan never did in the 1980s. They counter-retaliate, imposing tariffs on US goods. That’s exactly what China has done today in terms of buying soybeans and other products as well. They’ve increased their tariffs on US products and we’re in this trade war with China in a way that we never really were with Japan.

Japan never retaliated against the United States. The US-Japan trade conflict really faded by the early and mid-1990s when Japan’s economy began to flatline a bit. It remains to be seen what the end game is with China because they’re certainly experiencing slower growth. Whether they’re ever going to really back down or whether we’re going to diffuse this conflict remains to be seen.

Jordan Schneider: Could you talk in particular about the trade policy response to Japanese semiconductors?

Doug Irwin: That was a big issue at the time and the concern was twofold.

Japan had this tendency, in certain narrow product categories, to ramp up its exports very quickly, in a way that really surprised and hurt big US domestic industries. Autos is a classic case of that. Imports from Japan in the late 70s, early 80s, began to skyrocket.

The same was happening with semiconductors in the mid-1980s. They were investing a lot in the production of DRAMs (dynamic random access memory chips), basic memory chips in computers. They were just pushing the US industry out of that product category.

Many people thought that the whole US high-tech sector would suffer as a result and we wouldn’t produce any semiconductors. What happened as a result of that competition is Intel and others started producing more microprocessors and specialized chips rather than generic memory chips.

What the Semiconductor Industry Association wanted was two things. One, they wanted Japan to stop — what they were alleging was — the dumping of chips on the world market. Second, they wanted to open up Japan’s market to US-made semiconductors. The trade conflict in 1985-1987 centered around those two issues.

An agreement was signed in 1986 where Japan agreed to stop the dumping. Otherwise, the US could impose these special tariffs. There was also a secret side letter to that agreement where Japan agreed to ensure that 20% of semiconductor purchases in Japan would be to foreign vendors. That was a secret because it was a market share target. There was some trade barrier preventing Japan from importing semiconductors.

It was thought that, through various administrative practices, the Japanese semiconductor market really wasn’t open to foreign competition. They set up this market share target, which was eventually hit in the early 1990s. The failure to hit that target was one of the rationales for the US retaliation for noncompliance in terms of that agreement with Japan in 1987.

Semiconductors are not really remembered as much today in terms of the trade conflict. It was a big issue with Japan in the late 1980s.

This is something that comes up time and time again in US trade policy history. At any given point in time, there's going to be one issue which is considered overridingly important. All attention is focused on it. As little as five years later, it can be completely off the radar screen of trade policy officials.

Everything’s on Japan and semiconductors in 1987, but by 1992, it was basically almost a non-issue. That tends to happen a lot. There are a lot of these fads of trade policy concerns or trade policy issues that really just die away with time.

Jordan Schneider: They die away with time and die away with changing macroeconomic flows. As fear of Japan recedes and the real estate bubble bursts, and the stock market bursts with it, then it’s not so scary anymore that Japan’s semiconductors may be taking over the world.

Doug Irwin: Exactly.

NAFTA, the WTO, and competitive liberalization

Jordan Schneider: Now let’s come to NAFTA, which was almost not a priority for Clinton. Time and again, we see the influence of elite, worldly policymakers. Lloyd Bentsen of “You’re no Jack Kennedy” fame ends up talking Clinton into it.

Tell that story and talk a little bit about NAFTA’s path through Congress.

Doug Irwin: One of the themes of the book is that trade politics is always partisan. There are always parties that disagree over trade policy.

Generally, NAFTA, which had been proposed by Mexico, was supported by Republicans and George H. W. Bush. It was more opposed by Democrats who were concerned about the impact on US labor and things of that sort.

Ross Perot ran for president in 1992 against NAFTA. The Democrats were generally considered to be skeptical about NAFTA but it was also considered to be very important in terms of foreign policy. Here is a sort of economic rapprochement with Mexico, a country with whom we’ve had difficult relations for many decades. It would be hard to reject it out of hand.

Clinton was a New Democrat who supported globalization and wanted to support NAFTA. Many of his advisors, however, were skeptical about it. He was a bit cagey during the 1992 election campaign. He really wanted to support it. Many of his aides early on in his administration were saying, “This is going to divide Congress and divide your party. You really don’t want to push this and you want to delay it.” They did negotiate two side agreements on labor and the environment to make it more palatable to Democrats.

At some point, like with many trade policy decisions, you have to make a decision. There was this meeting of economic officials in the Clinton administration. The Secretary of the Treasury, Lloyd Bentsen — a Democrat who supported more trade with Mexico for economic reasons and foreign policy reasons — banged his hand on the table and said, “we got to stand up and support this. This is the right thing to do.”

That got everyone’s attention. He was held in high esteem by the president and other members in the White House. That pushed the decision. Yes, we’re going to back this thing and we’re trying to get it through Congress. It was quite difficult for a Democrat president to go against most Democrats and try to get it through Congress. President Clinton, to his credit, did it in 1993.

Jordan Schneider: It was a tough fight. Bentsen said, “it was pretty touch and go.” Lloyd Bentsen at one point said, “I courted some of these congressmen longer than I courted my wife.”

Doug Irwin: That gets back to that point.

When you're in the executive branch and you're negotiating these trade agreements, you're negotiating with two parties. One is the foreign countries that you're trying to deal with and the other is Congress and its domestic politics.

It was actually easy to come up with a trade agreement with Mexico. That wasn’t a hard negotiation. The hard part was convincing members of Congress to vote for it because trade politics is always difficult in Congress. There’s going to be a lot of opponents.

The people who are going to lose their jobs in the industries because they’re going to face foreign competition, they know that and they’re going to fight it tooth and nail. Whereas the beneficiaries, the consumers or other export industries, are much more diffuse and widely spread. They’re not as politically engaged or active as those who are going to oppose it. It took a lot of persuasion and arm-twisting to get Congress to approve NAFTA.

Jordan Schneider: At least in the age of globalists, which we can maybe start from the early nineties, it does seem like trade has a particular grasp on the imagination. People maybe give it more weight than the economics indicate.

You mentioned this idea of national security being a key factor in American trade policymaking. Can you expand on that a little more?

Doug Irwin: National security has always been in the backdrop of trade policy discussions, right back from the very beginning. One of the rationales for early tariffs in Alexander Hamilton’s famous “Report on Manufactures” in 1792, was that we need tariffs to be economically self-sufficient in certain areas. We can fight wars and not depend on imported war material. It’s always been there. It’s never been a primary concern because the US has never really faced big threats of invasion. 

During the Cold War when we were engaged in this conflict with the Soviet Union, there were provisions put in US trade law allowing the president to restrict imports on grounds of national security. This was if imports were impairing an industry that was deemed essential for national security.

That’s something that the Trump administration resurrected in the case of steel. They’ve also proposed it in the case of automobiles, although they haven’t proposed exactly what measures they’d like to take.

It’s always been a concern there. There’s always been an option to subordinate open trade policies to national security concerns. There’s also this question of abuse and whether it really was the case that we need to limit imports of steel on national security grounds or not.

Jordan Schneider: Speaking of other countries being upset at US trade policy, talk a little bit about the WTO. You argue that it was actually Reagan throwing around America’s might using the Section 301 clauses that got the world united into creating something like the WTO in the first place.

Doug Irwin: Yes, and certainly the dispute settlement system of the WTO, which is a key part of it.

In the late 1970s and 1980s there was a growth of voluntary export restraints. There’s something called the multi-fiber arrangement, which is limiting trade in textiles and apparel. There’s a sense that world trade rules were being circumvented and countries weren’t adhering to their obligations under the GATT.

The US wanted much greater enforcement of the rules. If we’re signing these agreements and coming up with these rules, they ought to be enforced in some way. In the administration’s opinion, there was no good mechanism in the GATT to do this. They started using Section 301 to address some of these things. Other countries hated this because the US would decide if some other country was a fair or an unfair trader. The US would grant itself the right to impose sanctions if a country was found in violation of some provision.

Basically, other countries said we need a neutral arbiter to determine if trade rules are being adhered to or not. We have to take this away from the US. It can't be just for the US to do this unilaterally. The US could get away with it because we're a large country. Other countries are really dependent on our market access.

How would Costa Rica or Jamaica ever enforce its trade rights? They don’t have the economic power to impose sanctions and discipline other countries and threaten them to get them to change their trade policies. We need some sort of independent arbitration mechanism to make this happen.

The Europeans and others basically said, “We’ll create this dispute settlement system in the WTO context as long as you stop using Section 301.” That was the implicit quid pro quo.

Jordan Schneider: Any reflections on the PNTR debate and Bush’s unilateralism?

Doug Irwin: The PNTR debate was about permanent normalized trade relationships with China. It was controversial at the time, in 1999-2000. It allowed China to enter into the WTO.  Recently the administration has raised questions about whether that was a good thing or not and whether we struck a hard enough deal.

There’s a re-litigating of that debate. Of course, the reason why is that China has become a huge economic power that’s created problems for American industries. There’s obviously a big rethinking of the US-China relationship.

The Trump administration thinks that we’ve been the loser there, so PNTR was a bad decision at the time. A lot of the former US trade reps have come out and defended that, saying that at the time we got the best deal we could and China actually gave a lot of concessions.

That’s just part of the trade policy debate, always second-guessing whether we got a good deal or not.

Jordan Schneider: On Bush’s unilateralism, this idea of competitive liberalization sounds good to me. It’s like a race to the top for trade deals.

What were the issues with it that actually turned trade into a much more partisan issue than it was pre-2000s?

Doug Irwin: Competitive liberalization was a phrase used by Robert Zoellick, who was the trade negotiator in the George W. Bush administration. It came about after the Uruguay Round that created the WTO. The idea was that the multilateral system wasn’t working so well in terms of generating new trade agreements.

To keep the system moving forward in terms of lowering trade barriers, we needed more bilateral agreements. That will put pressure on other countries to join up. You’ll get this wave of bilateral trade agreements that could help the multilateral system.

It ran out of steam by the late 2000s primarily because Congress became more reluctant to support them. It was forcing a lot of trade policy votes in Congress, which most members of Congress don’t like to do. The Democrats took over Congress, in 2006 or so. That put a damper on what the Bush administration could do.

The idea has something to it. You see what’s going on today in terms of the European Union and Canada and Mexico. The US has stepped back from trade agreements to some extent. These other countries and regions are really moving forward and pushing forward with more and more trade agreements.

There is this dynamic that’s going on. We see that with the Trans-Pacific Partnership. The US pulls out of it, but the other countries that are party to it continue moving forward with not just that trade agreement, but with other trade agreements with other countries as well.

There is something to the notion of competitive liberalization. Mexico, Canada, the EU, and others have really pushed forward on that.

Trump’s Trade Policy Pivot

Jordan Schneider: We've done plenty of shows about Obama’s trade policy so we’re going to skip forward to the present day. In your book, you end on an ambiguous note. You’re unwilling to commit to whether or not the Trump era is going to presage a real paradigm shift on the level of the Civil War or post-Great Depression trade policy.

However, you wrote a recent piece in Foreign Affairs with a former ChinaEconTalk guest, Chad Bown. You declare that even if Trump loses in 2020, “global trade will never be the same.”

What’s the right way to think about Trump’s current, as well as potential future, impact on American and global trade policy?

Doug Irwin: The reason why I was ambiguous in the book is because I actually finished the book in September of 2016. I sent it off to the publisher fully expecting that I’d be dealing with President Hillary Clinton and trade policy would be on no one’s radar screen. No one would care about the issue at all. The book would sink into oblivion and not receive any attention.

I did have a chance after the election to just write two or three pages speculating about what a Trump administration might bring. I tried to just speculate about where the Trump administration might take trade policy, whether we’d live up to the rhetoric or that was just overblown and not much would change.

The piece I wrote with Chad was a two-and-a-half-year assessment of the Trump administration in terms of trade policy. They have introduced some really big changes that cannot be easily undone. They fundamentally shifted where we are in terms of trade policy.

One thing he’s done is impose tariffs on imported steel. We’ve had steel tariffs in the past. There are some exemptions to them. We’ve also gotten rid of steel tariffs in the past. A new administration could get rid of these steel tariffs to repair relationships with other countries and to help out our downstream manufacturers who depend on cheap imported inputs. One can see that happening and that issue moves to the back burner.

Of course, the big one is China. That’s where what the Trump administration has done is much less reversible. By imposing these tariffs on imports from China, it becomes very difficult for any future president to get rid of them easily without reaching some sort of an agreement. It's going to be very difficult to reach an agreement with China. What the US wants them to do is essentially act like a market economy, open up and not have the government play such a large role.

Of course, that’s a fundamental contradiction in terms of where the Communist Party is coming from and the control it wants to have over the economy. I do see a more or less permanent medium-term breach between the US and China that won't go away even with a new administration. Some of the other things that we’ve had with trade friction with Mexico, and trade friction with the EU, a new administration can easily repair those relationships. 

Jordan Schneider: We’ve gone through a lot of American history here. Any suggestions for budding scholars who want to take a deeper look into trade history? Where should they dig? What sorts of questions do you think are particularly relevant today that are worth deeper exploration?

Doug Irwin: There’s so much out there that’s interesting. Obviously a lot of historical parallels can be explored. You can do some deep dives on some particular negotiations or some particular aspects of trade, going way back to the colonial period and through the years since.

One area that could be looked at more closely is the foreign archives in terms of how other countries have made adjustments to their trade policy, either in relation to the United States or just independently. We can shape more of a global history of trade policies.

Jordan Schneider: After I get my seven languages down, we’ll definitely take you up on that one. Any other final thoughts?

Doug Irwin: Just that it is amazing how there are these cycles in terms of how important trade policy is in American economic history.

It’s interesting to be present during this period now when trade is very much in the headlines and in the news, even if the news isn’t always good news.

Jordan Schneider: Oh to be alive in 2019. Doug Irwin, thanks so much for coming on ChinaEconTalk.

Doug Irwin: You’re most welcome. Thank you.

What makes DeepSeek great? How will DeepSeek’s moment impact the trajectory of AI in China and America?

To find out, ChinaTalk interviewed Kevin Xu, formerly of GitHub and Obama’s Press Office. Kevin is also the founder of Interconnected and Interconnected Capital.

In this knockout roundup, we explore…

• What DeepSeek does and doesn’t illustrate about Chinese innovation,

• Tensions between open-source cosmopolitanism and nationalism in DeepSeek and the broader Chinese tech community,

• DeepSeek’s organizational and talent management strategy, parallels to OpenAI, and what the fame will mean for the firm and Chinese AI policy,

• What DeepSeek should and may mean for the future of export controls and broader US innovation policy.

Have a listen on Spotify, iTunes, or your favorite podcast app.

Organizing Greatness

Jordan Schneider: Kevin, what makes DeepSeek great?

Kevin Xu: My one-liner description of DeepSeek for anyone catching up is that DeepSeek’s newest model is very capable, very affordable, and very open-sourced. DeepSeek as a company, startup, or what I would actually call more of a university-like AI research lab, is also very idiosyncratic.

They’re idiosyncratic because what they have accomplished and how they’re set up does not represent the norm of how most Chinese tech companies, US companies, or really any big tech companies around the world are set up. This comes out in three different dimensions that I wrote about in my newsletter right after New Year, after reading the V3 technical paper.

First, the lab has no pressure to commercialize whatsoever. While they have an API that charges some money, it’s probably at cost or slightly above cost just to recoup what they put into training the model. They are not OpenAI — they have no revenue expectations. They probably don’t have a sales team and likely never will. This has given the entire team significant freedom to quickly iterate and improve their model based on the latest open research.

Second, they run their own data center. They’ve been running their own data center since probably 2019, before ChatGPT and before expert control. This setup is quintessential to their roots as a quant fund, as all quant funds probably run their own on-prem data centers. This allows them to maximize the speed and efficiency of trading algorithms with their purchased hardware to get the most return for their LPs or fund. That level of hardware-to-software expertise from running their own data center has led to much of the mind-blowing innovation, discovery, or new paradigm that commentators have been discussing recently.

There is significant software-to-hardware optimization on the network bandwidth traffic balance loading. Recently, in the R1 paper, the engineers are literally going a layer below CUDA to work with the assembly language layer to control different parts of their GPUs, squeezing the most performance out of them to provide a model that is both very capable and affordable.

Because they have no commercial goals, strategy, or pressure, they’re willing to open source as much knowledge as they feel like. They’re probably the most open-source model we have in the market to date. All of this makes DeepSeek both very impressive and interesting, but also a unique example of what is happening in AI. Most people should pause before extrapolating a larger pattern from what DeepSeek has done.

Jordan Schneider: Let’s talk about hiring practices and labor design. This builds off a piece we just ran on the ChinaTalk newsletter from JS Tan, a PhD candidate at MIT who used to work in the cloud industry. There are at least three parallels between OpenAI and DeepSeek.

1. We have a very young CEO who is doing this out of the goodness of his heart, curiosity, and super long-run vision about how this is going to pay off — parallel number one to OpenAI. I remember in 2019, when a journalist asked Sam Altman how they were going to make money, he essentially said, “I have no idea. We’ll figure it out at some point.”

2. Parallel number two to OpenAI — DeepSeek is super small. OpenAI in the pre-ChatGPT era was in the hundreds. In 2020, it had literally 150 people, which is the same number of engineers that DeepSeek currently runs off.

3. They both have a bias towards young and truly exceptional talent that they’re willing to pay as much as needed to, outbidding the likes of ByteDance, Tencent, and Alibaba. There are all these lures of the OpenAI era where you had a 25-year-old intern or a 26-year-old hire basically coming up with key fundamental innovations that ended up unlocking GPT-3.

We’re at a point in AI engineering where having decades of experience is as much of a hindrance as it is an advantage in really trying to push the frontier of knowledge. To quote from Liang Wenfeng 梁文锋, the CEO of DeepSeek, “We need people who are extremely passionate about technology, not people who are used to using experience to find answers. Real innovation often comes from people who don’t have baggage.” A headhunter said they look for people with three to five years of work experience at most — any more than eight and you’re just a pass.

Generally, you hear stories about Chinese technology media and the Chinese tech ecosystem not hiring anyone under 30 or under 40, not because they’re inexperienced and can’t do the job, but because they can’t work a hundred hours a week, have kids, or might have a heart attack if pushed too hard. This is really different.

The other thing they’re doing — they have no KPIs, no hard organizational structure, no silos, no bake-offs between different teams competing against each other. That whole energy and ethos is, as you said Kevin, much more like a happy academic lab of young engineers super psyched to create the future. This contrasts with what’s presumably happening at Tencent, Alibaba, and even Google, where there is intense outside pressure to prove a return on investment in model making, which may not necessarily be the best way to deliver breakthrough innovations the way AI is set up today.

Share

Kevin Xu: There are many ways we can break down this unique cultural or organizational structure. Big shout out to JS Tan for a really great piece that I found myself nodding along to because it’s very spot on.

It’s important to understand how DeepSeek could pull this off. Coming back to what I said about no commercial strategy, they happen to be able to benefit from a quant fund that works right next door, which can basically fund whatever they need so far. Their approach to talent, if I can summarize what you said, Jordan, is that it minimizes politics and maximizes potential.

