PALO ALTO | The most important technology deals are no longer only negotiated between companies. They are negotiated through security reviews, export-control conditions, diplomatic timing and industrial-policy pressure.
Reuters reported that the United States has approved sales of Nvidia’s H200 artificial-intelligence chips to around 10 Chinese companies, including major firms such as Alibaba, Tencent, ByteDance and JD.com. But the report also said no actual shipments have occurred, leaving a major technology deal in limbo.
That gap between clearance and delivery is the story. The H200 is powerful enough to matter for advanced AI workloads, and Nvidia has a commercial interest in serving Chinese customers. But Washington has national-security concerns about cutting-edge chips, while Beijing wants to reduce dependence on U.S. technology and strengthen domestic alternatives.
Reuters reported that Nvidia CEO Jensen Huang joined a high-level U.S. delegation visiting China as he sought a breakthrough. The timing shows how chip sales have become part of broader U.S.-China diplomacy rather than ordinary corporate export activity.
The policy environment is unusual. Export controls are meant to prevent advanced technology from strengthening rival military or surveillance capabilities. At the same time, U.S. companies argue that legal sales help preserve American influence, revenue and standards in global technology markets. If Chinese customers turn entirely to domestic chips, U.S. firms lose both business and leverage.
China’s hesitation is also strategic. Buying U.S. chips can speed development, but it can also create dependence on a supplier that may be restricted by Washington at any time. Domestic chip investment is slower and harder, but it offers political control.
For AI developers, the uncertainty affects planning. Data centers, model training, cloud contracts and software roadmaps depend on knowing which chips will be available, when they will arrive and under what rules. A clearance that does not lead to deliveries does not solve a capacity problem.
For U.S. policymakers, the H200 case is a test of whether export controls can be precise. Too loose, and Washington risks weakening its own security goals. Too tight, and U.S. firms may lose markets while Chinese alternatives accelerate.
The H200 question will not be the last one. As AI models become more capable and chips become more specialized, export policy is likely to become increasingly granular. The future may be less about broad yes-or-no bans and more about conditional approvals, monitored use and deal-by-deal negotiation.
Additional Reporting By: Reuters; CGN Tech Blog analysis