Nvidia H200 Licenses Approved but Shipments Blocked by China

The U.S. granted export licences for Nvidia H200 chips to roughly 10 Chinese companies in December 2025, including Alibaba, Tencent, ByteDance and JD.com, with Lenovo and Foxconn authorised as distributors and allowances of up to 75,000 units per firm. No H200 units have been delivered because Chinese authorities bar their deployment inside the country and have told firms to confine any use of the chips to operations overseas.

The licences include a U.S. condition that H200 chips sold to Chinese buyers be used only within China. Chinese regulators issued an opposing rule, directing companies to limit the use of foreign GPUs to foreign operations and to prioritise domestically produced processors such as Huawei’s Ascend line. The two sets of rules are mutually incompatible and have prevented shipments from moving forward.

At a recent Senate hearing, Commerce Secretary Howard Lutnick testified that Chinese companies are concentrating investment on local suppliers and that China’s State Council has launched a supply-chain security review aimed at reducing dependence on U.S. semiconductors. During a presidential visit to Beijing, the U.S. leader brought Nvidia’s CEO Jensen Huang on the trip and said on departure that “something could happen” on chip exports. U.S. Trade Representative Jamieson Greer stated that semiconductor controls were not on the bilateral agenda during the talks.

Several Chinese technology firms have shifted product and development plans toward domestic chips. DeepSeek adapted its April release, DeepSeek V4, to run on Huawei’s Ascend processors for both training and inference. Executives at Tencent and Alibaba have indicated that Chinese GPU supply will rise through 2026, and Alibaba has said its T-Head proprietary GPUs have reached scaled mass production.

The impasse has measurable commercial effects. Nvidia’s revenue from China has declined to about 5% of total sales in recent quarters, down from more than 20% before tighter export controls. Nvidia’s guidance for the current quarter assumes zero revenue from China. U.S. regulators have authorised exports and named distributors, but Beijing’s deployment ban has kept deliveries on hold.

The current status is an approved-export, no-delivery situation: licences are in place, allowances assigned, and the hardware remains undelivered while Chinese policy requires that foreign GPUs not be used for domestic workloads. The pause has coincided with increased development and production activity around domestic AI hardware in China.