In a world like AI and AGI development, where none of us really know how to get there, we all have inklings. They reference OpenAI, look at what Llama is doing, probably look at what some other Chinese labs are doing, and have hunches of their own. It’s a killer to progress if you have a hunch in the middle of the night and then have to wait two weeks to get your GPUs before you can validate those hunches — based on my observation, that doesn’t exist in DeepSeek.

You have to minimize the organizational structure so you can get the resources you need almost immediately to validate these experimental needs. You need young talent that doesn’t have the baggage of experience, prestige, or previous accomplishments to be willing to try random ideas that may or may not succeed, with no ramification to failure.

Sam gave an interview to Bloomberg around New Year’s where he reminisced about that era of OpenAI. When they were trying to staff up OpenAI back in 2016 or so, they wanted to optimize for younger talent. They actually couldn’t get experienced talent to come work for them because building AGI at the time in Silicon Valley was a bit of a taboo. People didn’t want to be associated with it, preferring to work on something more utilitarian like self-driving, SaaS, or traditional machine learning applications.

DeepSeek has replicated that, maybe intentionally or not, in a way that is completely unencumbered from commercial need. Liang did try to raise venture capital, which goes into another nuance of the current state of Chinese VC — there’s zero appetite for the kind of idealistic vision or dream that DeepSeek is designed to pursue. He couldn’t raise any money even though he tried, but they had the people and some GPUs, so they gave it a shot.

You can contrast that with other well-known or well-funded AI labs in China, whether it’s Drupal, Moonshot, or obviously all the big models coming out of Alibaba, Baidu, and Tencent. They have either their internal ROI or commercialization goals to be accountable to, or they have state-backed or financial VCs from China to be accountable to. The way you build those companies is you have to hire well-known AI quantities to attract talent and convince investors to give you money. DeepSeek just didn’t need to do that. Maybe they wanted to at a point in time, but clearly, they’ve given up on that path, and it has served them very well.

Jordan Schneider: What’s interesting is the contrast between this playbook and other big Chinese industrial policy efforts. Semiconductors is probably the most obvious, where SMIC, famously founded by former TSMC engineers, and CXMT followed the same deal. Huawei poached enormous amounts of talent from Taiwan and South Korea to get its chip design and engineering up to speed.

There’s something really special and magical about AI today where decades of experience just happens to be more of a detriment than helpful. We are living in this incredible technological paradigm shift where people are discovering new laws, principles, and algorithms every day. Having the muscle memory of dozens of years of doing this back in a prior paradigm is actually more harmful than helpful, which is just the coolest thing.

Kevin Xu: Many folks are trying to fit what DeepSeek has done into some broader AI strategy out of China, another block into the industrial policy conversation you’re talking about. What SMIC has done makes total sense because qualitatively they’re trying to catch up to something that’s known — semiconductor manufacturing. TSMC has been way ahead of anything in China for a very long time, Samsung has been ahead.

Subscribe now

They poached Liang Mong-song 梁孟松, who was the former CTO-ish figure [excellent Asianometry video on him here]. While he didn’t get a CTO job at TSMC, he went to Samsung, helped Samsung catch up to TSMC, and then tried to play the same role with SMIC to help them catch up to Samsung and TSMC. All you need is one or two people who already know how this is done to get you to do something that already exists as quickly as possible. That means getting the experienced person, having their halo effect get you the talent and equipment, and then proceeding with state subsidies.

This is qualitatively different from AI research or AGI development, where we’re all just maybe six months, nine months ahead of each other into the unknown. None of the halo effect of experience — well, I shouldn’t say none of it matters, but it matters so much less than in other well-established areas of industry and technology, like chip manufacturing.

Jordan Schneider: Liang clearly, at least given his interviews, is the most AGI-pilled of all the Chinese AI lab leaders, as well as the most open-source pilled. Kevin, before launching your fund, you worked at GitHub for a long time and had significant interaction with the global open source community as well as the community in China. Let’s discuss how you see their energy baked into the team and motivation at DeepSeek.

Kevin Xu: I’ll explain this in two parts. Part one is open source itself, which at the basic level is a way to develop and build technology in the open, letting anyone find, download, share, and change the code built in the open. The assumption is that code itself has very little value in isolation — there’s no need to keep it behind a license or closed doors. Building in the open is the best way to build technology. This has been a 40-plus year movement.

The movement really took root in China in the mid-2000s. The Linux Foundation, the nonprofit steward of Linux, the largest open source operating system in the world, has been doing events and hosting conferences in China since the mid-2000s. Looking at Liang’s personal history, that’s when he started college. He was probably a star high school student, very technically gifted.

There’s a generation of Chinese entrepreneurs around that age group — late 30s, early 40s now — who are very much open source enthusiasts. They believe it’s one of the best things that happened to their own discovery and self-development. Now that they’ve come of age, they’re very pro-open source because of its benefits and positive-sum elements.

Coming back to DeepSeek’s idiosyncrasies, Liang could pull this off because he has a fund that could fund this research project with no need to directly commercialize the open source model, which is very different from other competitors.

From a cultural perspective, inside the Chinese open source community, there’s something I loosely describe as open source zeal or calling — “kai yuan qinghuai” (开源情怀). This manifests in a strong corps of engineers in China who love open source. They used to use mostly Western open source software to do their work, help their companies catch up, and learn more about technology.

About six or seven years ago, there was a turning point where many of these engineers, likely including Liang, wanted to produce open source themselves. They wanted to contribute, make their own open source projects, and have the world use them. They get incredibly excited when Western firms, especially Silicon Valley firms with brand recognition, use their database, microservices, sidecar, or whatever infrastructure package.

They provide free support, fix bugs in the middle of the night, and answer questions in forums. For them, it’s a source of validation and approval. They understand how little respect Chinese engineering and technology gets — whenever China develops something, it’s usually labeled as cheating or stealing. It’s somewhat nationalistic but also engineering pride — they want to prove their open source products can be used and loved by Western firms too. They care less about money than that validation.

This “开源情怀” (open source ethic) is evident in what DeepSeek is doing and hopefully will continue to do, though it’s an open question if they can continue to open source in perpetuity.

Jordan Schneider: Kevin, I’m going to hit you with another one — “崇洋媚外” (to revere foreign things and pander to foreigners). There’s this really interesting dichotomy. Looking closely at Liang’s two extensive interviews, the first one in 2023 was very much in that open source zeal and ethos — “I’m doing this thing, there’s no profit motive, it’s going to be good for humanity. It’s kind of crazy, but look how much we’re going to give to the world through our models and algorithms.”

Then in the July 2024 interview, the focus shifted more toward showing the world and Chinese engineers that they can achieve hardcore innovation (“硬核创新”). As we discussed at the beginning of this podcast, Chinese tech firms have this chip on their shoulder, poor organizational structure, poor incentives, and they believe all they can do is commercialize Western technology — just do consumer software stuff, not really push the frontier of innovation from a software perspective. Liang is saying, “Look, we’re doing this and it’s going to be awesome.”

A few weeks ago, Feng Ji 冯骥, the creator of Black Myth: Wukong, talked about DeepSeek as something that shapes China’s national destiny (国运). This tension between the open source, global surplus good that AI can bring and which country will do it first and make the most of it is really going to define the future of DeepSeek and the future of AI in general — where the world ends up falling on that divide.

Kevin Xu: The tension there is very palpable. They’re not irreconcilable, but there is a tension. Open source, as I described, is borderless by definition. Code speaks for itself. Anyone can leverage it, use it as this public good, which has its own drawbacks as well.

If you go to any open source conferences, which I’ve been to many up and down the stack, engineers all speak variants of English in different accents from around the world. Their community isn’t defined by nationality — it’s defined by whether you’re a C guy or a Rust person, a Kubernetes person or an OpenStax person. These are different camps of technology and open source that divide themselves into their own little tribes. This is completely counter to national identity as a whole.

When you come to the uniqueness of the Chinese engineering and Chinese open source community, there is this nationalistic pride or chip on their shoulder. They want to prove to the world that Chinese open source technology — which by definition becomes Chinese technology as a general matter — is good enough for the world to use. At least at this point, there’s no expectation for the world to pay them back for this contribution.

This goes into the idiosyncratic nature of DeepSeek’s team, which based on all estimates is entirely folks trained and studied in Chinese universities. Some might have done stints in foreign companies, but none for very long — none have spent 10 years at Google in Silicon Valley or Meta before getting poached back to work on stuff in China. This is a very unique identity.

It almost feels like a recruiting pitch. Every time Liang gives an interview, which is very rare, there’s always a recruiting element. He wants to attract the best and brightest because he has to fight against all the other big shops who probably offer more money than DeepSeek can match. What DeepSeek can provide is freedom to explore and that open source pride that’s very attractive to many engineers. Many engineers will work for less salary if the company is more open to sharing and more open source friendly.

Taking that into different directions regarding how he’s building the lab makes a lot of sense, but the tension is certainly there.

Jordan Schneider: Regarding the salary aspect, it’s easier to literally pay top dollar when you’re paying 25-year-olds, not 40-year-olds. The reporting actually works the other way where companies say they pay the most, but it’s much easier to pay the most for junior talent, which is less proven, versus the Stanford PhD who spent 10 years at Meta that everyone is bidding on.

Kevin Xu: They’re willing to pay high dollar to take a chance on younger, unproven talent, which many bigger companies either don’t do, or they push young talent into the system without much attention.

Jordan Schneider: Let’s talk about the future for DeepSeek. Continuing our OpenAI analogy — OpenAI didn’t have a hedge fund attached to them, but they had donations from Elon Musk, Sam Altman, and others.

Kevin Xu: They’re both nonprofit, in a sense.

Jordan Schneider: At a certain point, they realized that to scale both their product and research, they needed to partner with a hyperscaler, which turned out to be Microsoft — and now it’s Oracle and Dubai money. How do you think this plays out for DeepSeek from a corporate organization perspective after experiencing this moment?

Kevin Xu: My best guess is that DeepSeek has a good chance of maintaining its current structure for at least another year or two, which in AI is practically an eternity. They’re very well set up to do what they do, almost accidentally. I don’t know if Leung ever thought about doing DeepSeek when he started a hedge fund, but being a quant hedge fund allows for secrecy.

Recently, I learned that Liang worships Jim Simons, the founder of Renaissance Technologies, which is one of the most successful and secretive quant hedge funds ever. Liang wrote the Chinese foreword to the definitive biography on Jim Simons a few years ago — The Man Who Solved the Market by Gregory Zuckerman of Wall Street Journal, which is a fantastic read for anyone interested in financial history. Liang’s media shyness mirrors Simons’ approach. As long as his hedge fund continues performing well, they could maintain this structure for a long time. They don’t even have a PR department or typical big company infrastructure.

What might counterintuitively change this situation is if US export controls are loosened due to their development — which relates to the discussion about export controls being ineffective because of DeepSeek. If NVIDIA products become more accessible again, DeepSeek and other Chinese firms will immediately seek to acquire them. This scenario would require them to secure substantial funding quickly, either through partnering with a hyperscaler or taking outside investment. The dynamic could shift dramatically — even if you gave DeepSeek $100 billion tomorrow to build AI infrastructure, their deployment options would be limited, partially due to export controls.

Jordan Schneider: Even if export controls don’t change or get tighter, there’s still lots of compute available. We have a story coming out in a few days about China’s weird chip surplus, where apparently there are tons of chips floating around. Random local governments have them; ByteDance, Huawei, Alibaba, and Tencent — these guys are no slouches. The opportunity to scale inference and their experiments if DeepSeek gets a partnership is really remarkable.

I can only imagine the conversations going on in the hyperscalers who have their own AI labs. There’s this famous quote from Satya Nadella after ChatGPT dropped where he brings Peter Lee, who ran Microsoft Research, into the office and interrupts him mid-presentation to say, “OpenAI built this with 250 people. Why do we have Microsoft Research at all?”

Leaders at the hyperscalers are looking at their teams thinking, “We should just sign some partnership like Microsoft did with OpenAI because this is the golden goose. These guys and girls are just better than what we have.” If someone comes knocking on Liang’s door and says, “Look, export controls or not, we can give you 10x the compute budget you have now” — that’s a very interesting scenario.

Kevin Xu: I’ll put another company on people’s radar from an AI monitoring perspective on the Chinese side — Xiaomi. From what I understand, one of the key authors of the DeepSeek V3 development paper got poached by Xiaomi very recently to lead their AI division.

Xiaomi has been making a lot of noise about doing more with AI, going all-in on AI. They have all these EVs that are really good, they’re going to do self-driving, and they already have devices everywhere. They have a lot of channels to diffuse gen AI. That’s another company that doesn’t get talked about frequently when discussing Chinese AI.

To your point, Jordan, there will be many interesting forks in the road for DeepSeek that they didn’t anticipate. There’s a lot of speculation about whether DeepSeek’s release coinciding with Trump’s inauguration was intentional, or if it was to push back against this or destroy NVIDIA because their quant fund has an NVIDIA short position — all conspiracy theory mumbo jumbo that’s fun to read but shouldn’t be taken too seriously.

Looking at the timeline of the most important releases, DeepSeek V3, the base large language model — not the reasoning model — was released the day after Christmas. If you want to maximize your press impact in the Western world, you’ve chosen the worst time to release anything. Your PR person should be fired immediately.

Regarding R1’s release coinciding with Trump’s inauguration day — again, if you want to maximize impact, no one in the West is paying attention to anything that isn’t about President Trump. It took a week for us to really grasp what R1 meant after its release.

The Stargate announcement of $500 billion caught many people in the AI world by surprise, with consensus skepticism about whether they really need this or have the money. The R1 paper just sat there until people realized, “Wait, there’s this other thing we can use that costs so much less to train.” The numbers are obviously cherry-picked, but Stargate probably elevated DeepSeek in a way they never thought possible.

The only real forcing function for the DeepSeek release schedule is Chinese Lunar New Year. Just like American companies want to finish shipping before Christmas, Chinese companies want to complete their shipping before Chinese Lunar New Year so they can enjoy the holidays and return for another year of hustling.

Jordan Schneider: I was just thinking about other cultural changes about to hit them. The OpenAI attrition was interesting — there was all this Sam Altman skepticism, people didn’t believe him or thought he was a liar. They had that moment in December 2023 with this sort of false unity. Then all these people who used to be 25 years old, now in their early 30s, were like, “We’re done with this place and this guy.” OpenAI had a lot of senior research turnover over the past year.

DeepSeek is at the beginning of that cycle. Not to say Liang has any integrity issues, but just the sheer amount of money that other AI labs offer... things are going to change, perhaps not for the better, given all this attention.

US-China Relations and the Tragedy of Fame 偶像包袱

Jordan Schneider: Let’s talk about the political angle. Right before Chinese New Year, Liang gets a little glow-up with Li Qiang sitting down as the only AI representative in this work report preview. Now that the eye of Sauron is turning towards open source AI in China, what do you think that means for the firm and for open source AI in general?

Kevin Xu: I honestly have no idea. The Li Qiang meeting was very interesting to observe. For context, the setting was a work report where they regularly do these “top leader learn from industry leader” setups — what’s happening, what you’re doing, and in the most benign sense, how can the government help?

The most reliable leak from the meeting was that Liang predictably told the country’s leader they need more chips, they’re hardware constrained, export controls really hurt, and it would be great to have more chips. The other participants weren’t from tech — they were from different industries like sports, science, and robotics. Robotics people were actually overrepresented, and Liang was probably the only AGI model builder.

The fact that he’s now on the country’s top leaders’ radar is significant. To what extent? I honestly have no idea. Will this become part of some negotiation between President Trump and President Xi? If their last call mentioned TikTok, will their next call mention DeepSeek? That’s interesting to monitor.

There’s a chance the state could do a little too much to mess with DeepSeek’s little world that’s working really well, or they could let it be. Frankly, I have no idea how this will turn out.

Jordan Schneider: Before we get to the US angle, let’s look at the ledger of helpful versus harmful things. More chips? DeepSeek can get more chips by partnering with hyperscalers. That’s probably way less headache than trying to work with the Beijing municipal cloud, whose interconnects are probably already eaten away by rats.

Do they need tax incentives? No. R&D credits? Not really. Annoying things that could happen include golden shares, forcing the team to hang out in Yan’an 延安 for 10 days to see the caves and be more red, party cells, party meetings — a little annoying, but not the most annoying. It’s just something you have to deal with.

The bigger question to me is the long-term question of if and when open source models get too powerful for comfort, what does the Chinese government think about its firms pushing the frontier? Right now these models are fun, they’re fine — you can use them to write poetry, maybe book a restaurant for you, but you can’t just hack your city’s hospital system or read police reports by having a DeepSeek model on your laptop.

We’re on the trajectory where that sort of thing might happen. My base case is that once the models get more powerful, you’ll start to see more controls on what is and isn’t allowed to be released to the world. Now that folks on the Central Committee are going to be following this more closely than before, this open versus closed source dynamic with AI software — I see downside scenarios for a company that really prides itself on putting everything it does out into the public.

Kevin Xu: There are definitely many downside scenarios. Regarding how open source fits into the government perspective, at least from what I’ve seen in Chinese government releases, they’ve been interestingly very pro-open source so far in how they project their thinking about technology abroad, almost as a source of soft power, whether knowingly or not.

The Ministry of Industry Information Technology (MIIT) released part of their four-year report stating they want to see two to three very well-known, recognized Chinese open source projects by 2025. If they were to do the progress report today, DeepSeek would be on it — a great KPI moment for everybody in that ministry, whether they had a role in it or not.

More recently, Foreign Minister Wang Yi, speaking at the UN, talked about open source AI as something China wants to use to support the global south and help them develop their AI. At least from the outside perspective, they’re “very open source friendly.” From a projection perspective — how does that work when one of your labs is at the frontier or very competitive? These are just a few data points folks should keep in mind as we think about the distribution of outcomes for DeepSeek, from worst to best.

If any Chinese government leaders are listening and want helpful advice, a useful thing would be to unlock higher-quality data for training. There’s always going to be a data problem as far as I can see. Reinforcement learning has been put on the map again because of R1’s progress, so maybe synthetic data and models training each other can get us farther. But there’s always more quality data that could be unlocked, and the government has quite a lot of it.

Jordan Schneider: I want everyone to be fully into the AGI. We need that ancient Shang dynasty wisdom for our future AI overlords. To demonstrate part of the big freak out, I’m going to refer to this quote from Mark Zuckerberg on Joe Rogan:

Mark Zuckerberg: “If there’s going to be an open source model that everyone uses, like, we should want it to be an American model, right? There’s this great Chinese model that just came out — this company, DeepSeek, they’re doing really good work. It’s a very advanced model, and if you ask it for any negative opinions about Xi Jinping, it will not give you anything. If you ask it if Tiananmen Square happened, it will deny it.”

Jordan Schneider: My contention to you, Kevin, is this seems kind of overblown. The percentage of global queries that will be about Tiananmen or Xi Jinping thought or Xinjiang is basically zero. How would you conceptualize what companies, and countries more broadly, gain and lose from their model being the open source one that the world thinks is the coolest and wants to build on?

Kevin Xu: What Zuck said was very interesting. I wrote a post on my Interconnected newsletter about this too. Regarding whether Zuck was right or not, this has been part of his core strategic message for a few years — the “Xi or me” argument. Because Meta has been under US regulatory pressure for many years, one of his more effective arguments is that you can drive him down, but the world will just use TikTok, which they do, or use less Instagram, and that’s actually worse for US competitiveness.

