China’s carefully controlled H200 opening puts AI needs ahead of a full import reversal

Small batches of Nvidia’s H200 artificial-intelligence chips have reportedly entered mainland China in recent weeks, despite Beijing’s broader effort to limit reliance on foreign semiconductors. The development appears to be a tightly managed compromise: China’s leading AI companies gain access to powerful computing hardware, while the government continues to promote domestic chipmakers.

Who received the chips?

ByteDance and Tencent have each received about 10,000 H200 processors, according to two people familiar with the matter cited by the Financial Times. A few other Chinese technology groups could soon receive approval for shipments of a similar size, although the report did not identify those companies.

The quantities are substantial enough to support additional AI computing capacity, but they remain far below the level that might suggest a broad reopening of the Chinese market to Nvidia hardware.

Why were the shipments allowed?

The reported deliveries appear to be limited, government-approved exceptions rather than a general relaxation of import controls. The United States has cleared the companies to purchase as many as 100,000 H200 chips each, but Beijing is reportedly allowing only small batches into mainland China.

That approach reflects the tension at the heart of China’s AI policy. Chinese firms need access to advanced computing power to compete with U.S. rivals, yet Beijing also wants them to develop and use domestic alternatives rather than build a permanent infrastructure dependency on Nvidia.

Why does Hong Kong matter?

Chinese authorities have reportedly instructed companies to keep most of the H200 chips outside mainland China. Hong Kong is central to that arrangement because it operates outside the mainland’s customs border, allowing companies to hold or deploy the processors there without incorporating the bulk of the supply into the mainland market.

The arrangement is not without practical limits. Hong Kong reportedly lacks enough data-centre capacity and power infrastructure to support the full-scale deployment of a very large H200 inventory. That means the location offers regulatory flexibility, but not necessarily a complete solution for running all the chips at scale.

What the small shipments reveal

The deliveries highlight a two-sided technology squeeze. Washington has allowed only limited exports of Nvidia’s advanced AI hardware, while Beijing is rationing access even where U.S. approval exists. China is therefore trying to keep its AI companies competitive without weakening its push for semiconductor self-reliance.

Against that backdrop, the roughly 10,000 processors reportedly received by ByteDance and Tencent look less like a policy reversal than a controlled release: enough access to address immediate AI demand, but not enough to displace the strategic priority given to domestic chips.

How certain is the information?

A senior U.S. Commerce Department official said in July that “very few” H200 chips had so far been shipped to China and Hong Kong, describing the quantity as very small. Nvidia had not immediately commented at the time of the Reuters report.

Reuters explicitly presented the story as a report by the Financial Times based on sources familiar with the matter and said it could not independently verify the details. The reported shipment volumes, any additional recipients and the precise conditions attached to the chips should therefore be treated as unconfirmed information rather than an officially documented policy announcement.