A senior Commerce official told Congress this week that a “trivial” number of Nvidia H200 chips have been shipped to China under newly approved export licenses, the first deliveries of the H200 to Chinese buyers. The shipments are the clearest sign yet that Washington has chosen to sell into the Chinese market rather than wall it off, the opposite of what a bipartisan bill tried to require last year.

Jeffrey Kessler, the under secretary of commerce for industry and security, said the volume so far was small despite about $10 billion in approved licenses. Washington has cleared H200 sales to roughly ten Chinese companies, including Alibaba, Tencent, ByteDance, and JD.com, with each approved buyer allowed to purchase up to 75,000 chips, according to reporting on the approvals. Demand dwarfs that: Chinese firms are reported to have ordered more than two million H200s for 2026, against total Nvidia inventory of roughly 700,000.

The policy runs directly counter to the GAIN AI Act, a bipartisan measure introduced last year by Senators Jim Banks and Elizabeth Warren, among others, that would have required chipmakers to certify American customers had priority access to advanced AI chips before exporting comparable chips to countries of concern, China chief among them. It passed the Senate as an amendment to the annual defense bill but was left out of the House version and stripped in conference late last year; it never became law. Nvidia and the Semiconductor Industry Association opposed it; Microsoft and some AI-safety groups backed it.

The administration’s position, argued most publicly by White House AI adviser David Sacks, is that restricting sales does more harm than good, because it pushes Chinese buyers toward Huawei and helps domestic Chinese chipmakers scale. Sacks has framed export decisions around whether they help “the American tech stack dominate the world,” and warned that “excessive bureaucratic delays are a gift for Huawei.” The H200 approvals put that logic into practice: keep Chinese AI firms buying American silicon rather than building around it.

The result is a policy caught between two goals it cannot fully reconcile, denying China frontier compute while keeping China dependent on American suppliers. The H200 is a capable but prior-generation part, and the volume caps and Nvidia’s own supply shortage mean the approved sales are a fraction of what Chinese firms want, which is partly the point. But the direction is now set against the instinct that animated the GAIN AI Act, and against Washington’s own earlier flip-flops on chip controls: for now, the US is choosing to sell to China’s AI champions rather than make them wait behind American buyers, even as Huawei races to make the sale unnecessary.

Sources: CNBC, Bloomberg, Nextgov/FCW

–
By the Control Plane Editorial Team