When it comes to censorship in the model, I did some testing on this and need to update my post. With an open source model, it really depends on where you run it, not whether the model itself is censored. Every single model coming out of DeepSeek and the open source ones from Alibaba are probably trained from the world’s Internet knowledge as a starting point. Different things get fine-tuned out or fine-tuned in during post-training.

If you use these models on your own laptop by downloading them because they’re open source — I use Ollama — the behavior is very different than if you use it as a chatbot on their officially hosted website. The censoring or business logic that prevents your chatbot from saying potentially problematic things happens more in the cloud layer than the models themselves, though some occurs in the model as well.

There’s a third element — if you put DeepSeek on a third-party, non-official DeepSeek hosting environment, it also behaves differently. Perplexity, a leading AI search startup that’s really hot, is killing it with their DeepSeek deployment to get more users because they’re deploying it on their own US cloud server, promising no censorship, while still leveraging their best open source model, DeepSeek R1.

Regarding the broader global AI diffusion conversation — probably not many people would ask a chatbot about Tiananmen or other politically taboo topics in China. If they’re not getting the information they should, that’s problematic. Right now, these open source models really shine in business adoption settings. If you’re a large company wanting to use AI to summarize customer service scripts, help salespeople be more efficient, or summarize internal knowledge bases, OpenAI APIs are likely too expensive.

Having this super affordable option that you can deploy on a third-party cloud — it’s now on Azure, a very interesting late development. Azure has embraced DeepSeek as one of their first-party models you can just click and use. Microsoft’s pragmatic CEO is being shareholder-friendly — people want DeepSeek, so let’s put them on Azure.

Microsoft isn’t worried about the supposedly censorious characteristics of DeepSeek or any Chinese open source model because you don’t have that effect if you serve them in the right cloud setting for the right business use case. Most businesses just want to get AI going with a cost structure that makes economic sense. Nobody wants a $200 per month pro subscription account that doesn’t last long. People want cheaper AI, which is what open source brings to the industry.

This validates Meta’s open source strategy — pushing out Llama without concern for profit because they don’t need Llama to make money, really eroding the closed source model moat that OpenAI and Anthropic have been erecting to protect themselves. The big question going forward is how long that could last.

Jordan Schneider: Another argument beyond spreading Xi Jinping thought is the slightly more nuanced one that DeepSeek can now run on Ascend 910s, Huawei’s AI chip. The argument goes that if people standardize on DeepSeek, and DeepSeek partners with Huawei to make their chips, this could make Huawei chips — which are currently not all that competitive with NVIDIA’s — a much more exciting buying prospect. This applies both domestically in China and for global cloud providers who, thanks to the AI diffusion rule, will have a harder time getting Western chips. It’s analogous to how Huawei was able to outcompete Ericsson and Nokia.

Kevin Xu: Beyond the hardware, from a US competitiveness perspective, what should really concern people about the Huawei ecosystem benefiting from this is DeepSeek engineers’ ability to work below CUDA to really maximize their NVIDIA GPU. That level of low-level engineering could potentially help Huawei’s software ecosystem.

The bigger struggle for people to adopt anything that’s not NVIDIA is that CUDA is still dominant. Huawei has been trying to promote their own equivalent of CUDA, working closely with PyTorch, the open source training framework. To really mold that ecosystem and push developers away from CUDA — not necessarily NVIDIA GPUs — there needs to be more sharing of how DeepSeek has optimized this to make the software layer of the Ascend series more competitive.

This could erode the CUDA moat that, by extension, is sort of the American moat — something many people still don’t fully grasp. Even if Ascend hardware has better FLOPS and better performance at all scales, people still won’t use it extensively because of CUDA. This could really flip that situation. That’s another interesting unknown where DeepSeek comes in.

Jordan Schneider: Here’s our transition to export controls, because Huawei’s limitation is on the manufacturing side, which comes back to their inability to import semiconductor manufacturing equipment — that’s why we’re in this whole situation. Let me quote this clip from President Trump:

President Trump: “Today and over the last couple of days, I’ve been reading about China and some of the companies in China, one in particular coming up with a faster method of AI and much less expensive method. And that’s good because you don’t have to spend as much money. I view that as a positive, as an asset. So, I really think if it’s fact and if it’s true — and nobody really knows if it is — but I view that as a positive because you’ll be doing that too, so you won’t be spending as much and you’ll get the same result, hopefully.

The release of DeepSeek AI from a Chinese company should be a wake-up call for our industries that we need to be laser-focused on competing to win because we have the greatest scientists in the world. Even Chinese leadership told me that. They said you have the most brilliant scientists in the world in Seattle and various places, but Silicon Valley — they said, there’s nobody like those people. This is very unusual. When you hear about DeepSeek, when you hear somebody come up with something — we always have the ideas, we’re always first. So I would say, that’s a positive. That could be very much a positive development. Instead of spending billions and billions, you’ll spend less and you’ll come up with hopefully the same solution. Under the Trump administration, we’re going to unleash our tech companies and we’re going to dominate the future like never before.”

Jordan Schneider: Whoever wrote this I am sure is a ChinaTalk listener. DM me, we should chat.

Kevin Xu: They must be! I was very surprised and mildly impressed by how quickly DeepSeek made it into President Trump’s speech to a group of mostly Republican congressional leaders. This was a very US domestic event. He was talking to the Republican conference about his legislative agenda — tax cuts, deporting illegal immigrants, and so on. Then he just worked DeepSeek in there on the day of the big market crash that DeepSeek triggered.

What he said at the time was very complimentary, almost like a tough love message. “US Industry, you got to get your act together. Look at this DeepSeek thing — they’re kicking our butt.” But then it’s paired with, “The Chinese leaders tell us we have the best people, the best scientists, so we shouldn’t rest on our laurels.” That messaging has changed a little already over time, so we’ll see where this really ends up from a policy perspective.

I have to begrudgingly give kudos to the Trump White House speechwriting team for working that in so quickly. As a former communications person myself, they’re on top of it. Now regarding export controls, we’ve seen rumors of them tightening further under the Trump administration.

Jordan Schneider: Let’s roll another clip from Howard Lutnick’s confirmation hearing.

Howard Lutnick: “We’ve got to find a way to back our export controls with a tariff model, so that we tell China, ‘You think we are your most important trading partner — when we say no, the answer is no.’ It’s a respect thing. They’ve disrespected us, they’ve figured out ways around it. I do not believe that DeepSeek was done all above board — that’s nonsense. Okay? They stole things, they broke in, they’ve taken our IP. It’s gotta end. I’m going to be rigorous in our pursuit of restrictions and enforcing those restrictions to keep us in the lead, because we must stay in the lead.

…

I take a very jaundiced view of China. I think they only care about themselves and seek to harm us. And so we need to protect ourselves. We need to drive our innovation.

And we need to stop helping them. Open platforms, Meta’s open platform — let DeepSeek rely on it. Nvidia’s chips, which they bought tons of and found their ways around it, drive their DeepSeek model. It’s got to end. If they’re going to compete with us, let them compete, but stop using our tools to compete with us. I’m going to be very strong on that. I’m thrilled to oversee BIS, and I’m thrilled to coordinate and empower BIS with tariffs that will improve the strength. When we say no, the answer’s got to be no.”

Jordan Schneider: In the first news cycle of this, when I did my podcast last week with Miles Brundage, there was concern that this would be a turning point in the export control debate. People, myself included, were already worried about Trump turning export controls into a package deal he’ll use to get them to buy more soybeans. This is at least an initial interesting marker by the Trump administration that no, this stuff isn’t necessarily on the table when it comes to America striking a deal.

Let me take a step back, Kevin, and ask you more broadly. There’s something really sad about this. What DeepSeek has done and the team it has is something really special. People talk about what the US and China can collaborate on, and the first thing they point to is medical collaboration — wouldn’t it be great if we worked together to cure cancer? There’s going to be so much positive social good generated from artificial intelligence.

If the politics were different in China, if the government was different than it is today, that one piece of the Trump comment where he says “This is awesome, everyone’s gonna save a lot of money, this is gonna be cheaper and better for the world” — if it was almost any other country on the planet, that would be the reaction. But it’s not, because of all the reasons we’ve talked about in 500 other ChinaTalk episodes we don’t need to get into right now. I’m curious to what extent, Kevin, you think there is any sort of third path when it comes to US-China relations, strategic competition and artificial intelligence.

Kevin Xu: President Trump’s comment sounds more like a Silicon Valley investor, while Lutnick, who was supposed to be the Wall Street free market guy, sounds more like a hardcore DC China export control supporter. From a tactical perspective, it’s important for people to understand that export controls have worked reasonably well to achieve their goal of keeping Chinese AI progress from advancing further. Without export controls, Chinese AI labs would have gone much further than where DeepSeek or Alibaba has gone right now.

That’s a tactical question. The tactics arise from the strategy of “small yard, high fence.” People conflate strategy with tactics. “Small yard, high fence” is the why question. Export control is the how question. The how flows from the why. The big question mark now is whether the Trump administration challenges or changes the why of our relationship with China when it comes to AI competition. That will change export control one way or another in its entirety and could change the whole game. Many folks who have been hyperventilating about export control failing miss the difference between what is tactical and what is strategic.

Jordan Schneider: Let’s tease out some of those futures. On one hand, we have Dario Amodei saying we need to cut them off from the H100s, squeeze them as much as we can because we’re about to run into the singularity. We want liberal democracies to be able to decide how that works, not the CCP.

Then we have VC investor Trump saying this is great for the world — American companies need to put up or shut up, step up their game. It’s not my job to bail you out, and if you lose, whatever — it’s a global good, this is open source, things will work out, you all get great AI tutors. What other visions can you conjure, Kevin, of how this could potentially play out?

Kevin Xu: Dario’s essay is very much worth a read. There’s the benign humanity benefit path, and then there’s the strictly military competition, head-to-head path. One of the justifications in Dario’s essay for his argument is the military perspective — if China gets to AGI before the democratic or Western world does, they will apply it to military settings.

Coming back to the cancer-curing setting, nobody would argue or care whether the cancer solution from AGI is a democratic or non-democratic solution. If there’s a third way to parse this difficult path forward, can we disaggregate the use cases? There are obvious use cases where the nature is conflict — that’s military, that’s war, there’s no question about that. Then there’s business productivity, efficiency, all the enterprise AI stuff I mentioned. Finally, there’s the humanity benefit aspect — health.

Watching the Stargate press conference at the White House, they justified this not through military terms, but healthcare. That’s what Larry Ellison said. It’s potentially finding a cure for cancer, which Sam Altman mentioned. That’s the benign message justifying the billions of dollars of AI investment they hope to have in the United States.

There’s this interesting dichotomy that only the West can find the cure to cancer first, then license it to the rest of the world. Hopefully, this gets diluted over time when open source is more widely accepted from a positive-sum perspective. But we must be mindful that AI could be used for conflict as well, or for nuclear development.

Anthropic deserves kudos — based on what I know, they work with the US Government’s Nuclear Review Commission to red-team their new models, ensuring they don’t accidentally leak important and harmful information related to nuclear power or nuclear weapons, not just to adversaries but more importantly to non-state actors and terrorist organizations.

This represents the middle-of-the-road AI safety lens that deserves more discussion without preventing the world from finding a cure for cancer earlier, just because of this paradigm of conflict we can’t escape.

Jordan Schneider: Reading Dario’s export control essay in parallel with his October 2024 piece “Machines of Loving Grace” is instructive. With his biomedical background, he painted a clear picture of how medical research will be accelerated. But on the other hand, you have missile systems, drone swarms, and all the nastiness that can come from it.

The base case might be that it’s probably impossible to stop the PLA from benefiting from AI. They’re going to get the chips first, they’ll be able to squeeze all the secrets out of the best engineers and impress talent. They might not do it efficiently, but what the US, China, and the rest of the world can try to minimize is the crazy leakage — like instead of one kid going crazy and shooting up a school, they go crazy and release a bioweapon that kills half of humanity.

Kevin Xu: The latter case concerns me much more — either rogue actors or non-state actors, as opposed to reasonably irrational state actors literally blowing things up for unknown reasons. Not being deeply versed in military matters, I’m skeptical of what the latest AI models can really do for making weapons more harmful. The cyber attack angle sounds more plausible to me. From a physical weapons perspective, it’s hard for me to assess.

Even from the United States perspective, the Department of Defense has been deploying AI using Palantir software and similar tools. The biggest benefit people discuss is better, more efficient supply chain management — that’s an enterprise AI application, not an offensive one. It makes your back office and supply chain, which are very complicated in the DoD military setting, much more efficient. That’s no different from IBM wanting to use AI to make their organization leaner.

Jordan Schneider: The bull case is AI-driven scientific advancements which can then be weaponized. There are many naughty technical problems where solutions could be revolutionary — underwater radar, laser weapons — lots of sci-fi possibilities. With the equivalent of a thousand von Neumanns running at a thousand hertz per second, you can imagine developing some pretty crazy capabilities that could give you a dramatic military advantage.

AI isn’t there yet, but Dario believes it might be soon. Given the pace of advancement, it’s not crazy to consider that a double-digit possibility. With that premise, you have to take all this export control stuff really seriously. That’s where I come out on the issue. It’s unfortunate because there will be global social costs to restricting who can work on this technology and pursue its upside potential. But hopefully we’ll get there too.

Kevin Xu: Open source makes that diffusion answer very straightforward — it will just be out there. With open source, if you want to make money from it, most of the time you don’t even know who’s using it if they don’t tell you. That’s both the beauty and curse of building open source software. It’s always been the case and always will be. Hopefully the benign version wins, but we shouldn’t be complacent about the possibility of the non-benign version becoming reality.

Outtro Music

We released a new podcast yesterday interviewing Kevin Xu of on DeepSeek. Check it out on Apple Podcasts, Spotify, or your favorite podcast app.

By open sourcing its models, DeepSeek opened the door for any curious developer to learn from and build upon its cost-saving innovations.1

This is the open-source ideal: free exchange of ideas in the global researcher’s sandbox that allows clever and creative ideas to compound. Proponents of OS models argue that it can accelerate science and innovation, improve transparency, distribute governance, and enhance market competition.

Still, the debate on open versus closed source rages in the AI community. It’s been hard for anyone to provide a definitive near-term commercial reason for why a firm would open source its AI models.2

In the Chinese tech space, this pragmatic sentiment is common. Baidu founder Robin Li 李彦宏 put it bluntly:

Open source is a kind of IQ tax 智商税...

When you rationally consider what value a large model can bring to you and at what cost, you should always choose a closed-source model…Many times, a model may seem useful, but when you calculate the costs, it’s not cost-effective so customers abandon it.

This is why I said that open-source models cannot beat closed-source models.

Alibaba maintains its open-source Qwen, but makes money by upselling APIs, cloud services, and computing infrastructure to customers. Kai-fu Lee’s 李開復 start-up, 01.AI, released Yi-34B as a way to “give back” to the Chinese developer community, but the company ultimately targets state-of-the-art proprietary models as the basis for its commercial offerings.

DeepSeek, on the other hand, has no near-term profit strategy.

In a 2023 interview, CEO Liang Wenfeng 梁文锋 very clearly stated that there is no commercial rationale for DeepSeek’s dream of “limitless research.” This sentiment brings to mind a quote from OpenAI’s Sam Altman in 2019, when he said, “I have no idea how we may one day generate revenue.”

High-Flyer Quant says it isn’t in it for the returns, either. The quantitative hedge fund financing DeepSeek recently emphasized that High-Flyer’s AI model research will not be used for stock trading: “It has nothing to do with finance. … What we care about is long-term social value.”

We should take these statements of principle at face value — this isn’t a government front, since the way DeepSeek has moved is so antithetical to traditional Chinese government-backed industry. It’s more accurate to say that DeepSeek’s employees, largely composed of young homegrown talent, are driven by something other than money-making. Is that madness, one interviewer asked? Liang’s response was telling:

Liang: I’m unsure if it’s madness, but many inexplicable phenomena exist in this world. Take many programmers, for example — they’re passionate contributors to open-source communities. Even after an exhausting day, they still dedicate time to contributing code.

Waves: There is a sense of spiritual reward in it.

Liang: It’s like walking 50 kilometers — your body is completely exhausted, but your spirit feels deeply fulfilled.

Waves: Do you think curiosity-driven madness lasts long-term?

Liang: Not everyone can stay passionate their entire life. But most people, in their younger years, can wholeheartedly dedicate themselves to something without any materialistic aims.

These young Chinese developers’ intense passion to work on open-source projects is sometimes called “the open source ethos” 开源情怀, which Kevin Xu from Interconnected explains well:

Most engineers are thrilled if their open-source projects — a database, a container registry, etc. — are used by a foreign company, especially a Silicon Valley one. They’d tack on free labor on top of already free software, to fix bugs, resolve issues, all day all night. It’s all for the validation and approval.

Implicit in this “zeal” or “calling” is an acute awareness that no one in the West respects what they do because everything in China is stolen or created by cheating. They are also aware that Chinese firms have been taking for free lots of open source tech to advance, but they want to create their own, contribute, and prove that their tech is good enough to be taken for free by foreign firms — some nationalism, some engineering pride.

In his more recent interview, Liang shared a similar insight. It’s easier to recruit engineers to solve tough problems if they have a thirst to prove themselves, he explained: “Top talents in China are underestimated because there’s so little hardcore innovation happening at the societal level, leaving them unrecognized.”

That is now changing. Thanks to recent open-source models, DeepSeek has earned global recognition and respect from engineers around the world. But will China’s government see it the same way?

Beijing’s approach to open source

The Chinese government has already expressed some support for open source 开源 development.

In 2018, a (since-deleted) white paper and the formation of the China AIOSS Development Alliance 中国人工智能开源软件发展联盟 brought open-source AI into the spotlight. An open ecology would be achieved, the white paper asserts, by cultivating OS communities and talent, promoting standards, establishing funding mechanisms, improving the intellectual property rights regime, and strengthening security reviews.

Top-level policy plans3 reference open source when discussing software and technology ecosystems, but not AI specifically. Two separate five-year plans also advocate for the benefits of open source.4

This positive political stance toward its domestic open-source community stems from a government and industry desire to reduce Chinese dependence on foreign software.

The government’s push for open source in the early 2000s — including the creation of several OS software alliances and a locally developed “Red Flag Linux” 中科红旗 — was a way to limit the impact of Microsoft Windows operating systems. Later, the Ministry of Industry and Information Technology designated Gitee as China’s national “independent, open-source code hosting platform” to replace GitHub, which it has struggled to censor. In the Chinese chip industry, open-source projects are being explored as a way to reduce dependence on closed-source Western ecosystems.5

Huawei exemplifies how geopolitics can color support for Chinese open source. In 2019, the president of Huawei’s consumer software group warned that without its own OS community, “All of China’s software communities will be in great danger the moment something happens beyond our control.” Huawei is obviously referring here to the struggle with US sanctions.

How it could all go wrong

So, will the Chinese government allow DeepSeek’s team to continue with the excitement of their nerd show-and-tell?

Unfortunately, potential liabilities from AI technology could push the government away from open source despite all the positive rhetoric.

