President Trump approved Nvidia H200 AI chip exports to ten Chinese technology firms during the US-China summit on May 15, with Alibaba, Tencent, and ByteDance among the approved buyers. Jensen Huang joined the US delegation on a late invitation from Trump. The approval comes with three conditions: no military use, strict security standards, and a requirement that Nvidia certify sufficient US inventory before any China sale ships. The unusual fourth condition: the US Treasury will receive 25% of the revenue generated from approved China sales.
25% revenue-share to the federal government on private chip sales is the number to focus on. It is unprecedented for commercial semiconductor transactions. The most analogous structure is the Cold War-era foreign military sales arrangements where the Defense Department took a cut on US-export weapons platforms. Those were government-brokered defense deals; this is a commercial sale by a publicly-traded company. Applied to Nvidia, a 25% revenue cut translates to roughly 10-12% of operating margin assumed by the buyer if absorbed through pricing, depending on the specific transaction structure. Wall Street is reading this as Nvidia bearing the cost. Chinese buyers passing it through to end-customers would be the alternative reading. Whether that holds in practice depends on whether the structure is tax-like (Nvidia pays before recognizing revenue) or fee-like (Nvidia collects and remits). The legal mechanism has not been disclosed. Chinese buyers are reportedly hesitant to finalize purchases despite the approval, citing domestic political pressure to favor Huawei silicon. The reciprocal risk is real: a Chinese government action against the approved buyers could leave the deals in limbo.
For Nvidia shareholders: the 25% structure is the number to model. For Chinese chip-procurement teams: the approval window is open but expect domestic friction. For other US chip exporters: assume this becomes the template, not the exception.