The performance gap between US and Chinese AI models may be widening due to second-order effects of chip controls. By limiting inference at scale, the controls reduce the volume of customer interactions and feedback Chinese firms receive. This starves them of the data needed to identify and patch model weaknesses on diverse, real-world tasks.
Related Insights
Evaluating export controls by asking if China is still advancing is the wrong metric. The true test is the counterfactual: where would China be *without* the restrictions? The controls act as a significant handicap in a competitive race, not a complete stop, and it's highly likely China would be ahead of the U.S. in AI without them.
Unable to compete globally on inference-as-a-service due to US chip sanctions, China has pivoted to releasing top-tier open-source models. This serves as a powerful soft power play, appealing to other nations and building a technological sphere of influence independent of the US.
The most dangerous policy mistake would be reverting to a 'sliding scale' that allows China to buy chips that are a few generations behind the cutting edge. In the current era of AI, performance is aggregatable. China could simply purchase massive quantities of these slightly older chips to achieve compute power equivalent to frontier systems.
A nation's advantage is its "intelligent capital stock": its total GPU compute power multiplied by the quality of its AI models. This explains the US restricting GPU sales to China, which counters by excelling in open-source models to close the gap.
Contrary to the narrative of a close race, Huawei's AI chips are falling further behind NVIDIA's. The performance gap is projected to widen from a 5x difference to a 17x difference within two years. Shockingly, Huawei's next-generation chip is actually projected to be less powerful than its current leading model, indicating significant production struggles.
Contrary to their intent, U.S. export controls on AI chips have backfired. Instead of crippling China's AI development, the restrictions provided the necessary incentive for China to aggressively invest in and accelerate its own semiconductor industry, potentially eroding the U.S.'s long-term competitive advantage.
The US ban on selling Nvidia's most advanced AI chips to China backfired. It forced China to accelerate its domestic chip industry, with companies like Huawei now producing competitive alternatives, ultimately reducing China's reliance on American technology.
China's superior ability to rapidly build energy infrastructure and data centers means it could have outpaced US firms in building massive AI training facilities. Export controls are the primary reason Chinese hyperscalers haven't matched the massive capital spending of their US counterparts.
China is compensating for its deficit in cutting-edge semiconductors by pursuing an asymmetric strategy. It focuses on massive 'superclusters' of less advanced domestic chips and creating hyper-efficient, open-source AI models. This approach prioritizes widespread, low-cost adoption over chasing the absolute peak of performance like the US.
The US-China AI race is a 'game of inches.' While America leads in conceptual breakthroughs, China excels at rapid implementation and scaling. This dynamic reduces any American advantage to a matter of months, requiring constant, fast-paced innovation to maintain leadership.