Operating systems can’t disseminate information and power to the public in the way that AI can.

An open-source AI model grants the public broad access, usage, and customizability — which can’t easily be moderated or rescinded. There are general AI safety risks.

And beyond that, with the prospect of future advancements of AI, an outspoken chatbot might not be the only threat on the government’s radar. In our recent show with Miles Brundage, Jordan puts it this way:

We’re not far from a world where, until systems are hardened, someone could download something or spin up a cloud server somewhere and do real damage to someone’s life or critical infrastructure. It’s not just about reading outputs about Tiananmen — it’s about democratizing power by giving individuals access to an incredibly powerful technology that has the potential to cause real social damage. China does not let civilians buy guns — once open-source AI really gets weapons-grade, and one person can shut the lights off in a city, is that really something the CCP will allow to proliferate without any control? I don’t see that as a world state that government officials in Beijing, or the West for that matter, will accept.

Today, China’s generative AI regulations lack specific guidance for open-source providers. As regulators attempt to balance the country’s need for control with its ambition for innovation, DeepSeek’s team — driven by curiosity and passion rather than near-term profit — might be in a vulnerable spot.

The uncertain future of DeepSeek and open source

At least for now, though, the Chinese government sees the benefits of open-source AI.

To be clear, the only government statement6 we have on DeepSeek (for now) endorses open source:

DeepSeek’s greatest strength lies in its open-source approach, which empowers researchers worldwide…This highlights the importance of transcending a narrow, competitive mindset. China’s technological progress can and should contribute to humanity on a broader scale. Ultimately, what is good for China can be good for the world, and vice versa. As long as Chinese technology aligns with global progress, no barriers—no matter how high or deep—can halt its development.

And DeepSeek’s success has inspired more dialogue in China about the advantages of open source. Chinese AI startup MiniMax released several open-source models with the hope that “there will be encouragement for good work and criticism for bad work, and people outside will be able to contribute.” Chinese analysts pointed out that cost-effective open-source models support widespread access and adoption, including to countries in the Global South. See a viral post by an engineer refuting Dario’s export control essay:

Dario raises a critical question: What would happen if China gains access to millions of high-end GPUs by 2026–2027? His answer is this—if China cannot obtain this computing power, the U.S. will enter a phase of "unipolar AI dominance" and could solidify its advantage long-term through AI’s self-reinforcing mechanisms. However, if China does acquire these resources, the U.S. may face a prolonged “AI arms race.”

But my perspective is this: Whether unipolar or bipolar, AI development has already irreversibly entered a phase of global diffusion. The U.S. will not monopolize AI, China will not be contained, and nations like Europe, Japan, India, and others will not remain absent. Variables such as export controls, model competitions, and capital flows may influence the pace of the race, but they cannot halt the world’s march toward more advanced forms of AI.

DeepSeek is not an endpoint, but a signal—its significance lies not in “defeating” anyone, but in proving that the world has entered an irreversible era of large-scale AI competition.

When thinking about DeepSeek, two things can be true at once:

• DeepSeek likely chose to open source its models for the same reason developers from around the world choose to open source: out of genuine faith in the value of an open, global research community — to show off their accomplishments and inspire others to build upon their work.

• At the same time, as AI models become more powerful, governments might have an incentive to step in and take command.

[Jordan: Beyond direct government intervention, DeepSeek’s accomplishments will open doors for corporate partnerships bearing orders of magnitude more compute than DeepSeek runs today. DeepSeek has been slept on domestically in China — it has no outside funders, no fancy returnees with foreign degrees and experience at top Western labs, no government contracts, and no popular consumer apps. This will change. Just like what happened to OpenAI with Microsoft post-ChatGPT, China’s top hyperscalers — ByteDance, Alibaba, Tencent, and Huawei — will come knocking. Given Liang’s commitment to hardcore innovation and admission that compute access is holding him back, he might not pass up the opportunity of extra compute for R&D. This could scale up DeepSeek to really push the frontier of innovation with the major AI players in the West — but doing so could also unleash the same sort of forces that led OpenAI to splinter and caused its top researchers to scatter to the winds.]

Does open-source AI have a future in China? Will the lure of hyperscaler compute pull DeepSeek away from its open-source ideals? Will the government intervene for the sake of security and control?

Subscribe to ChinaTalk as we continue to follow this story.

We’re looking for advertisers. Respond to this email to connect.

Jordan: So, R1 is really good at poetry. Let’s close with its meditation on DeepSeek and its future with open-source models. I fed it this article (initially it refused, telling me in Chinese “Sorry, I haven’t learned how to think about these types of questions, I’m good at math, coding, logical topics, so please let’s chat about those things.” “对不起，我还没有学会如何思考这类问题，我擅长数学、代码、逻辑类的题目，欢迎与我交流.” Then I got ChatGPT to summarize the piece above, fed it back in, told it to write an award-winning contemporary poem, and after a few rounds it came out with this.

The Unlicensed Sky

Guest piece by , a PhD Candidate at MIT’s international development program, researching the political economy of innovation in the US and China with a focus on cloud computing. He was previously a software engineer and writes on substack here.

As of December 2024, DeepSeek was relatively unknown.

Then its base model, DeepSeek V3, outperformed leading open-source models, and R1 broke the internet. No easy feat when operating with less compute than western labs.1

Their breakthroughs raise two key questions:

1. How did DeepSeek outcompete Chinese AI incumbents, who have thrown far more money and people at building frontier models?

2. What does DeepSeek’s success tell us about China’s broader tech innovation model?

DeepSeek’s success is not just a product of technical ingenuity, but also deeply rooted in its unique approach to labor relations. Chinese tech firms are known for their grueling work schedules, rigid hierarchies, and relentless internal competition. DeepSeek’s flat management structure, in contrast, focuses on empowering its workers with autonomy and creating a collaborative environment.

DeepSeek is hardly a product of China’s innovation system. The company is neither a state-led project nor a direct beneficiary of China’s AI-focused industrial policies. Rather, it was self-funded by a former hedge-fund manager and emerged from the periphery of China’s tech landscape. While DeepSeek makes it look as though China has secured a solid foothold in the future of AI, it is premature to claim that DeepSeek’s success validates China’s innovation system as a whole.

How DeepSeek Broke the Tired 996 Playbook

To appreciate why DeepSeek’s approach to labor relations is unique, we must first understand the Chinese tech-industry norm.

Perhaps the most notable aspect of China’s tech sector is its long-practiced “996 work regime” — 9 a.m. to 9 p.m., six days a week. This workplace culture emerged during the rise of China’s digital economy in the mid-2000s and solidified during the hyper-competitive years that followed. Employees are kept on a tight leash, subject to stringent reporting requirements (often submitting weekly or even daily reports), and expected to clock in and out of the office to prevent them from “stealing time” from their employers. Employers set demanding key performance indicators (KPIs) and practice “stack ranking,” a performance management system where employees are ranked against each other.2

Since the mid-2010s, these grueling hours and draconian management practices were a staple of China’s tech industry. The long hours were considered a basic requirement to catch up to the United States, while the industry’s punitive management practices were seen as a necessity to squeeze maximum value out of workers.

Indeed, speed and the ability to rapidly iterate were paramount during China’s digital growth years, when companies were focused on aggressive user growth and market expansion. The primary goal was to quickly and continuously roll out new features and products to outpace competitors and capture market share. This relentless pursuit of expansion demanded a workforce that functioned like a well-oiled machine. As a result, employees were treated less as innovators and more as cogs in a machine, each performing a narrowly defined role to contribute to the company’s overarching growth objectives.

Such labor relations can be seen at Pinduoduo, a rising challenger to Alibaba’s dominance in e-commerce. The company is infamous for requiring an extreme version of the 996 work culture, with reports suggesting that employees work even longer hours, sometimes up to 380 hours per month. Management uses digital-surveillance tools — including location-tracking systems — to measure employee productivity. Even bathroom breaks are scrutinized, with employees reporting that prolonged absences can trigger disciplinary action. Those who fail to meet performance benchmarks risk demotion, loss of bonuses, or even termination, leading to a culture of fear and relentless pressure to outperform each other.

Since the late 2010s, however, China’s internet-user growth has plateaued, and key digital services — such as food delivery, e-commerce, social media, and gaming — have reached saturation. Those developments have put the efficacy of this model under strain.

But instead of focusing on developing new value-added digital innovations, most firms in the tech sector, even after public backlash about the 996 working schedule, have doubled down on squeezing their workforce, cutting costs, and relying on business models driven by price competition. This approach comes at a cost: stifling creativity, discouraging independent problem-solving, and ultimately hindering China’s ability to engage in long-term innovation-based competition.

New Approach to Talent

DeepSeek’s approach to labor relations represents a radical departure from China’s tech-industry norms. Since its founding in 2023, the company has eschewed the hierarchical and control-heavy management practices standard across China’s tech sector. Instead, it has built a workplace culture centered on flat management, academic-style collaboration, and autonomy for young talent.

The team size is deliberately kept small, at about 150 employees, and management roles are de-emphasized. Research groups are formed based on specific goals, with no fixed hierarchies or rigid roles. Team members focus on tasks they excel at, collaborating freely and consulting experts across groups when challenges arise. This approach ensures that every idea with potential receives the resources it needs to flourish. Liang Wenfeng 梁文峰, the company’s founder, noted that “everyone has unique experiences and comes with their own ideas. They don’t need pushing. … When an idea shows potential, we allocate resources from the top down.” To that end, DeepSeek actively avoids the performative aspects of traditional tech workplaces. There are no weekly reports, no internal competitions that pit employees against each other, and famously, no KPIs.

On the human capital front: DeepSeek has focused its recruitment efforts on young but high-potential individuals over seasoned AI researchers or executives. Many of DeepSeek’s researchers, including those who contributed to the groundbreaking V3 model, joined the company fresh out of top universities, often with little to no prior work experience. Said one headhunter to a Chinese media outlet who worked with DeepSeek, “they look for 3-5 years of work experience at the most. Any more than 8 and you’re just a ‘pass’ for them.” Liang explains the bias towards youth: “We need people who are extremely passionate about technology, not people who are used to using experience to find answers. Real innovation often comes from people who don't have baggage.” While other Chinese tech firms also prefer younger candidates, that’s more because they don’t have families and can work longer hours than for their lateral thinking.

Heavy emphasis is placed on educational background and competition achievements. The company is known to reject candidates who’ve achieved anything but gold in programming or math competitions. And beyond a cultural commitment to open source, DeepSeek attracts talent with money and compute, beating salaries offered by Bytedance and promising to allocate compute for the best ideas rather than to the most experienced researchers.

This hiring practice contrasts with state-backed firms like Zhipu, whose recruiting strategy has been to poach high-profile seasoned industry recruits — such as former Microsoft and Alibaba veteran Hu Yunhua 胡云华 — to bolster its credibility and drive tech transfer from incumbents.

[Jordan: this strategy has worked wonders for Chinese industrial policy in the semiconductor industry. Poaching experienced talent from TSMC and Samsung has been integral to SMIC, Huawei and CXMT’s success. But AI engineering is in a unique moment where young lateral thinking often trumps talent trained in the pre-transformer era.]

DeepSeek’s success highlights that the labor relations underpinning technological development are critical for innovation. While many of China’s tech giants have focused on squeezing maximum output from overworked employees, DeepSeek has demonstrated the transformative potential of a supportive and empowering workplace culture. By breaking away from the hierarchical, control-driven norms of the past, the company has unlocked the creative potential of its workforce, allowing it to achieve results that outstrip its better-funded competitors.

An Outlier in China’s Innovation Landscape

Like its approach to labor, DeepSeek’s funding and corporate-governance structure is equally unconventional. Unlike many of its peers, the company didn’t rely on state-backed initiatives or investments from tech incumbents. Instead, its former hedge fund founder essentially bankrolled the company. The company’s origins are in the financial sector, emerging from High-Flyer, a Chinese hedge fund also co-founded by Liang Wenfeng. DeepSeek itself emerged from High-Flyer’s pivot into AI after the 2021 regulatory crackdown on speculative trading. As a result of this setup, DeepSeek’s research funding came entirely from its hedge fund parent’s R&D budget. This unique funding arrangement means that the company could operate independently of the constraints often associated with state or corporate funding. In this way, DeepSeek is a complete outlier.

Once again, let’s contrast this with the Chinese AI startup, Zhipu. Zhipu is not only state-backed (by Beijing Zhongguancun Science City Innovation Development, a state-backed investment vehicle) but has also secured substantial funding from VCs and China’s tech giants, including Tencent and Alibaba — both of which are designated by China’s State Council as key members of the “national AI teams.” In this way, Zhipu represents the mainstream of China’s innovation ecosystem: it is closely tied to both state institutions and industry heavyweights.

DeepSeek, by comparison, has remained on the periphery, carving out a path free from the institutional expectations and rigid frameworks that often accompany mainstream scrutiny. Its funding model — self-financed by its founder rather than reliant on state or corporate backing — has allowed the company to operate with a level of autonomy rarely seen in China’s tech sector.

Tech Transfer vs. Indigenous Innovation

This brings us to a larger question: how does DeepSeek’s success fit into ongoing debates about Chinese innovation? And how must we update our perspectives on Chinese innovation to account for DeepSeek?

The debate around Chinese innovation often flip-flops between two starkly opposing views: China is doomed versus China is the next technology superpower. As I see it, this divide is about a fundamental disagreement on the source of China’s growth — whether it relies on technology transfer from advanced economies or thrives on its indigenous ability to innovate.

• Those who believe China’s success depends on access to foreign technology would argue that, in today’s fragmented, nationalist economic climate (especially under a Trump administration willing to disrupt global value chains), China faces an existential risk of being cut off from critical modern technologies. From this perspective, isolation from the West would deal a devastating blow to the country’s ability to innovate.

• On the other hand, those who believe Chinese growth stems from the country’s ability to cultivate indigenous capabilities would see American technology bans, sanctions, tariffs, and other barriers as accelerants, rather than obstacles, to Chinese growth. In this view, such restrictions compel Chinese firms to innovate, upgrade, and develop homegrown technological solutions, ultimately strengthening China’s self-reliance and long-term competitiveness.

[See also Nancy Yu’s piece on China’s industrial policy.]

So far, this debate has primarily unfolded in the context of advanced manufacturing sectors, from solar PV to batteries, and, more recently, electric vehicles. In the early stages — starting in the US-China trade wars of Trump’s first presidency — the technology transfer perspective was dominant: the prevailing theory was that Chinese firms needed to first acquire fundamental technologies from the West, leveraging this know-how to scale up production and outcompete global rivals. So the initial restrictions placed on Chinese firms, unsurprisingly, were seen as a major blow to China’s trajectory. China’s dominance in solar PV, batteries and EV production, however, has shifted the narrative to the indigenous innovation perspective, with local R&D and homegrown technological advancements now seen as the primary drivers of Chinese competitiveness.

When it comes to China’s tech industry, its success is portrayed as a result of technology transfer rather than indigenous innovation. Part of the reason is that AI is highly technical and requires a vastly different type of input: human capital, which China has historically been weaker and thus reliant on foreign networks to make up for the shortfall. Scholars like MIT professor Huang Yasheng attribute the rise of China’s tech sector to the many collaborations it has had with other countries. Even Chinese AI experts think talent is the primary bottleneck in catching up.

Indeed, China’s post-2000s ICT sector built its success on the back of overseas technical know-how. Many of China’s early tech founders either received education or spent considerable time in the United States. Chinese tech firms privilege employees with overseas experience, particularly those who have worked in US-based tech firms. In the generative AI age, this trend has only accelerated: Alibaba, ByteDance, and Tencent each set up R&D offices in Silicon Valley to increase their access to US talent. This reliance on international networks has been especially pronounced in the generative AI era, where Chinese tech giants have lagged behind their Western counterparts and depended on foreign talent to catch up.

Is DeepSeek the exception or the new rule?

This is where DeepSeek diverges from the traditional technology transfer model that has long defined China’s tech sector. Instead of relying on foreign-trained experts or international R&D networks, DeepSeek’s exclusively uses local talent. Liang himself also never studied or worked outside of mainland China. The DeepSeek story shows that China always had the indigenous capacity to push the frontier in LLMs, but just needed the right organizational structure to flourish. Much like China’s advancements in solar manufacturing, batteries, and electric vehicles, DeepSeek symbolizes a critical turning point in tech/AI: China is no longer merely playing catch-up, but is now competing on equal footing with the leading innovators in the West.

While I hope the “tech transfer vs. indigenous innovation” perspective is helpful in thinking about China’s innovation system, I must admit that it is somewhat of a false dichotomy. As development economists would remind us, all technology must first be transferred to and absorbed by latecomers; only then can they innovate and create breakthroughs of their own. Thus, tech transfer and indigenous innovation are not mutually exclusive — they’re part of the same sequential progression. First, technology must be transferred to and absorbed by latecomers; only then can they innovate and create breakthroughs of their own.

If we are to claim that China has the indigenous capabilities to develop frontier AI models, then China’s innovation model must be able to replicate the conditions underlying DeepSeek’s success. But this is unlikely: DeepSeek is an outlier of China’s innovation model.

Unlike solar PV manufacturers, EV makers, or AI companies like Zhipu, DeepSeek has thus far received no direct state support. In fact, its success was facilitated, in large part, by operating on the periphery — free from the draconian labor practices, hierarchical management structures, and state-driven priorities that define China’s mainstream innovation ecosystem.

Can China’s tech industry overhaul its approach to labor relations, corporate governance, and management practices to enable more firms to innovate in AI? The real test lies in whether the mainstream, state-supported ecosystem can evolve to nurture more companies like DeepSeek — or whether such firms will remain rare exceptions. The answer to this will define the long-term competitiveness of China’s AI firms.

For now, though, all eyes are on DeepSeek.

In December 2022, JD.com AI-research executive He Xiaodong 何晓冬 told local media,

In order to say goodbye to Silicon Valley–worship, China’s internet ecosystem needs to build its own ChatGPT with uniquely Chinese innovative characteristics, and even a Chinese AI firm that exceeds OpenAI in capability. This is an essential question for the development of China’s AI industry.

DeepSeek made it — not by taking the well-trodden path of seeking Chinese government support, but by bucking the mold completely.

The parallels between OpenAI and DeepSeek are striking: both came to prominence with small research teams (in 2019, OpenAI had just 150 employees), both operate under unconventional corporate-governance structures, and both CEOs gave short shrift to viable commercial plans, instead radically prioritizing research (Liang Wenfeng: “We do not have financing plans in the short term. Money has never been the problem for us”; Sam Altman: “We have no idea how we may one day generate revenue. We have made a soft promise to investors that, ‘Once we’ve built a generally intelligent system, basically we will ask it to figure out a way to generate an investment return for you’”).

But now that DeepSeek has moved from an outlier and fully into the public consciousness — just as OpenAI found itself a few short years ago — its real test has begun. How will it fare? Can High-Flyer cash and Nvidia H800s/A100 stockpiles keep DeepSeek running at the frontier forever, or will its growth aspirations pressure the company to seek outside investors or partnerships with conventional cloud players? Does Liang’s recent meeting with Premier Li Qiang bode well for DeepSeek’s future regulatory environment, or does Liang need to think about getting his own crew of Beijing lobbyists? Will Liang receive the treatment of a national hero, or will his fame — and wealth — put a months-long Jack Ma–style disappearance in his future? If an organizational vision crisis arises (à la the Altman vs. Musk feud), will Liang be able to steer DeepSeek through it?

