Chinese regulators have given the green light for the first imports of Nvidia's H200 AI accelerators, approving the shipment of several hundred thousand chips, according to people familiar with the matter. The clearance was issued while Nvidia Chief Executive Officer Jensen Huang was in China this week.
Sources said the initial allotment was largely directed to three large Chinese internet companies, while additional firms are queued for subsequent batches. Those familiar with the decision declined to name the recipients because of the sensitivity surrounding the approvals.
This move represents a notable relaxation following months of tight review of advanced AI hardware entering China. The approvals underscore the country's demand for high-end computing capacity to support data centers and accelerate AI development, even as authorities continue efforts to foster domestic alternatives.
Regulators granted authorization for the chips during the Nvidia CEO's visit, a timing detail cited by people who spoke on condition of anonymity. The reported clearance covers several hundred thousand H200 units, and allocations were concentrated among a small number of major internet firms in China. Other companies remain on waiting lists for later shipments.
Observers note the sensitivity of the approvals process was reflected in the decision by those familiar with the matter not to disclose who received the initial allocations. The episode highlights the tension Chinese policymakers face - balancing the immediate need for cutting-edge AI computing power with long-term ambitions to build domestic semiconductor and AI capabilities.
While the authorization is limited to initial shipments and concentrated recipients, it indicates regulators are prepared to permit some imports of advanced AI hardware to meet near-term infrastructure and development requirements. The longer-term stance on such imports and the pace at which approvals may be extended to a broader set of companies were not specified by the sources.
Sectors affected: data centers, cloud services, AI development, and semiconductor suppliers supporting AI hardware.