Nvidia Plans H200 Shipments to China, Existing Inventory to Be Used

Nvidia has informed Chinese customers that it intends to begin delivering H200 AI accelerator modules to approved buyers in China by mid February 2026, company communications and people familiar with the discussions show. The initial shipments would come from existing inventory and would be limited in scale while Nvidia prepares to expand production capacity later next year.

Sources described the first tranche as roughly 5,000 to 10,000 modules. Because each module contains multiple H200 chips, that quantity translates to about 40,000 to 80,000 individual H200 processors. Reporting on the numbers has sometimes presented the figure as individual chips, but the underlying shipments are organized and contracted at the module level, industry executives said.

Nvidia plans to open orders for new H200 production capacity in the second quarter of 2026. That planned capacity ramp is intended to follow the initial inventory shipments and would allow wider fulfillment if regulatory and commercial conditions permit. For now, the company is positioning its on hand stock as a stop gap to meet approved demand before a fuller manufacturing and distribution schedule is established.

The timing follows a recent U S policy decision permitting certain advanced chip sales to China subject to a 25 percent fee, a condition announced by the administration in December. That policy move and a concurrent government review of export rules have created a narrow window for companies and regulators to reconcile national security concerns with commercial demand for high performance processors. At the same time Chinese authorities had not approved the reported H200 purchases as of today, and officials in Beijing were said to be considering possible conditions on approvals, including requirements to bundle purchases with domestically produced chips.

The accounts of Nvidia’s plans come from unnamed people familiar with the matter who declined to be identified because the discussions were private. Nvidia and China’s Ministry of Industry and Information Technology did not immediately respond to requests for comment.

Financial markets reacted to the developments, with Nvidia shares ticked higher in pre market trading as investors weighed the prospect of resumed sales into China, a major growth market for high performance compute. For Chinese technology firms and cloud operators, access to H200 class hardware would represent a significant boost in available compute power for training and running large language models and other advanced applications. Analysts caution however that the effective cost of imports could rise if the 25 percent fee is applied in practice, and that any Chinese requirements to bundle orders with local chips could alter the economics and logistics of deployment.

Key milestones to watch in coming months include formal approvals from Chinese regulators, the precise terms under which U S export conditions will be implemented, and the opening of orders for new H200 production capacity in the second quarter of 2026. Those outcomes will determine whether the initial shipments described by sources materialize and how rapidly full scale sales might resume.