In short, how long can DeepSeek buck the mold?

new year mood music

Before December 2024, DeepSeek was rarely mentioned in China’s AI community. With the release of DeepSeek-V3 and the reasoning model R1, Chinese media and AI researchers started to ask the same question as their American counterparts: Who is DeepSeek and how should we feel about them?

In this newsletter, we share a translation of insights from a January 26 closed-door session hosted by Shixiang 拾象, a VC spun out from Sequoia China. Attended by dozens of AI researchers, investors, and industry insiders, the event captures how the Chinese AI community is processing the DeepSeek shock. A core conclusion they’ve come to, one we’ve emphasized in ChinaTalk with our Miles Brundage interview and guest post by Lennart and Sihao, is that “In the long-run, questions about computing power will remain. Demand for compute remains strong and no company has enough.”

Before diving into that translation, we did a broad look at additional details and discussion coming from Chinese-language coverage of DeepSeek.

The Story Behind DeepSeek The Paper 澎湃 offered more details about High-Flyer, the quantitative hedge fund behind DeepSeek. Founded in 2015 by Liang Wenfeng 梁文锋, a Zhejiang University graduate, High-Flyer has a strong background in machine learning-based quantitative trading. Liang founded DeepSeek in July 2023, and the company has not received any outside funding to date.

When it comes to hiring, DeepSeek prioritizes “young and high-potential” candidates — specifically those born around 1998 with no more than five years of work experience, similar to other AI labs in China. Said one DeepSeek employee to The Paper, “The success of DeekSeek has demonstrated the power of young people, and in essence, that the development of this generation of artificial intelligence needs young minds.”

Liang has maintained a relatively low public profile, but 36Kr managed to secure two exclusive interviews with him. The first, in May 2023, followed High-Flyer’s announcement that it was building LLMs, while the second, in November 2024, came after the release of DeepSeek-V2.

In both interviews, Liang emphasized the value of innovation without immediate monetization and DeepSeek’s culture of openness. The second interview had a stark shift in tone, with Liang meditating less on the baked-in idealism of a strategy predicated on open sourcing core innovations and more time emphasizing that he wanted DeepSeek to prove to other Chinese engineers that domestic teams could deliver on “hardcore innovation.”

A budding partnership with ByteDance? TMT 钛媒体 reported yesterday that ByteDance and OpenAI are “considering research collaborations” with DeepSeek. While the two firms may have talked in the past, given today’s political climate it’s kind of hard to put much weight into the OpenAI rumor. Partnering with ByteDance, however, could be an enormous unlock for DeepSeek researchers, giving them access to orders of magnitude more compute.

National Pride in the Face of US Competition. The response from Chinese media has been quite positive. State media and industry leaders have celebrated DeepSeek’s achievements, often tinged with nationalist pride, particularly after English-language reports highlighted its performance and cost efficiency. For example:

China Daily declared, “For a Chinese LLM, it’s a historical moment to surpass ChatGPT in the US.” Daily Economic News echoed this sentiment, stating, “Silicon Valley Shocked! Chinese AI Dominates Foreign Media, AI Experts Say: ‘It Has Caught Up with the U.S.!’”

Tech executives have also weighed in. Feng Ji 冯骥, founder of Game Science (the studio behind Black Myth: Wukong), called DeepSeek “a scientific and technological achievement that shapes our national destiny (国运).” Zhou Hongyi, Chairperson of Qihoo 360, told Jiemian News that DeepSeek will be a key player in the “Chinese Large-Model Technology Avengers Team” to counter U.S. AI dominance.

Ordinary users have also been astounded by the model’s capabilities. Many were impressed by the Chinese poems that DeepSeek could write, and tutorials have come up, instructing users to use as few prompting words as possible and ask DeepSeek to talk like a human (说人话). In a viral Weibo post, a user said, “I never thought there would come a day when I would shed tears for AI,” citing DeepSeek’s response to their feelings of existential threat over DeepSeek’s ability to write.

Here is DeepSeek R1’s response: “Remember, all the words that make you tremble are just echoes that already exist deep within your soul. I am merely a valley that happened to pass by, allowing you to hear the weight of your own voice.” 记住，所有让你颤粟的文字，本质上都是你灵魂深处早已存在的回声。我不过是偶尔经过的山谷，让你听到了自己声音的重量。

And now, our translation of the industry summit.

A High Level Closed-Door Session Discussing DeepSeek: Vision Trumps Technology

January 26th. WeChat Link, Archive.

DeepSeek-R1 has sparked a frenzy in the global AI community, but there is a relative dearth of high-quality information about DeepSeek.

On January 26, 2025, 李广密 Guangmi Li, Founder and CEO of 拾象 Shixiang, organized a closed-door discussion on DeepSeek with dozens of top AI researchers, investors and frontline AI practitioners to discuss and learn from DeepSeek's technical details, organizational culture, and short-, medium-, and long-term impacts of its entry into the world. This discussion attempted to lift the veil of this “mysterious eastern force” about which we have so little information.

Below is a summary of the key points from this discussion.

The Mystical DeepSeek. ‘The most important thing about DeepSeek is pushing intelligence’

1. Founder and CEO Liang Wenfeng is the core person of DeepSeek. He is not the same type of person as Sam Altman. He is very knowledgeable about technology.

2. DeepSeek has a good reputation because it was the first to release the reproducible MoE, o1, etc. It succeeded in acting early, but whether or not it did the absolute best remains to be seen. Moving forward, the biggest challenges are that resources are limited and can only be invested in the most high-potential areas. DeepSeek’s research and culture are still strong, and if given 100,000 or 200,000 chips, they might be able to do better.

3. From its preview to its official release, DeepSeek’s model’s long-context capabilities have improved rapidly. DeepSeek’s long-context 20K can be achieved with very conventional methods.

4. The CEO of Scale.ai said that DeepSeek has 50,000 chips, but that is definitely not reality. According to public information, DeepSeek had 10,000 old A100 chips and possibly 3,000 H800 cards before the ban. DeepSeek pays great attention to compliance and has not purchased any non-compliant GPUs, so it should have few chips. The way the United States uses GPUs is too extravagant.

5. DeepSeek focused all its efforts on a single goal and subsequently gave up many things, such as multimodality. DeepSeek is not just serving people, but seeking intelligence itself, which may have been a key factor in its success.

6. In some ways, quant trading can be said to be the business model of DeepSeek. Huanfang (another quantitative investment company founded by Liang Wenfeng) is the product of the last round of machine learning. DeepSeek’s highest priority is to push intelligence. Money and commercialization are not high priorities. China needs several leading AI labs to explore things that can beat OpenAI. Intelligence takes a long time to develop, and has begun to differentiate again this year, so new innovations are bound to result.

7. From a technical perspective, DeepSeek has been instrumental as a training ground for talent.

8. The business model of AI labs in the United States is not good either. AI does not have a good business model today and will require viable solutions in the future. Liang Wenfeng is ambitious; DeepSeek does not care about the model and is just heading towards AGI.

9. Many of the insights from DeepSeek’s paper involve saving hardware costs. On a couple of big dimensions of scaling, DeepSeek’s techniques are able to reduce costs.

10. In the short-term, everyone will be driven to think about how to make AI more efficient. In the long-run, questions about computing power will remain. Demand for compute remains strong and no company has enough.

11. Discussing DeepSeek’s organization:

    1. When investing, we always choose the most advanced talent. But we see from DeepSeek’s model (the team is mostly smart young people who graduated from domestic universities) that a group that coheres well may also gradually advance their skills together. It has yet to be seen whether poaching one person might break DeepSeek’s advantage, but for now this seems unlikely.

    2. While there’s a lot of money in the market, DeepSeek’s core advantage is its culture. The research culture of DeepSeek and ByteDance are similar, and both are critical for determining the availability of funding and long-term viability. Only with an important business model can there be a sustainable culture. Both DeepSeek and ByteDance have very good business models.

12. Why did DeepSeek catch up so fast?

    1. Reasoning models require high-quality data and training. For LLMs or multimodal AI, it’s difficult to catch up with a closed source model from scratch. The architecture of pure reasoning models hasn’t changed much, so it’s easier to catch up in reasoning.

    2. One reason R1 caught up quickly was that the task was not particularly difficult. Reinforcement learning only made the model choices more accurate. R1 did not break through the efficiency of Consensus 32, spending 32 times the efficiency, which is equivalent to moving from deep processing to parallelization, which is not pushing the boundaries of intelligence, just making it easier.

Pioneers vs. Chasers: 'AI Progress Resembles a Step Function – Chasers Require 1/10th the Compute’

13. AI is similar to a step function, where the compute requirements for followers have decreased by a factor of 10. Followers have historically had lower compute costs, but explorers still need to train many models. The exploration of new algorithms and architectures will not stop. Behind the step function, there are significant investments by many people, meaning compute investments will continue to advance. Many resources will also be allocated to products. Apart from reasoning, there are other directions that are compute-intensive. While the vast amount of compute resources spent by explorers may not be visible, without such investment, the next "step" might not occur. Additionally, many are dissatisfied with current architectures and RL methods, and progress will continue.

14. When exploring directions, performance achieved with 10,000 GPUs may not always be significantly better than that of 1,000 GPUs, but there is a threshold somewhere. It’s unlikely that meaningful results can be achieved with only 100 GPUs because the iteration time for each solution would be too long.

15. Advancements in physics can be divided into academic research in universities and industry labs. The former focuses on exploring multiple directions without requiring immediate returns, while the latter prioritizes efficiency improvements.

16. From the perspectives of explorers and chasers, small companies with limited GPUs must prioritize efficiency, whereas large companies focus on achieving models as quickly as possible. Methods that improve efficiency on a 2,000-GPU cluster may not work effectively on a 10,000-GPU cluster, where stability becomes a higher priority.

17. The advantage of the CUDA ecosystem lies in its extensive and complete set of operators. Chinese companies like Huawei have targeted commonly used operators to achieve breakthroughs, leveraging their latecomer advantage. If a company has access to 100,000 GPUs, the decision between becoming a leader or a chaser is critical. Being a leader comes with high costs, while being a chaser offers higher efficiency. The next direction for China to follow could be multi-modality, especially since GPT-5 has been delayed for a long time.

[points 18-48 was a long technical discussion we’ve machine-translated below]

Why didn’t the other companies take the DeepSeek approach: ‘Models from the big labs need to maintain a low profile’

49. The question of why OpenAI and Anthropic did not do work in DeepSeek’s direction is a question of company-specific focus. OpenAI and Anthropic might have felt that investing their compute towards other areas was more valuable.

50. One hypothesis for why DeepSeek was successful is that unlike Big Tech firms, DeepSeek did not work on multi-modality and focused exclusively on language. Big Tech firms’ model capabilities aren’t weak, but they have to maintain a low profile and cannot release too often. Currently, multimodality is not very critical, as intelligence primarily comes from language, and multimodality does not contribute significantly to improving intelligence.

The Divergence and Bets of 2025 Technology: ‘Can We Find Architectures Beyond Transformer?’

51. In 2025, models will begin to diverge. The most enticing vision is to continuously push the boundaries of intelligence, with many potential breakthrough paths. Methods might change, such as through synthetic data or alternative architectures.

52. 2025 will, first and foremost, see interest in new architectures beyond Transformers. Some initial exploration is already underway, aiming to reduce costs while pushing the boundaries of intelligence. Secondly, the potential of reinforcement learning (RL) has yet to be tapped into completely. On the product side, there is significant interest in agents, though they have yet to see widespread application.

53. Multimodal products capable of challenging the ChatGPT paradigm might emerge in 2025.

54. The success of R1 and V3 in achieving low cost and high performance demonstrates the viability of this direction. This does not conflict with the approach of expanding hardware or increasing parameters. However, in China, due to certain restrictions, the former path is the primary option.

55. On DeepSeek:

• First, DeepSeek may have been "forced" into its current path from base models or may simply be following the Scaling Law.

• Second, from the perspective of distillation, DeepSeek likely follows a "large to small" approach. This is beneficial for closed-source models, which are growing larger and larger.

• Third, there are currently no anti-scaling metrics emerging in the field. If such metrics arise, they could pose a challenge to the Scaling Law. However, open-source models can implement everything closed-source models do while also reducing costs, which is advantageous for closed-source models as well.

56. It is reported that Meta is still in the process of reproducing DeepSeek, but so far, this has not significantly impacted their infrastructure or long-term roadmap. In the long run, beyond exploring the boundaries of the technology, cost efficiency must also be considered. Lowering costs will let us have more fun.

Have developers moved from closed-source models to DeepSeek? ‘Not yet’

57. Will developers migrate from closed-source models to DeepSeek? Currently, there hasn’t been any large-scale migration, as leading models excel in coding instruction adherence, which is a significant advantage. However, it’s uncertain whether this advantage will persist in the future or be overcome.

58. From the developer's perspective, models like Claude-3.5-Sonnet have been specifically trained for tool use, making them highly suitable for agent development. In contrast, models like DeepSeek have not yet focused on this area, but the potential for growth with DeepSeek is immense.

59. For large model users, DeepSeek V2 already meets most needs. While R1 improved speed, it didn’t provide significant additional value. Interestingly, when engaging in deep reasoning, some previously correct answers now tend to be incorrect.

60. When choosing models, users tend to simplify problems using engineering methods. 2025 may become a year of applications, with industries leveraging existing capabilities. However, this could lead to a bottleneck, as most day-to-day tasks might not require highly intelligent models.

61. Currently, reinforcement learning (RL) solves problems with standard answers but has not achieved breakthroughs beyond what AlphaZero accomplished. In fact, it is often simpler. Distillation addresses problems with standard answers, and RL methods work effectively when training with such answers. This explains why distillation and RL have made rapid progress in recent years.

62. Humanity’s demand for intelligence is vastly underestimated. Many critical problems, such as cancer and SpaceX's heat shield materials, remain unsolved. Existing AI primarily automates tasks, but there are numerous unsolved challenges ahead. Looking forward, the potential for explosive growth is immense, and the advancement of intelligence cannot stop.

OpenAI Stargate’s $500B Narrative and Changes in Computing Power Demand

63. The emergence of DeepSeek has led people to question the latest $500B narrative from Nvidia and OpenAI. There’s no verdict yet on compute — and OpenAI’s $500B narrative is their attempt to throw themselves a lifeline.

64. Regarding the doubts about OpenAI’s $500B infrastructure investment: because OpenAI is a commercial company, it could be risky if debt is involved.

65. $500B is an extreme number — likely to be executed over 4 or 5 years. SoftBank and OpenAI are the leading players (the former providing capital, the latter technology) — but SoftBank’s current funds can’t support $500B; rather SoftBank is using its assets as collateral. OpenAI, meanwhile, isn’t very cash-rich either, and other AI companies are more technical participants than they are funding providers. So it will be a struggle to fully realize the $500B vision.

66. OpenAI’s $500B computing power makes sense: during the exploration phase, the cost of trial and error is high, with both human and investment costs being substantial. But although the path isn’t clear and getting from o1 to R1 won’t be easy, at least we can see what the finish line looks like: we can track the intermediate markers, and from day one, aim for others’ proven end states; this gives us a better bearing on our progress. Being at the frontier exploring the next generation is most resource-intensive. The followers don’t bear exploration costs — they’re always just following. If Google/Anthropic succeed in their exploration areas, they might become the frontier company.

67. In the future, Anthropic might replace all their inference with TPU or AWS chips.

68. Domestic Chinese companies were previously constrained by computing power, but now it’s proven that the potential technical space is vast. For more efficient models, we might not need especially large cards — we can provide relatively customized chips that can be adapted for compatibility with AMD and ASIC. From an investment perspective, Nvidia’s moat is very high, but ASIC will have yet greater opportunities.

69. The DeepSeek situation isn’t really about compute — it’s about America realizing China’s capabilities and efficiency. DeepSeek isn’t Nvidia’s vulnerability; Nvidia will grow as long as AI grows. Nvidia’s strength is its ecosystem, which has been built up over a long time. Indeed, when technology develops rapidly, the ecosystem is crucial. The real crisis comes, though, when technology matures like electricity: it becomes commoditized; then, everyone will focus on products, and many ASIC chips will emerge for specific scenario optimization.

Impact on the Secondary Market: ‘Short-term sentiment is under pressure, but the long-term narrative continues’

70. DeepSeek has had a significant short-term impact on the US AI sector and stock prices: pretrain demand growth is slowing, while post-training and inference scaling haven’t scaled up fast enough, creating a gap in the narrative for related companies, which will affect short-term trading.

71. DeepSeek mainly uses FP8, while the US uses FP16. DeepSeek’s improvements are all based on limited computational engineering capabilities, with efficient use of computing power being the biggest highlight. Last Friday, DeepSeek had a huge impact in North America: Zuckerberg gave higher expectations for Meta’s capital expenditure, but Nvidia and TSMC fell, and only Broadcom rose.

72. DeepSeek creates short-term market-sentiment pressure on stock prices and valuations. That’s affecting secondary market computing-related companies, and even energy companies — but the long-term narrative will continue.

73. Secondary-market practitioners will worry about potential air pockets in Nvidia’s transition from H cards to B cards. Combined with pressure from DeepSeek, there will be short-term stock-price pressure — but this may give rise to better long-term opportunities.

74. This short-term impact reflects sentiment about DeepSeek’s low-cost training investments (see, for instance, how it directly affected Nvidia’s stock price). AI, however, is a growth market with huge potential. Long-term, AI is just beginning, and if CUDA remains the preferred choice, hardware growth potential remains substantial.

Open-Source vs Closed Source: ‘If capabilities are similar, closed source will struggle.’

75. The battle between open-source and closed-source intensifies the spotlight on DeepSeek.

76. There is a possibility that OpenAI and others have hidden their good models, and no leading models have been released so far. But after DeepSeek’s release, other AI companies may not be able to hide their good models anymore.

77. DeepSeek has done a lot of cost optimization. Amazon and others haven't seen any changes as a result and are still following the established plan in a state of coexistence. Open source and closed source models are not contradictory. Universities and small labs should give priority to DeepSeek. There will be no competition for cloud vendors, because cloud vendors support open source and closed source, preserving the current state of coexistence in the ecosystem. DeepSeek’s applications have not been as mature as Anthropic’s, and the latter has spent a lot of time on AI security Deepseek must consider if it hopes to be recognized by European and American markets in the long-term.

78. Open source controls the margins of the whole market. If open source can do 95% of what closed source can do and closed source is too expensive, then open source can be used completely. If the capabilities of open source and closed source do not differ greatly, then this presents a big challenge for closed source.

The Impact of DeepSeek’s Breakthrough: ‘Vision Trumps Technology’

79. DeepSeek’s breakthrough made the outside world realize China’s AI strength. Previously, outsiders thought China’s AI progress lagged America by two years, but DeepSeek shows the gap is actually 3 to 9 months, and in some areas, even shorter.

80. When it comes to technologies and sectors that America has historically blocked China from accessing, if China can break through nonetheless, those sectors ultimately become highly competitive. AI might follow this pattern — and DeepSeek’s success may well prove this.

