The debate over whether the United States can maintain its lead in artificial intelligence has entered a sharper phase. According to the Financial Times, China is approaching the development of its own frontier AI capabilities — a prospect that demands a recalibration of American technology policy and export control strategy.
The argument, advanced in the FT's analysis, is straightforward in its logic: if China is on the cusp of building AI systems comparable to the most advanced Western models, the window for the US to leverage its technological edge is narrowing. The editorial thesis here is not whether competition exists — that much is settled — but whether the tools Washington currently wields are adequate for the contest ahead.
Export Controls and Their Limits
The US has relied heavily on semiconductor export restrictions as its primary lever for slowing China's AI progress. The Biden-era controls on advanced chips, tightened in successive rounds, were designed to deny Chinese labs access to the compute necessary for training large-scale models. The logic was temporal: every month of delayed access to cutting-edge hardware was a month of preserved American advantage.
But the effectiveness of this approach has always been contested. China's domestic chip industry, while still trailing TSMC and Nvidia in raw performance, has shown a capacity for workarounds — from stockpiling chips before restrictions took effect to developing alternative architectures that extract more from less. DeepSeek's emergence earlier this year as a competitive open-weight model, reportedly trained with fewer high-end chips than Western counterparts, underscored a discomforting possibility: that compute constraints may slow but not stop Chinese AI development. If the FT's framing is correct and China is nearing its own frontier-class systems, the question shifts from delay to deterrence — and the policy toolkit looks thinner.
Beyond Hardware: The Broader Contest
The AI competition between the US and China extends well beyond chips. It encompasses talent pipelines, data ecosystems, regulatory frameworks, and the willingness of governments to fund basic research at scale. China has advantages in several of these domains. Its vast domestic data pools, relatively permissive approach to data collection, and centralized capacity to direct industrial policy all serve as accelerants. The US, meanwhile, benefits from deeper capital markets, a more mature startup ecosystem, and the gravitational pull of Silicon Valley on global talent — though immigration policy remains a persistent friction point.
What makes this moment distinct is the convergence of capability and intent. China's AI ambitions are not new, but the gap between aspiration and execution appears to be closing faster than many Western analysts anticipated even a year ago. The policy implication is that export controls, while necessary, function as a single instrument in what needs to be a multi-layered strategy. Investments in domestic semiconductor manufacturing — through programs like the CHIPS Act — represent one additional layer. Coordinated allied frameworks with Japan, the Netherlands, and South Korea on technology transfer represent another. But none of these measures operate in isolation, and each carries trade-offs: tighter controls risk accelerating China's push for self-sufficiency, while looser ones risk direct capability transfer.
The framing of this contest as one that "America must prepare for" carries an implicit acknowledgment that preparation is incomplete. Whether the US approach evolves toward a more integrated strategy — one that pairs restriction with acceleration of domestic capability — or remains anchored primarily in denial-based export policy may determine the shape of AI competition for the rest of the decade. As both nations continue to pour resources into frontier AI development, the question of whether technological advantage can be sustained through policy alone remains very much unresolved.
With reporting from Financial Times — Technology
Source · Financial Times — Technology
