Recent reports highlight just how quickly China has closed the gap with the U.S. on AI. Bloomberg recently reported that China is rapidly closing the performance gap while already winning on cost and global adoption. Stanford University found earlier this year that China has “nearly erased” America’s AI lead. Forbes and BCG now describe the competition as a duopoly: two increasingly incompatible tech stacks, with the window for flexibility closing fast.
Alvin Graylin, a well-respected technology executive in Asia, points the finger for these developments right at export controls. In conversations with the founders and CEOs of China’s leading GPU companies, he has learned that many of them would have gone under without American export controls. Hardly anyone was buying Chinese GPUs before the ban. Indeed, the ban is the only reason several of these companies are still around. A look at key Chinese AI companies underscores that fact. Biren Technology “filed” a 22-fold revenue surge, directly attributable to the domestic market the U.S. chip ban created. Cambricon reported its first-ever full-year profit, with revenue up 450 percent.
The most telling admission came from Huawei’s rotating chairman. Xu Zhijun who explicitly thanked the U.S. government for its export restrictions, saying they “supercharged” China’s semiconductor research and development. Cut off from American suppliers, Huawei built its own chip architecture under pressure. Then it launched a Model-as-a-Service platform across nine countries. The restrictions were designed on the assumption that Huawei couldn’t build its own stack. Huawei is now selling that stack across the globe.
Huawei’s recent export pitch to Egypt suggests Xu wasn’t exaggerating. Huawei proposed a plan for Egypt to build AI data centers from the ground up, including 1,408 Ascend 950-series chips for a training cloud, 600 more for inference clusters, on a 12-month delivery timeline. Winning the contract would have made Huawei a key AI provider to Africa’s second-largest economy and established a foothold in the Middle East, the same region where the U.S. has aggressively invested in the more lucrative markets of the UAE and Saudi Arabia. It would also mark the first confirmed export of Huawei’s Ascend accelerators after more than a year of attempts, proving there is real foreign demand for Chinese chips.
It’s important to understand that once developers build on Huawei’s models and companies migrate operations to its cloud infrastructure, they don’t switch back. BCG warns the window for companies to operate across both tech stacks is rapidly closing. Countries are being forced to choose. Every month that window stays open, the cost of pivoting back to American providers grows higher.
Washington is still tied to the idea that the country with the best model is winning the AI race, but that fixation ignores the power of “good enough but cheaper” models for many customers. And China seems to be closing the quality gap as well. More importantly, it has grasped a fundamental truth: models come and go, but infrastructure endures—and the country whose technology powers the world will set the terms of the AI age. To do so, Huawei is pushing beyond its domestic market with lower cost models. Chinese AI and semiconductor companies’ revenues are surging with China’s own tech executives crediting U.S. restrictions with accelerating their drive for self-reliance. The simple reality is that export controls have not stopped China’s advance but accelerated it. Four years is long enough to admit that the controls are not working — and allow American firms to compete globally. The current system doesn’t.