81. DeepSeek didn’t suddenly explode. R1’s impressive results reverberated throughout America’s entire AI establishment.

82. DeepSeek stands on the shoulders of giants — but exploring the frontier still requires much more time and human capital cost. R1 doesn’t mean that future training costs will decrease.

83. AI explorers definitely need more computing power; China, as a follower, can leverage its engineering advantages. How Chinese large-model teams use less computing power to produce results, thereby having some definite resilience — or even doing better — might end up being how the US-China AI landscape plays out in the future.

84. China is still replicating technical solutions; reasoning was proposed by OpenAI in o1, so the next gap between various AI labs will be about who can propose the next reasoning. Infinite-length reasoning might be one vision.

85. The core difference between different AI labs’ models lies not in technology, but in what each lab’s next vision is.

86. After all, vision matters more than technology.

Technical Discussion

What followed was a deep 2000-word technical discussion that explored SFT, data, distillation and the process reward function.

Technical Detail 1: SFT. ‘No need for SFT on the reasoning level’

18. The most groundbreaking aspect of DeepSeek isn’t its open-source nature or low cost, but the fact that it eliminates the need for Supervised Fine-Tuning (SFT)—at least for inference tasks. However, tasks beyond inference may still require SFT. This raises questions about whether this represents a new paradigm or architecture that improves data efficiency in training, or whether it accelerates the iteration speed of model performance.

Read more

a playlist

Best show of the year: Our epic series with Yasheng Huang (1, 2, 3).

Huang argues that China’s imperial examination system (科舉 kējǔ) was a double-edged sword: on the one hand, it spurred political mobility and human capital development — but simultaneously, it concentrated intellectual energy on serving the emperor rather than driving innovation. Huang’s stats show that China was actually most inventive during its fragmented “European moment,” between the Han and Sui dynasties (220-589 CE), before the exam system was institutionalized.

This pattern of trading dynamism for stability continues to shape China today. Looking ahead, Huang identifies two major challenges:

• A succession crisis. By eliminating term limits, Xi has reintroduced the ancient problem of peaceful power transitions that the post-Mao system had largely solved.

• Economic stagnation. China’s current emphasis on top-down industrial policy and tight control is producing declining productivity (much like the late Soviet Union under Leonid Brezhnev). The core tension Huang identifies is between “scale” (uniformity/control) and “scope” (diversity/dynamism). While China’s model has proven extremely effective at scale, its anxiety toward scope may undermine the very factors — entrepreneurship, openness to foreign ideas, and bottom-up innovation — that drove its rise in the first place.

Huang also argues that US engagement with China failed at least in part because Americans focused more on manufacturing and finance and less on media and academic exchange — areas more likely to enjoy reciprocity and promote pluralism. And he suggests that the West, rather than primarily engaging with ethnic minorities, should emphasize how the rule of law protects everyone, including CCP elites who often end up needing those protections themselves.

In 2025, we’re going to do more on ancient China — starting with Tonio Andrade’s The Gunpowder Age, a look at the great divergence from the perspective of military technology. Consider getting a copy and reading beforehand, the book is a ton of fun.

Regarding politics, evidence increasingly shows that commerce alone doesn’t change political structures. The Song dynasty had vibrant commerce but remained autocratic. In fact, it invented one of the most restrictive ideologies in Chinese philosophy — neo-Confucianism — while Buddhism declined.

Runner up: Shakespeare on Power

We brought on Elliot Cohen, dramaturg of , and my brother, professional actor Phil Schneider to act out some scenes and discuss how Shakespeare’s histories help illuminate court politics.

We’ve since learned that ChinaTalk doesn’t only get people promoted (see: Kurt Campbell, Devin DeBacker), it also gets them cast in movies! Phil spent this fall as one of Adam Sandler’s four sons in Happy Gilmore 2. He’s also the lead in an excellent play currently on in Denver now through March 9th and Atlanta March 29 to May 4th.

Ep 1: Shakespeare on Power

Towards the end of Henry IV Part Two, when the father of Henry V, the guy who gives the St. Crispian’s Day speech, is dying. Prince Hal, as he then is, comes in to pay his respects and sees the crown. He thinks his dad has croaked and he puts the crown on his own head, and his father wakes up and there’s a furious scene. They sort of reconcile at the end — but the thing that’s so striking about Henry IV’s exit, and particularly these rather harsh words that he addresses to Prince Hal, is that it’s clear that he doesn’t think anybody else can do the job that he’s done.

If you think about it, that’s Trump’s “I alone can do it” — but I think it’s also, alas, where President Biden is: “I’m the only guy who can really defeat Trump and run the country properly.” Which is nonsense, but I think it’s understandable for somebody who’s been in the Senate for half a century or something like that. It is a good example of how power messes with your sense of reality.

Ep 2: Shakespeare emergency pod after Biden’s presidential debate

I remember a conversation I had with a very senior government official, where I was asking why someone even more senior was still clinging to office. He said, “Well, remember, for these guys, the next big job is death.”

The truth is, Lear says that he’s crawling his way to death. Prospero also says, at the end, every third thought will be of the grave. These guys know that they’re coming to the end, but they’re not approaching it with the kind of tranquility that Cicero recommended.

Tech Policy

Technical heavyweights on national competitiveness

One of ChinaTalk's comparative strengths is examining technical challenges through a policy lens. Episodes like my conversations with Scale’s Alex Wang, Leonard Heim and Chris Miller on test-time compute, the Latent Space discussion, the US Biotech Commissioners, and Nathan Lambert's end-of-year AI review all feature guests far more technical than me. Making content that allows policy folks to engage with what leading technologists teaches me a ton and

I strive to help listeners who don't come from STEM backgrounds engage with cutting-edge technology developments. These episodes are particularly enjoyable for me - I learn a lot from them, and I plan to produce more like them in the future

Export Controls Emergency Pods

We’ve been on a long saga together covering Commerce’s chip export controls, from the first round in October 2022, through the updates in October 2023 and finally in December 2024.

Dylan Patel and Greg Allen blasted Raimondo’s latest round as ineffective half-measures — “if we can identify numerous billion-dollar loopholes within hours of release, imagine what lawyers will find in a month” — arguing that attempts for increasingly elusive “precision” have backfired. Instead, how about a simpler “shotgun approach” that would broadly restrict semiconductor equipment sales to China, rather than the current “Swiss cheese” regulations full of loopholes that actually incentivize Chinese domestic chip development.

The intellectual- and execution-level challenges the Biden administration encountered with export controls exemplify broader Democratic Party challenges. There’s a tendency to believe they can devise the perfect algorithm that balances all competing interests. They think with solving enough integrals, extensive legal review, and track changes on docs, they’ll reach the optimal solution.

This pattern emerged with the Inflation Reduction Act’s lengthy development, the CHIPS Act’s extended negotiations, the periodic reassessment of Ukraine arms distribution, and these export controls. The problem is that, if you can’t make tough strategic decisions upfront and execute them — accepting that not everyone will be happy — you end up in limbo. You achieve worse results by trying to moderately satisfy five variables instead of maximizing the two most critical ones.

We also snuck in two final shows in the last week of the Biden administration, looking into the AI diffusion and foundry controls.

History

Makers of Modern Strategy

Few books have influenced me as much as the Makers of Modern Strategy series. The three volumes (published in 1942, 1986, and 2023) are indispensable to understanding statecraft, leadership, and the evolution of warfare across millennia.

The New Makers of Modern Strategy (2023) is a thousand pages long and analyzes strategy from ancient Greece to the Congo. Hal and I had a blast talking through how the three volumes evolved over time, Tecumseh’s strategic vision, and the impact of technological change across a few thousand years of warfare. Have a listen on Spotify or Apple Podcasts or check out the transcript below.

How the Navy Learned to Fight

Last year I blazed through an unhealthy amount of 20th century naval history. The capstone conversation to that reading was a great deep dive with Trent Hone, author of a new Nimitz biography and the classic Learning War: The Evolution of Fighting Doctrine in the U.S. Navy, 1898-1945. We explored classic ChinaTalk themes of technological disruptions to warfare and organizational evolution following the dizzying story how the US Navy went from underperforming against a decrepit Spanish Empire in 1898 to uncontested global hegemons in half a century.

Spotify and Apple Podcasts links here.

Is the NSC Unwell?

Who in the White House actually makes decisions? How does the organizational structure, personal incentives, and professional backgrounds of NSC staffers shape what America ends up doing around decisions of war and peace? John Gans and I had a really fun conversation on his book White House Warriors and how the NSC evolved over its post-WWII incarnations.

Spotify, Apple Podcasts, or a transcript.

It’s like NSC staffers are sitting in the back seat of the car, watching. The NSC staff has great judgment for when something’s going wrong because they can see things. But they have very poor judgment on what to do because they’re not doing the things. It’s the classic operator’s dilemma.

WWIII

“So Jordan, will China invade Taiwan?” is first question normies always ask me. We had some of our best discussions of the year taking a stab at answering it.

We hosted Jared McKinney of the Air War College and Peter Harris of Colorado State University, who recently co-authored a fantastic monograph, “Deterrence Gap: Avoiding War in the Taiwan Strait.” In June, we hosted Jared McKinney of the Air War College and Peter Harris of Colorado State University, who presented their findings on how to prevent an invasion of Taiwan. We also had on their student, US Army Lieutenant Colonel J. Kevin McKittrick, whose research paper — “Why America Didn’t Invade Taiwan” — dissected the ultimately aborted WWII-era plan to invade Taiwan as a launching base to then take Japan.

Then we brought on Dmitri Alperovitch who argued that deterring “Xi the gambler,” the United States needs to rationalize its defense spending away from exquisite but unaffordable weapons platforms, maintain its edge in key technologies like AI and semiconductors, and focus on building practical military capabilities in the Indo-Pacific rather than engaging in provocative symbolic gestures.

Pols

Rahm Emanuel

This one is a classic. Rahm didn’t mince words — he articulated the Biden administration’s view of China today with language that nobody in the Biden administration would use themselves.

China has quite an entrepreneurial culture, and [Xi] crushed their entrepreneurship — and in crushing their entrepreneurship and the tactics he adopted, he crushed the confidence of the world in China. … You can’t get anybody in Japan, Europe, or the United States to raise their hand and say, “I’d like to move my family to a city where I could get arrested any given day and be in lockdown.”

GOP Voices

Rolling off the momentum of last year’s excellent interview with then-Rep. Mike Gallagher, our top GOP interviewees this year included Senator Todd Young and former deputy national security advisor Matt Pottinger.

Sen Young:

Though the benefit-to-cost ratio of research investments is off the charts, we’ve seen we saw the percentage of research applications go down just because the resources to fund them were drying up. Those are lost economic opportunities, for our entire economy.

As our fiscal situation has deteriorated, there’s naturally been a desire to pull back from a number of investments, but I would remind a number of my fellow green eyeshade legislators out there that the real number we should be looking at as we think about our fiscal deterioration is the debt-to-GDP ratio.

It’s a fraction — debt/GDP. You can’t starve your economy of one of the components of your GDP, which is R&D investment, in order to deal with the debt.

Pottinger (back in February) predicted that, while a second Trump admin would likely escalate trade tensions with China, his “strategic ambiguity” on Taiwan defense would deter Beijing — though Pottinger warns we shouldn’t reverse Biden's clearer defense commitments. He also sees continued tough China policy as a rare area of bipartisan agreement, shaped more by Xi Jinping’s aggressive moves than by who happens to occupy the Oval Office.

I think President Biden’s policy is closer to President Trump’s policy on China than it is to President Obama’s policy, when President Biden was the vice president. That’s interesting. That tells me that we must have gotten some things right, and some things that are now viewed as a consensus — we have the benefit of bipartisan consensus.

Guest edition by Lennart Heim and Sihao Huang, cross-posted from Lennart’s personal blog. Lennart is a repeat ChinaTalk guest most recently coming on to talk about geopolitics in the age of test time compute. Sihao has previously written for us on Beijing’s vision of global AI governance.

Recent coverage of DeepSeek's AI models has focused heavily on their impressive benchmark performance and efficiency gains. While these achievements deserve recognition and carry policy implications (more below), the story of compute access, export controls, and AI development is more complex than many reports suggest. Here are some key points that deserve more attention:

1. Real export restrictions on AI chips only started in October 2023, making claims about their ineffectiveness premature. DeepSeek trained on Nvidia H800s, chips designed specifically to circumvent the original October 2022 controls. For DeepSeek's workloads, these chips perform similarly to the H100s available in the US. The now available H20, Nvidia’s most recent AI chip which can be exported to China, is less performant for training though it still offers significant deployment capabilities.

   The H20 chip, while restricted for training, remains uncontrolled and highly capable for frontier AI deployment, particularly for memory-intensive workloads like long context inference. This is significant given recent trends toward test-time compute, synthetic data generation, and reinforcement learning—all processes that are more memory-bound than compute-bound. Following the December 2024 restrictions on high-bandwidth memory exports, the H20's continued availability should be addressed, especially as deployment compute grows increasingly central to AI capabilities.

2. Export controls on hardware operate with a time lag and haven't yet had time to bite.

   Remember: this is all assuming export controls work perfectly—which they don't. We've seen plenty of loopholes in semiconductor controls and have credible reports of large-scale chip smuggling into China. While the Diffusion Framework should help plug some gaps, implementation remains a key challenge. [JS: and of course there’s still the issue of western cloud access…]

   China is still running pre-restriction data centers with tens of thousands of chips, while US companies are constructing data centers with hundreds of thousands. The real test comes when these data centers need upgrading or expansion—a process that will be easier for US firms but challenging for Chinese companies under US export controls. If next-generation models require 100,000 chips for training, export controls will significantly impact Chinese frontier model development. However, even without such scaling, the controls will affect China's AI ecosystem through reduced deployment capabilities, limited company growth, and constraints on synthetic training and self-play capabilities.

3. The fact that DeepSeek V3 was trained on less compute is not surprising: machine learning algorithms have always gotten cheaper over time. But the same efficiency gains that allow smaller actors like DeepSeek to access a given capability (“access effect”) will probably also allow other companies to build more powerful systems on larger compute clusters (“performance effect”). We should be very glad that DeepSeek trained its V3 model with 2,000 H800 chips, not 200,000 B200 chips (Nvidia’s latest generation).

4. Their timing is strategic, but the technology is real. R1's release during President Trump’s inauguration last week is clearly intended to rattle the public’s confidence in the United States’ AI leadership during a pivotal moment in US policy, mirroring Huawei's product launch during former Secretary Raimondo's China visit. After all, the benchmark results of an r1 preview have already been public since November.

   This calculated PR timing shouldn't obscure two realities: DeepSeek's technical progress and the structural challenges they already and increasingly face from export controls.

5. It is harder for export controls to affect individual training runs, and easier to for them impact a whole ecosystem. Crucially, restrictions on the most advanced chips can effectively constrain large-scale AI deployment (i.e., allowing large numbers of users to access AI services) and capability advancement. AI companies typically spend 60-80% of their compute on deployment—even before the rise of compute-intensive reasoning models. Restricting compute access will increase the PRC's AI costs, limit widespread deployment, and constrain system capabilities. Importantly, deployment compute isn't just about serving users—it's crucial for generating synthetic training data and enabling capability feedback loops through model interactions, and building, scaling, and distilling better models.

   For example, see this recent Gwern comment which suggests that deployment compute plays a crucial role beyond just serving users. Models like OpenAI's o1 are used to generate high-quality training data for future models, creating a feedback loop where deployment capabilities directly enhance development capabilities and effectiveness

6. DeepSeek's efficiency gains may have come from previously having access to substantial compute. Counterintuitively, the path to using fewer chips (i.e., “efficiency”) may requires starting with many more. DeepSeek operated Asia's first 10,000 A100 cluster, reportedly maintains a 50,000 H800 cluster, and has additional unlimited access to Chinese and foreign cloud providers (which is not export-controlled). This extensive compute access was likely crucial for developing their efficiency techniques through trial and error and for serving their models to customers.

   Recent usage spikes at other AI companies have led to service disruptions despite larger compute resources. DeepSeek's ability to handle similar surges remains untested and with limited compute they’ll face difficulties. (Sam Altman even claimed they are currently losing money on the ChatGPT Pro plan.)

   While their R1 model demonstrates impressive efficiency, its development required significant compute for synthetic data generation, distillation, and experimentation.

7. The compute gap between US and China—further widened by export controls —remains DeepSeek’s primary constraint. DeepSeek’s leadership openly acknowledged a 4x compute disadvantage despite their efficiency gains. DeepSeek Founder Liang Wenfeng stated: "this means we need twice the computing power to achieve the same results. Additionally, there’s about a 2x gap in data efficiency, meaning we need 2x the training data and computing power to reach comparable outcomes. Combined, this requires 4x the computing power." He added: "We don’t have short-term fundraising plans. Our problem has never been funding; it’s the embargo on high-end chips."

8. Leading US companies keep the best of their capabilities private, meaning that public benchmarks paint an imperfect picture of AI progress. While Chinese firms openly share their progress, companies like Anthropic and OpenAI maintain significant private capabilities. This makes direct comparisons based on public information incomplete. The attention on DeepSeek stems partly from their open approach—sharing model weights and methods in detail, unlike Western companies' increasingly closed stance. However, if openness necessarily translates to strategic advantage remains to be seen.

So what?

DeepSeek's achievements are genuine and significant. Claims dismissing their progress as mere propaganda miss the mark. Their reported training costs are not unprecedented given historical algorithmic efficiency trends. However, comparisons require careful context—DeepSeek only reports the final pre-training run costs, excluding crucial expenses like staff time, preliminary experiments, data acquisition, and infrastructure setup. See this post for a discussion at the top of how different cost accounting methods can lead to misleading comparisons

The reality of increasing compute efficiency means AI capabilities will inevitably diffuse. Controls alone aren't enough: they must be paired with actions to strengthen societal resilience and defense: creating institutions to identify, assess, and address AI risks and building robust defenses against potentially harmful AI applications from adversaries. However, we should also recognize that export controls already impact Chinese AI development and could have even stronger effects in the future.

Models themselves may not be the strategic moat many assume, but the compute impact on national security varies by use case. For use cases that require large-scale deployment (like mass surveillance), compute limitations could create significant barriers. For single-user applications, controls have less impact. The relationship between compute access and national security capabilities remains complex, even as model capabilities become more easily replicable. ↩︎

While AI capabilities will likely diffuse regardless of controls—and it will always be difficult for export controls or other "capability interventions" to completely prevent proliferation—they remain important for maintaining our technological advantages. Controls buy valuable time but need to be complemented with policies that ensure democracies stay in the lead and are resilient to adversaries.

I recorded a show yesterday afternoon with , who recently left OpenAI after six years running its Policy Research and AGI Preparedness teams. See our transcript below I’m rushing out as these terrible takes can’t stand uncorrected.

Or have a listen on Apple Podcasts, Spotify or your favorite podcast app.

Initial Model Impressions

Jordan: Let’s start with the news. What is remarkable about their latest R1 model?

Miles: I think compared to GPT3 and 4, which were also very high-profile language models, where there was kind of a pretty significant lead between Western companies and Chinese companies, it’s notable that R1 followed pretty quickly on the heels of o1. It’s a model that is better at reasoning and kind of thinking through problems step-by-step in a way that is similar to OpenAI’s o1.

