Nvidia has informed its customers in China of its intention to begin shipping the first H200 chip, the company’s second-most powerful AI accelerator, before New Year’s Eve. The information was reported by Reuters, citing three sources familiar with the company’s plans.
According to reports, initial supplies are expected to be covered using existing stocks. Expected shipments range between 5,000 and 10,000 modules, for an estimated total of 40,000 to 80,000 H200 chips destined for the Chinese market.
The same sources report that Nvidia has also communicated to customers the possibility of expanding production capacity. Orders for new production capacity are expected to open in the second quarter of 2026, although the timing remains subject to further evaluation.
The situation, however, remains uncertain. Currently, Chinese authorities have not yet approved any requests to purchase H200s , and the delivery schedule could be subject to change based on government decisions. One of the sources consulted clarified that the entire plan depends on official authorization, without which operational certainty remains unclear.
Neither Nvidia nor the Chinese Ministry of Industry and Information Technology have commented on the matter.
If the shipments are confirmed, this would be the first shipment of H200 chips to China. The sales were authorized this year by US President Donald Trump, subject to the payment of a 25% commission. Reuters also reported last week that the Trump administration has launched an interagency review to examine license applications for H200 exports to Beijing.
The decision represents a departure from previous policies. Both the Trump and Biden administrations had imposed restrictions on the sale of advanced artificial intelligence chips to China, citing national security concerns.
The H200 belongs to the Hopper family, a previous generation of Nvidia accelerators. While it has been joined by the newer Blackwell chips, it remains a widely used solution for AI workloads. Currently, the company’s production is primarily focused on the Blackwell line and the upcoming Rubin series, which has contributed to the H200’s lack of availability on the market.
The US administration’s decision comes at a time when China is investing significantly in developing a domestic supply chain for AI chips. Locally produced solutions have not yet achieved the performance of the H200, and any authorization for imports could raise concerns about a potential slowdown in domestic technological progress.
According to Reuters, Chinese officials convened an extraordinary meeting earlier this month to discuss the issue and assess whether to approve shipments. Among the options under consideration is the possibility of tying each purchase of H200 chips to the use of a certain quota of semiconductors produced in China.
For Chinese technology groups like Alibaba Group and ByteDance, which are interested in the H200, a potential completion of supplies would represent a significant leap in computing power. The processing power of imported H200s would be approximately six times greater than that of the downgraded H20 chips currently available on the domestic market.
Follow us on Google News to receive daily updates on cybersecurity. Contact us if you would like to report news, insights or content for publication.
