Nathan Goes to China #3: US-China Relations, the Art of the AI Deal & the Road to Pax Robotica

What It Covers

Nathan Labenz delivers the third installment of his China series, arguing that the US goal toward China should shift from "winning" to achieving a "Pax Robotica" — mutual AI-era prosperity. He assesses China's AI capabilities, articulates the Chinese perspective on US export controls, and proposes concrete bilateral deal structures involving chips, solar, batteries, and robotics.

Key Questions Answered

• •Reframe the goal state: US China policy discourse skips defining a win condition. Labenz argues the target should be "Pax Robotica" — shared prosperity and AI benefits for both nations — rather than dominance. Without an explicit goal, policy defaults to zero-sum competition, which raises the probability of an AI arms race analogous to nuclear proliferation, where the sword of Damocles remains permanently overhead with no civilian dividend.
• •China's AI rise is reversion to the mean, not anomaly: China appeared on the SuperGLUE benchmark leaderboard at number one as of 2022, before ChatGPT launched. Chinese five-year plans included AI as a strategic priority in 2016 — the same year Google declared itself AI-first. Roughly 12 Chinese companies have shipped frontier or near-frontier LLMs, matching the 11 US companies identified. Treating Chinese AI strength as sudden fast-following misreads a decade of systematic investment.
• •Export controls carry a hidden cost — paradigm divergence: The most underappreciated risk of chip export controls is not that China catches up anyway, but that US and Chinese AI tech stacks diverge enough that neither side can verify the other's capabilities. This mutual opacity is precisely how AI competition converts into cold or hot war. Maintaining a shared AI paradigm — where both sides can read each other's research and models — functions as a structural de-escalation mechanism.
• •The "century of humiliation" framework drives Chinese chip policy reaction: Chinese leadership interprets US semiconductor export controls through the lens of their 1839–1949 period of foreign domination, when technological inferiority enabled invasion and forced concessions. Proposals like Anthropic's "coalition of democracies" strategy — building decisive AI advantage to make China an offer it cannot refuse — read to Chinese officials as a direct threat of regime change, hardening resolve rather than producing compliance.
• •Proposed deal structure — chips for clean energy and robotics: Labenz outlines a concrete bilateral trade: the US offers advanced chips and potentially chip-making know-how in exchange for Chinese expertise in solar manufacturing, battery technology, and robotics — sectors where China now leads. This recouples the economies without creating unacceptable dependencies on either side, and reduces the incentive for China to accelerate domestic chip development purely for strategic self-sufficiency.
• •China's open-source AI posture is a credible signal, not just rhetoric: China has published AI research in English, released open-weights models with no usage restrictions, and established the World AI Cooperation Organization. The Tsinghua University AI Safety Hub lists Constellation, LISA, and SASH as explicit inspirations and funds international student exchanges. DeepSeek's January 2025 RL-zero paper — demonstrating chain-of-thought metacognition without supervised fine-tuning — was released openly. These actions are harder to reverse than statements, making them more credible commitments.
• •US structural advantages require active maintenance, not passive assumption: The US holds three durable advantages China cannot replicate: immigrant assimilation (language accessibility, cultural openness), English as the global research lingua franca, and venture capital startup culture. However, restricting Chinese student visas and talent pipelines directly erodes the first two. Labenz recommends expanding high-skill immigration, clarifying that chip export rules do not restrict AI safety research collaboration, and building state capacity in manufacturing rather than relying solely on exclusion.