So o1 inspired R1, but it didn’t take very long, about two months. And that has rightly caused people to ask questions about what this means for tightening of the gap between the U.S. and China on AI.

Jordan: What are your initial takes on the model itself?

Miles: I think it’s good. Honestly, there’s a lot of convergence right now on a pretty similar class of models, which are what I maybe describe as early reasoning models. It’s similar to, say, the GPT-2 days, when there were kind of initial signs of systems that could do some translation, some question and answering, some summarization, but they weren't super reliable.

We’re at a similar stage with reasoning models, where the paradigm hasn’t really been fully scaled up. But certainly, these models are much more capable than the models I mentioned, like GPT-2. But it’s notable that this is not necessarily the best possible reasoning models. It's just the first ones that kind of work.

And there are several models like R1, Alibaba’s QwQ. R1 is probably the best of the Chinese models that I’m aware of. And then there’s a bunch of similar ones in the West. So there’s o1. There’s also Claude 3.5 Sonnet, which seems to have some kind of training to do chain of thought-ish stuff but doesn’t seem to be as verbose in terms of its thinking process. And then there is a new Gemini experimental thinking model from Google, which is kind of doing something pretty similar in terms of chain of thought to the other reasoning models.

They’re all broadly similar in that they are starting to enable more complex tasks to be performed, that kind of require potentially breaking problems down into chunks and thinking things through carefully and kind of noticing mistakes and backtracking and so forth.

Jordan: When you read the R1 paper, what stuck out to you about it?

Miles: I mean, honestly, it wasn’t super surprising. There were some interesting things, like the distinction between R1 and R1.0 — which is a riff on AlphaZero — where it’s starting from scratch rather than starting by imitating humans first. But broadly speaking, I wasn’t surprised because I worked at OpenAI and was familiar with o1. I spent months arguing with people who thought there was something super fancy going on with o1. They were saying, “Oh, it must be Monte Carlo tree search, or some other favorite academic technique,” but people didn’t want to believe it was basically reinforcement learning—the model figuring out on its own how to think and chain its thoughts.

For some people that was surprising, and the natural inference was, “Okay, this must have been how OpenAI did it.” There’s no conclusive evidence of that, but the fact that DeepSeek was able to do this in a straightforward way — more or less pure RL — reinforces the idea. Just today I saw someone from Berkeley announce a replication showing it didn’t really matter which algorithm you used; it helped to start with a stronger base model, but there are multiple ways of getting this RL approach to work.

DeepSeek basically proved more definitively what OpenAI did, since they didn’t release a paper at the time, showing that this was possible in a straightforward way. It has interesting implications. It also speaks to the fact that we’re in a state similar to GPT-2, where you have a big new idea that’s relatively simple and just needs to be scaled up.

Squaring Distilled Models and The Stargate Announcement

Jordan Schneider: The piece that really has gotten the internet a tizzy is the contrast between the ability of you to distill R1 into some really small form factors, such that you can run them on a handful of Mac minis versus the split screen of Stargate and every hyperscaler talking about tens of billions of dollars in CapEx over the coming years.

Square the circle for me here, Miles. What does and doesn’t R1 tell you about to what extent compute is going to be necessary to reap the gains of AI in the coming years?

Miles Brundage: It’s a great question. And, you know, for those who don’t follow all of my tweets, I was just complaining about an op-ed earlier that was kind of saying DeepSeek demonstrated that export controls don’t matter, because they did this on a relatively small compute budget.

I think that’s the wrong conclusion. I think it certainly is the case that, you know, DeepSeek has been forced to be efficient because they don’t have access to the tools — many high-end chips — the way American companies do. And they’ve said this quite explicitly, that their primary bottleneck is U.S. export controls. As the CEO said in an interview you translated, “Money has never been the problem for us; bans on shipments of advanced chips are the problem.”

That doesn’t mean they wouldn’t prefer to have more. Turn the logic around and think, if it’s better to have fewer chips, then why don’t we just take away all the American companies’ chips? Clearly there’s a logical problem there.

Certainly there’s a lot you can do to squeeze more intelligence juice out of chips, and DeepSeek was forced through necessity to find some of those techniques maybe faster than American companies might have. But that doesn’t mean they wouldn’t benefit from having much more. That doesn’t mean they are able to immediately jump from o1 to o3 or o5 the way OpenAI was able to do, because they have a much larger fleet of chips.

People are reading too much into the fact that this is an early step of a new paradigm, rather than the end of the paradigm. These are the first reasoning models that work. This is the first demonstration of reinforcement learning in order to induce reasoning that works, but that doesn’t mean it’s the end of the road. I think everyone would much prefer to have more compute for training, running more experiments, sampling from a model more times, and doing kind of fancy ways of building agents that, you know, correct each other and debate things and vote on the right answer. So there are all sorts of ways of turning compute into better performance, and American companies are currently in a better position to do that because of their greater volume and quantity of chips.

Implications of DeepSeek’s Model Distillation

Jordan Schneider: Can you talk about the distillation in the paper and what it tells us about the future of inference versus compute? The premise that compute doesn’t matter suggests we can thank OpenAI and Meta for training these supercomputer models, and once anyone has the outputs, we can piggyback off them, create something that’s 95 percent as good but small enough to fit on an iPhone.

Miles: It’s unclear how successful that will be in the long term. Even if you can distill these models given access to the chain of thought, that doesn’t necessarily mean everything will be immediately stolen and distilled.

There are rumors circulating that the delay in Anthropic’s Claude 3.5 Opus model stems from their desire to distill it into smaller models first, converting that intelligence into a cheaper form. They apparently want to control the distillation process from the large model rather than letting others do it.

Companies will adapt even if this proves true, and having more compute will still put you in a stronger position. The implications for APIs are interesting though. If someone exposes a model capable of good reasoning, revealing these chains of thought might allow others to distill it down and use that capability more cheaply elsewhere.

Some companies have started embracing this trend. OpenAI provides a fine-tuning service, acknowledging the benefits of smaller models while keeping users on their platform rather than having them use their own model.

The space will continue evolving, but this doesn’t change the fundamental advantage of having more GPUs rather than fewer. Consider an unlikely extreme scenario: we’ve reached the absolute best possible reasoning model — R10/o10, a superintelligent model with hundreds of trillions of parameters. Even if that’s the smallest possible version while maintaining its intelligence — the already-distilled version — you’ll still want to use it in multiple real-world applications simultaneously.

You wouldn’t want to choose between using it for improving cyber capabilities, helping with homework, or solving cancer. You’d want to do all of these things. This requires running many copies in parallel, generating hundreds or thousands of attempts at solving difficult problems before selecting the best solution. Even in this extreme case of total distillation and parity, export controls remain critically important.

To make a human-AI analogy, consider Einstein or John von Neumann as the smartest possible person you could fit in a human brain. You would still want more of them. You’d want more copies. That’s basically what inference compute or test-time compute is — copying the smart thing. It’s better to have an hour of Einstein’s time than a minute, and I don’t see why that wouldn’t be true for AI.

No, You Don’t Just Give Up on Export Controls

Jordan Schneider: For the premise that export controls are useless in constraining China’s AI future to be true, no one would want to buy the chips anyway.

Miles: Exactly. People sometimes conflate policies having imperfect results or some negative side effects with being counterproductive. While export controls may have some negative side effects, the overall impact has been slowing China’s ability to scale up AI generally, as well as specific capabilities that originally motivated the policy around military use.

Jordan Schneider: What’s your fear about the wrong conclusion from R1 and its downstream effects from an American policy perspective?

Miles: My main concern is that DeepSeek becomes the ultimate narrative talking point against export controls. While I don’t think the argument holds, I understand why people might look at it and conclude that export controls are counterproductive. They are being efficient — you can’t deny that’s happening and was made more likely because of export controls.

However, the more extreme conclusion that we should reverse these policies or that export controls don’t make sense overall isn’t justified by that evidence, for the reasons we discussed. My concern is that companies like NVIDIA will use these narratives to justify relaxing some of these policies, potentially significantly. This might have some marginal positive impact on companies’ revenue in the short term, but it wouldn't align with the administration’s overall policy agenda regarding China and American leadership in AI.

If you’re DeepSeek and currently facing a compute crunch, developing new efficiency methods, you’re certainly going to want the option of having 100,000 or 200,000 H100s or GB200s or whatever NVIDIA chips you can get, plus the Huawei chips. Having access to both is strictly better. This is a straightforward case that people need to hear — it’s clearly in their benefit for these export controls to be relaxed. We shouldn’t be misled by the specific case of DeepSeek.

Those familiar with the DeepSeek case know they wouldn’t prefer to have 50 percent or 10 percent of their current chip allocation. Nobody wants fewer chips. Yes, you have to be more efficient when you have less, but everyone would prefer to have more, and relaxing our policies would only help them.

Jordan Schneider: A longer-term question might be: if model distillation proves real and fast following continues, would it be better to have a more explicit set of justifications for export controls?

Honestly, I always thought the Biden administration was somewhat disingenuous talking about “small yard, high fence” and defining it solely as military capabilities. That’s where the compute will go first, but if you’re talking about long-term strategic competition, much of how the Cold War was ultimately resolved came down to differential growth rates. When considering national power and AI’s impact, yes, there’s military applications like drone operations, but there’s also national productive capacity. From that perspective, you want a hundred von Neumanns rather than five to help with broader economic growth, not just hardening missile silos.

Miles: I agree about the somewhat disingenuous framing. The U.S. clearly benefits from having a stronger AI sector compared to China’s in various ways, including direct military applications but also economic growth, speed of innovation, and overall dynamism.

There are multiple reasons why the U.S. has an interest in slowing down Chinese AI development. However, to be clear, this doesn’t mean we shouldn’t have a policy vision that allows China to grow their economy and have beneficial uses of AI. We don’t necessarily need to choose between letting NVIDIA sell whatever they want and completely cutting off China. There should probably be something more nuanced with more fine-grained controls.

But that requires work to sort out and requires properly staffing the Department of Commerce to implement detailed agreements around keeping certain technologies for civilian purposes while preventing others from going to military uses [See Greg Allen, Emily Benson (now working at BIS), and Bill Reinsch on some very sensible and affordable reforms to upgrade BIS’ capabilities]. Without that capacity and without innovation in technical tooling, potentially including trackers on chips and similar measures, we’re forced into this all-or-nothing paradigm.

It’s unfortunate because this situation has numerous negative consequences. There are legitimate beneficial uses for AI in China, but we’re currently stuck between these extreme choices because we haven’t invested in those long-term fundamentals. Commerce can barely turn around rules in response to NVIDIA’s latest chips, let alone implement anything more sophisticated.

Open-Sourced Frontier Models

Jordan: What does it mean that this model got open-sourced?

Miles: It’s super interesting. DeepSeek is similar to Meta in being explicitly pro-open source — even more so than Meta. They’ve made an explicit long-term commitment to open source, while Meta has included some caveats.

I’m not sure how much we should believe that commitment. As AI systems become more capable, both DeepSeek employees and the Chinese government will likely start questioning this approach. I don’t actually believe it will continue, and I’m not convinced it’s in the world's long-term interest for everything to always be open-sourced.

Many things should be open-sourced, though. When things are open-sourced, legitimate questions arise about who’s making these models and what values are encoded in them. There are also potential concerns that haven’t been sufficiently investigated — like whether there might be backdoors in these models placed by governments. From a U.S. perspective, there are legitimate concerns about China dominating the open-source landscape, and I’m sure companies like Meta are actively discussing how this should affect their planning around open-sourcing other models.

The U.S. government needs to strike a delicate balance. It would be a mistake to lock in a policy of unconditional support for open source forever. The decision to release a highly capable 10-billion parameter model that could be valuable to military interests in China, North Korea, Russia, and elsewhere shouldn’t be left solely to someone like Mark Zuckerberg. The government needs to be involved in that decision-making process in a nuanced way. However, completely cutting off open source would also be a mistake. Both companies and the U.S. government should be considering how to respond to DeepSeek’s current leadership in this space.

Jordan Schneider: This is my base case too. The Trump administration just recently said they were going to revoke the AI executive order — the only thing remaining really was the notification requirement if you’re training a giant model.

This doesn’t seem sustainable as a steady state, either in China or in the West. These models are fine, cute, and fun now — they’re not really super dangerous. But we’re not far from a world where, until systems are hardened, someone could download something or spin up a cloud server somewhere and do real damage to someone’s life or critical infrastructure. That’s not a world government officials in Beijing or the West want to live in.

The eye of Sauron has now descended upon DeepSeek. Li Qiang, the Chinese premier, invited DeepSeek’s CEO to an annual meet-and-greet with the ten most notable Chinese people they select each year. The AI representative last year was Robin Li, so he’s now outranking CEOs of major listed technology companies in terms of who the central leadership decided to give shine to. That’s impressive, but it also means the Chinese government is really going to start paying attention to open-source AI.

In the past, there have been some industries where it was particularly helpful for Chinese industry to coalesce around open-source. With RISC-V, there’s no social stability risk of people using that instruction set architecture instead of ARM to design chips. Letting models run wild in everyone’s computers would be a really cool cyberpunk future, but this lack of ability to control what’s happening in society isn’t something Xi’s China is particularly excited about, especially as we enter a world where these models can really start to shape the world around us. Color me skeptical.

Miles: We haven’t done serious hardening yet. We don’t have CAPTCHA systems and digital identity systems that are AI-proof over the long term without leading to Orwellian outcomes. There are open vulnerabilities to AI systems running wild in the West. We’re also not well-prepared for future pandemics that could be caused by deliberate misuse of AI models to produce bioweapons, and there continue to be all sorts of cyber vulnerabilities.

We could eventually reach a point where we’ve built those defenses and feel more confident letting it rip, at least in the U.S. I don’t know whether China is ready for this kind of wild west situation of AIs running everywhere, being customized on devices, and fine-tuned to do things that might differ from the Party line. There might be a scenario where this open-source future benefits the West differentially, but no one really knows. The technology is still developing — it’s not in a steady state at all. I’m an open-source moderate because either extreme position doesn't make much sense.

Jordan: The Chinese regulatory architecture around bringing models to market has entirely focused on content moderation. You basically need to submit it and score well on questions about Tiananmen, Xinjiang, Xi, and the constitution. But you can do that and still be able to turn off the lights in a city. Once we live in that future, no government — any government — wants random people having that ability.

That world is probably a lot more likely and closer thanks to the innovations and investments we’ve seen over the past few months than it would have been a few years back. So, buckle up.

Miles: These reasoning models are reaching a point where they’re starting to be super useful for coding and other research-related purposes, so things are going to speed up. Despite some folks’ views, not only will progress continue, but these more dangerous, scary scenarios are much closer precisely because of these models creating a positive feedback loop.

I don’t think it’s going to happen overnight. For now, humans are in the driver’s seat of the research process, but these are extremely useful tools that DeepSeek, Meta, and others are using internally to improve their productivity. Even though a year feels like a long time — that’s many years in AI development terms — things are going to look quite different in terms of the capability landscape in both countries by then.

Export Controls Will Still Be a Key Piece of the Competition

Jordan: Closing thoughts, Miles?

Miles: No one believes the current export control system is perfect. Some people would prefer it to be stronger in some ways or weaker in others, but the main thing we should remember is that imperfect is not the same as counterproductive.

When people say “DeepSeek clearly shows X, Y, and Z,” they’re often pointing to examples of imperfections, like how we haven’t completely stopped Chinese AI progress, or how it led to more efficiency in specific contexts. That’s very different from saying it’s counterproductive.

Even if it had been counterproductive in the past, that doesn’t necessarily mean we’re stuck with the current policy. Perhaps they’ve invested more heavily in chips and their own chip production than they would have otherwise — I’m not sure about that. But they’re still behind, and export controls are still slowing them down.

Whether it’s the perfect policy or whether everything was done exactly right in the past is a separate question from whether we should maintain broadly similar direction with some course corrections versus reversing it entirely. My worry is that this will be taken as a sign that the whole direction is wrong, and I don't think there's any evidence of that.

Jordan: Got a song to take us out on?

Miles: How about something from Max Richter? Maybe something from The Leftovers, which I’d also like to plug as a good show. There’s this song called “The Departure” from the season one soundtrack of The Leftovers by Max Richter, which is very pleasant to listen to.

Jordan : Great. Perfect way to take us into our weekend. Miles, thanks so much for being a part of ChinaTalk.

Miles: Yeah, thanks so much for having me.

China AI Policy Updates

Trying a new thing this week giving you quick China AI policy updates led by Bitwise. Let us know if you like it!

DeepSeek CEO attends meeting with Premier Li Qiang

DeepSeek — the quiet giant leading China’s AI race — has been making headlines. Its latest r1 model, an open source model with comparable performance to o1 at a fraction of the cost, has turned the internet upside down. We’ll be covering the geopolitical implications of the model’s technical advances in the next few days.

Having flown under the radar domestically, policymakers in Beijing at the highest level have now officially taken notice. DeepSeek CEO Liang Wenfeng 梁文锋 attended a symposium hosted by Premier Li Qiang 李强 on January 20. This event is part of the deliberation and revision process for the 2025 Government Work Report, which will drop at Two Sessions in March.

Liang thus far has maintained an extremely low profile, with very few pictures of him publicly available online. One domestic reporter noted after seeing the state media video of the meeting, “The legendary figure in China’s AI industry is even younger in real life than expected. He takes great care of his skin, and at first glance, you might think he’s a student representative.”

At a similar symposium in January last year, Baidu’s Robin Li 李彦宏 was among the attendees. Not bad for Liang, beating out CEOs of China’s biggest tech companies.

Liang’s invitation should be interpreted as political recognition of DeepSeek’s critical place in China’s AI ecosystem. Attention like this is double-sided. Rising to the ranks of a “national champion” can open doors for both private and state-backed investment, as well as deliver government contracts (though past interviews indicate this probably isn’t what Liang is after…). However, it also may invite additional scrutiny and burdens. As we’ve covered in recent months, Chinese AI regulatory barriers are relatively low and narrowly focused on content moderation. As the Chinese political system starts to engage more directly, however, labs like DeepSeek may have to deal with headaches like government Golden Shares. It is also still an open question just how today’s regulators feel about closed- vs. open-source AI. In the coming weeks, we will be exploring relevant case studies of what happens to emerging tech industries once Beijing pays attention, as well as getting into the Chinese government’s history and current policies toward open-source development.

Unfortunately, the official readout of the meeting provides little detail on what was actually discussed and only notes that attendees “shared their insights on addressing current development challenges and improving this year’s government work.”

Apart from Liang Wenfeng, eight other experts attended, with robotics as the only other industry getting special attention:

• Zhang Hui 张辉: Dean of the Economics School at Peking University;

• Ren Shaobo 任少波: trained political economist and Party Secretary of Zhejiang University;

• Liu Jun 刘珺: Vice President and General Manager of the ICBC (Industrial and Commercial Bank of China);

• Wei Hongxing 魏洪兴: Chairman of AUBO Robotics 遨博智能 (a firm we covered in this piece last month);

• Chen Xuedong 陈学东: Chinese Academy of Engineering academician and expert in robotics;

• Chen Hongyan 陈红彦: Director of the National Center for Ancient Book Protection and specialist in cultural preservation;

• Du Bin 杜斌: Deputy Director of Beijing Union Medical College Hospital;

• Zou Jingyuan 邹敬园: Gymnast and Olympic medalist.

To catch up on China and robotics, check out our two-part series introducing the industry.

Policy support for the data labeling industry

The National Data Administration 国家数据局, a government entity established in 2023, has released “opinions” to foster the growth of the data labeling industry. The policy aims to harness China’s vast data resources and diverse application scenarios to drive this emerging sector forward.

Goals by 2027:

• Achieve an average annual growth rate of over 20%.

• Build a “relatively complete industrial ecosystem” for data annotation, including the development of influential, innovative enterprises and specialized annotation hubs.

The policy emphasizes advancing core technologies such as multimodal annotation, large model annotation, and quality evaluation. It also calls for the establishment of industry standards for data annotation, particularly in sectors like agriculture, manufacturing, healthcare, and smart cities.

Specific support measures include:

• Tax incentives: Implement policies such as R&D expense deductions and tax benefits for high-tech enterprises to reduce costs for data annotation businesses.

• Government procurement: Regions and departments are encouraged to allocate funds for purchasing data products and annotation services.

• Cost reduction: Promote the use of data vouchers 数据券, algorithm vouchers 算法券, and computing power vouchers 算力券 to lower operational costs for data annotation enterprises.

• Investment promotion: Encourage government funds to increase investments in the data annotation industry.

• Talent development: Cultivate and attract high-level professionals in data annotation through talent programs, revised national occupational standards. Encourage partnerships between enterprises, universities, and research institutions to promote training, continuing education, and certification of skills.

Additionally, the policy underscores the importance of AI safety in data annotation, with a focus on strengthening privacy protection, AI alignment, and security assessments.

Our take: High-quality data annotation is crucial for cutting-edge AI development. For example, Scale AI, a US-based firm specializing in this field — whose CEO, Alex Wang, we interviewed last year — recently raised $1bn at a $14bn valuation. Similar Chinese firms currently appear to be behind: Scale AI’s 2024 revenue was around 10x that of leading comparable Chinese firms like DataTang 数据堂 and Data Ocean 海天瑞声. It is unlikely that this new policy will do much to completely change dynamic, but the attention shows that the government recognizes the strategic importance of these firms and intends to continue helping them on their way.

State Council opinions on government guidance funds

In early January, the Chinese State Council released high-level “opinions” on improving government guidance funds, following discussions in December.

State-backed funds are now essential to China’s tech ecosystem. With foreign venture capital retreating and limited domestic private investment, local governments account for roughly 80% of all investments, making them the dominant limited partners (LPs). Understanding the challenges these funds face — and how the State plans to address them — is critical.

The “opinions” propose several measures to address the shortcomings of government-backed investment funds:

1. Encouraging collaboration to avoid duplication

   • National and local funds are urged to coordinate and focus on specialization, preventing redundant investments.

2. Stronger specialization of funds

   • Industrial investment funds: Support industrial chain modernization and critical sectors.

   • Venture investment funds: Drive early-stage, high-risk, and innovative projects.

3. Broadening exit strategies. The “Opinions” stress the need to expand exit pathways, including:

   • Private equity secondary markets (S-Funds): These funds specialize in purchasing stakes in private equity investments.

   • Mergers and acquisitions (M&A): Funds can exit by selling their stakes to strategic investors or companies looking to expand through acquisitions.

   • Listing on multi-tiered capital markets: Funds can sell their stakes through platforms like the National Equities Exchange and Quotations (NEEQ) (also called “New Third Board” 新三板) and regional equity markets.

4. Encouraging risk-taking and long-term investments

   • Focus on early-stage, high-risk projects, adopt “invest early, invest small, invest long-term” strategies, and extend fund durations to support projects requiring sustained development.

The above is essentially a list of the current shortcomings of government guidance funds:

• Duplication of efforts: Funds compete to support every high-tech industry in every city instead of fostering specialized clusters with agglomeration effects.

• Lack of specialization: Funds try to cover too many industries without building deep expertise.

• Limited exit strategies: Start-ups over-rely on IPOs. In Silicon Valley, only 5% of exits come from IPOs, while 95% are acquisitions. In China, the “better to be the head of a chicken than the tail of a phoenix” 宁当鸡头，不做凤尾 mindset discourages acquisitions, limiting exit options and ecosystem dynamism.

• Short-term mindset: Funds prioritize low-risk investments in established companies to ensure returns, rather than taking risks on transformative, high-impact technologies.

The “Opinions” correctly identify these issues, but the bigger question is: What can the State Council actually do to address them effectively?

New AI Standardization Committee under MIIT

The Ministry of Industry and Information Technology (MIIT) has established a new AI Standardization Technical Committee, numbered MIIT/TC1.

This committee’s responsibility spans five major areas.

1. Foundational and general standards:

   • Developing standards for AI terminology, evaluation and testing, reference architectures, and operations and maintenance.

2. Support infrastructure:

   • Creating standards for datasets, foundational hardware, and software platforms.

3. Algorithms and models:

   • Formulating standards for foundational large models and industry-specific large models.

4. Operations management:

   • Establishing guidelines for the application of large models, application maturity, and application development management.

5. Safety governance:

   • Developing standards to identify and prevent AI risks, ensure safety governance, address technological ethics, and safeguard data and information security.

The committee is comprised of 41 members, with the secretariat hosted by the China Academy of Information and Communications Technology (CAICT) — an MIIT-affiliated think tank. Professor Zheng Zhiming 郑志明 from Beihang University has been appointed as the Chairperson.

The other members include experts from major research institutions, universities, and companies, such as the three major telecom operators (China Mobile, China Telecom, and China Unicom), Baidu, Tencent, iFLYTEK, Huawei, Alibaba, SenseTime, and Unitree Robotics 宇树科技.

The big question on our mind now: How will this committee position itself vis-à-vis existing AI standard-setting bodies, such as the TC260 and SAC/TC28?

China leads the world in positive AI sentiment

According to a new Ipsos poll, China is the most optimistic about AI’s ability to create jobs out of the 33 countries surveyed, up there with Indonesia, Thailand, Turkey, Malaysia and India. The 77% of Chinese agreeing with the statement “AI will lead to many new jobs created in my country” contrasts pretty dramatically with America’s 36%. What this means for regulatory barriers for diffusion

Taiwan needs to scrap its Public Debt Act

Joseph Webster is a senior fellow at the Atlantic Council and edits the independent China-Russia Report. This article represents his personal opinions.

Taiwan’s debt levels are far too low. Taiwan’s low central government debt-to-GDP ratio, capped at 40.6% by the Public Debt Act, is abnormally low compared to other developed economies and limits its ability to address pressing security challenges. Given the security challenges facing the island, Taiwan must revoke the Public Debt Act and invest wisely in military kit and other whole-of-society resilience measures.

In 2023, Taiwan’s debt-to-GDP ratio stood at 29.1 percent, the sixth lowest of the 41 economies in the International Monetary Fund’s “advanced” classification. Moreover, Taiwan’s public debt has fallen significantly since peaking in 2012. While central government frugality is usually highly commendable, this policy is wildly inappropriate for Taiwan, given its unique conditions.

Taiwan’s perilous security environment demands greater investments. The CCP has repeatedly declared its intent to subjugate Taiwan, by force if necessary, and is building the military capabilities to do so. With the world’s largest navy and a vast dual-use civilian fleet, the PRC is escalating coercive measures, including large-scale military exercises, blockades, and potential kinetic actions, demonstrating both intent and growing capability.

Given these fraught security conditions, it is astonishing — infuriating to some — that Taiwan continues to underinvest in defense. Taiwan’s defense outlays stand at 2.5 percent of GDP, above the 2 percent baseline for NATO members, but also far below its needs. SIPRI estimates PRC military expenditures totaled $309 billion in 2023, more than 17 times the ROC’s outlays. While Taiwan should not be expected to approach total PRC military spending or conventional capabilities, it can procure “a large number of small things” and make itself indigestible via a porcupine strategy based on asymmetric capabilities.

Subscribe now

And Taiwan’s holistic security needs extend beyond just military affairs. Taiwan, which faces a real risk of a quarantine or blockade, is more than 95 percent reliant on seaborne energy imports. Taiwan is already the world’s largest per-capita coal consumer in the electricity sector, but its power needs are only increasing due to the demands of data centers (including those for AI), electric transportation, and more. Mitigating Taiwan’s serious and growing energy security challenges will require substantial investment in indigenous nuclear energy, offshore and onshore wind, and next-generation solid-state batteries, which could play a major role in a cross-Strait contingency. US LNG could enhance Taiwan’s energy security, limit urban air pollution, and reduce bilateral trade deficits — all of which are increasingly important again in Washington, DC.

Taiwan’s Public Debt Act hampers essential security investments, particularly in military readiness. With rising risks from Beijing and an increasingly complex relationship with Washington, Taipei should repeal the act to prioritize critical security spending.

What role have patents played across history in building national power, and how in the 21st century have the US, China, and EU played their cards differently with respect to IP?

To discuss, ChinaTalk interviewed Adam Mossoff, professor at the Antonin Scalia Law School at GMU.

Listen on Apple Podcasts, Spotify, or check out the YouTube video.

We explore…

• How the patent system has shaped American society since independence,

• The extent to which patent policy caused the great divergence between China and the west,

• Whether Elon’s misunderstanding of patents will become the dominant attitude of the second Trump administration,

• The Patent Trial and Appeal Board (PTAB) and other threats to the U.S. innovation ecosystem,

• How to reconcile China’s IP theft with robust domestic patent law,

• What the U.S. can do to facilitate innovation while competing with China in emerging technology.

Thanks to The Innovation Alliance for sponsoring this episode. The Innovation Alliance is a coalition of research and development-based technology companies representing innovators, patent owners, and stakeholders who believe in the critical importance of maintaining a strong patent system that supports innovative enterprises of all sizes.

The Right to Innovate

Jordan Schneider: Let’s go back to 1787. Why was patent policy so uncontroversial that it was written directly into the Constitution without any debate during the Constitutional Convention in Philadelphia?

Adam Mossoff: Many people don’t realize patents are in the Constitution. It grants inventors, creators, and artists the power to secure exclusive rights. While it doesn’t use the terms “patent” or “copyright,” the basis for such rights is Article 1, Section 8 of the Constitution. This marked the first time in human history that protections for intellectual property appeared in a country’s founding document.

Several core reasons explain this inclusion. The Founders recognized that creators and innovators morally had a right to the fruits of their productive labors. Following John Locke’s notion that we have the right to mix our labor with things in the world to create property, they applied this principle to innovators and creators. George Washington was what we’d now call an angel investor. He invested in steamboat inventors in Virginia.

Jordan Schneider: But Adam, what were his returns like?

Adam Mossoff: Not very good, from what I understand.

Noah Webster, of Webster’s First American Dictionary fame, traveled to Philadelphia to explain to the Constitutional Convention how he struggled to sell copies of his dictionary due to varying copyright protections across different states. He needed national protection for his rights since he had a book to sell nationwide.

Robert Fulton, famous for his steamboats, demonstrated some of his early prototypes to the delegates as well. They took a break from the convention and went to the Delaware River to watch Fulton’s prototype in action.

The delegates were intimately aware of these issues and understood their importance, especially for a new country lacking an industrial base. The nation was primarily agrarian at the time, and the Founders recognized that patent rights would be key to growing America’s innovation economy.

Jordan Schneider: Many funny things come out of that origin story. This is downstream, of course, of people being upset with England, the first modern country to set up a version of this patent system. The royal prerogative was baked into the whole British ecosystem, with patents being granted at the whim of the Crown and assigned to one person — and the Crown never had to pay any royalties. It was much less comprehensive and more arbitrary, though better than whatever they were doing in the Frankish kingdom.

Webster was an early patent lobbyist — congratulations, we have a rich legacy here. The inventors showing up to do tech demos in Philadelphia is like Sam Altman going to D.C. the month before GPT-4 comes out to tell the U.S. government how much AI will matter to America.” Some things never change.

Now we have this policy where the Constitution says intellectual property is actually property and should be something you can monetize. How did that play out in the Republic’s first decades?

Adam Mossoff: Before I answer your question, I’d like to note something about the inventors at the Constitutional Convention. The Founders were really sensitive to and aware of the abuse of government monopolies by the Crown and England. They recognized that these patents weren’t monopolies.

The copyright and patent clause, which authorizes Congress to secure exclusive rights to inventors and creators for a limited time, is the only place in the Constitution before the Bill of Rights where you’ll find the word “right” used. When they use the term “exclusive right,” that means property right. President Washington, in his first address to Congress, called for enacting copyright and patent statutes. Congress did this with the Patent Act of 1790 and the Copyright Act of 1790. These were among the first pieces of legislation Congress enacted, because they were recognized as absolutely fundamental to economic growth and property rights more broadly.

They secured these not as royal grants or gifts from the Crown, but as property rights. This meant inventors could enter the market, make transactions, and commercialize their work — just as people commercialize property rights now by selling houses, computers, or renting rooms. Inventors immediately began engaging in what we now call licensing.

With licensing, inventors could invent without manufacturing. That enabled them to divide up the labor and specialize.

Patent owners invented what we now call the franchise business model — an intellectual property licensing business model. While we associate franchises with fast food like McDonald’s or Wendy’s, it’s actually an IP licensing model where the owner licenses others to manufacture, run, and sell products. Samuel Morse did this with the telegraph and Morse code.

Charles Goodyear did this when he invented vulcanized rubber — he has no affiliation with the Goodyear Tire and Rubber Company, which was created 40 years after his death and precipitated a trademark lawsuit with his family over his name. Goodyear was just an eccentric inventor. He wrote a two-volume book on all the cool things you could do with rubber — you can make boats out of it, you can make shoes out of it. Then to prove his point, he had copies of this book bound in rubber. He didn’t want to be a manufacturer, he wanted to license other people to make these products.

This quickly distributed these innovations into the marketplace, incorporating a lot more people in the process.

Patent licensing democratized these inventions, leading to an explosion of new commercial products and services from the United States in its first 50-60 years.

Jordan Schneider: Let’s underline that for a second. The British model required the Crown to decide, and if you got the patent, you had to be the one to make and sell the product. The problem is there aren’t that many people like Elon Musk — we’ll get to him later. Not many engineering geniuses or tinkerers, county clerks and such, also have that rapacious capitalism gene to build factories, empires, and sales teams.

In the late 18th and early 19th century, the American model allowed you to get rich and spread your invention across the country and economy more broadly without needing to be blessed with both engineering genius and business genius.

Adam Mossoff: Perfectly stated. Abraham Lincoln famously identified the U.S. patent system as one of the three great achievements in human history, the first being language and the second being the written Constitution. He said the patent system “added the fuel of interest to the fire of genius.” Lincoln knew of which he spoke — he’s actually the only U.S. President to have received a patent, which was for an invention he created before his presidency, in 1848.

Jordan Schneider: What was Lincoln’s patent?

Adam Mossoff: It was a method for lifting boats over sandbars and other obstructions in the Mississippi River. It’s unclear whether it was actually deployed or if he made any money from it, but Lincoln also worked as a lawyer representing other patent holders in Illinois. He represented Cyrus McCormick, who invented the mechanized reaper — the first true labor-saving device in human history. It dramatically increased the efficiency of food production and proved Malthus wrong. Food production went through the roof after the invention of the mechanized reaper. Lincoln represented McCormick in many lawsuits against infringers of his reaper patent. He knew the patent system very well.

Jordan Schneider: Coming out of paternity leave, all I did was read Civil War history. There’s this incredible moment — example #574 of why Lincoln was an incredible person. He got onto this McCormick case, which was going to be the biggest case of his career. The case gets transferred out from Illinois and then McCormick decided to bring in some big guns from the East Coast who had actually gone to law school. Edwin Stanton shows up, then the hottest lawyer on the planet. He takes one look at Lincoln, who’s gangly and wearing tattered clothes, with his messy 400-page brief, calls him a “damned long armed Ape” and literally doesn’t acknowledge throughout the entire case.

Lincoln’s takeaway from that incredibly rude and disrespectful experience wasn’t resentment, but rather, “Oh my God, I now know what it’s like to play in the big leagues. I need to go back and up my game. I’m thankful for this opportunity to have been the equivalent of an unused pinch hitter on a World Series baseball team.” The fact that he later appointed Stanton as his Secretary of War 15 years later shows his incredible magnanimity. So yes, in my book Lincoln gets to be on Mount Rushmore even though he didn’t make money from his patent.

Adam Mossoff: Those people’s ROI is that we have an amazing country that survives incredible ups and downs.

Jordan Schneider: He gets credit for all the GDP growth post-1865.

Adam Mossoff: Yeah, I mean, he saved the Union. Kind of a small thing.

Jordan Schneider: Let’s come back to Lincoln’s idea that the patent system added the fuel of interest to the fire of genius.

We did an epic four-hour, three-part show with Yasheng Huang a few months ago. Professor Huang argues that one of the key mistakes Chinese civilization made was that the exam system, starting in the 700s, stripped out all the economic returns and societal esteem from doing inventions. Conversely, during the Industrial Revolution, tinkerers in the UK and the U.S. became rich and famous for doing science — which hadn’t happened for basically all of human history before then.

Adam Mossoff: It’s a fascinating phenomenon. China was incredibly inventive — they invented paper money, gunpowder, one of the first modern massive fleet ships, incredible ship technologies, the abacus, and many other things. But then invention shifted to the West, to Europe, and eventually to the United States.

Today we distinguish between invention and innovation because humans are naturally inventive. People invent for all sorts of reasons — prestige, personal interest, joy — but inventing something in your lab or garage is not the same as innovation that’s actually used by consumers, mass-produced, deployed in the marketplace, and sold through stores or over the internet.

That was the key feature in the United States in recognizing patents as property rights, which serves both a direct financial interest and a democratization function. Anyone can invent and get this property right, just as anyone can be a farmer, worker, or podcaster if they choose that career. Property rights serve as the basis for entering contracts and deals to commercialize, engage with manufacturers, and create profit while consumers get the product in the marketplace.

Jordan Schneider: There’s this American national stereotype of a nation of pioneers, self-willed, doing your own thing. The UK stereotype is the British Academy of Sciences, which is more centralized. Yes, in the early years of the Industrial Revolution, you had this very distributed system with local inventors, but pretty quickly you start to have a more aggressive caste system and hierarchy in science. What’s your take on institutions versus national character when it comes to the American story of engineering-based businesses in U.S. History?

Adam Mossoff: It’s a great question. There’s a tendency to be reductionist about things and say it’s all one thing — nature or nurture — but why can’t it be both? It’s more of a symbiotic relationship between strong intellectual and cultural norms about individualism in our society, about individuals pulling themselves up by their bootstraps, as they said in the 19th century.

Many examples illustrate this. Charles Goodyear, Cyrus McCormick, Samuel Morse was a professor of art at NYU when he invented the electromagnetic telegraph. Samuel Colt whittled his first wooden version of the Colt Revolver as a shipmate. Dr. Leo Baekeland was a well-established, famous chemist in Europe who immigrated to the United States to continue his chemistry experiments. He became famous as the inventor of synthetic plastic, patented around 1906, and trademarked as Bakelite.