China is permitting small batches of Nvidia's H200 chips to enter the mainland to assist domestic AI companies in keeping pace with the US. This move indicates a flexible approach to acquiring high-end chips, easing the pressure from US export controls.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
China is allowing small, batch-by-batch imports of Nvidia's H200 chips to ease the computing power anxiety of its domestic AI companies. This strategy is both a pragmatic response to U.S. export controls and a reflection of China's deep-seated shortcomings in high-end computing power.
According to the Financial Times, China is allowing small batches of Nvidia H200 chips into the mainland, with ByteDance and Tencent each receiving about 10,000 processors, and more companies may follow soon. This policy adjustment aims to help domestic AI companies keep pace with the U.S. in the competition.
Notably, the U.S. allows each company to purchase up to 100,000 H200 chips, but China's release scale is far below this cap, reflecting Beijing's cautious approach. Additionally, purchases require approval from China's National Development and Reform Commission (NDRC), indicating that this is not a full opening but a managed admission.
The H200 chip is at least two generations behind Nvidia's most powerful chips, which cannot be sold to China due to U.S. export controls. This means that even if Chinese AI companies obtain H200 chips, they will struggle to compete with the top U.S. level in hardware performance.
China hopes to support domestic chip manufacturers like Huawei, but whether Huawei and others can provide sufficient high-end computing power in the short term remains unknown. Although Hong Kong is also allowed to receive H200 chips, the region lacks data centers and power infrastructure, making it difficult to become an effective computing power hub.
At the model technology level, Chinese labs are catching up quickly. Moonshot launched K3, Alibaba released Qwen3.8, DeepSeek introduced DeepSeek V4, and Z.ai recently released GLM-5.3. These developments show that China is competitive in algorithms and model design.
However, in terms of available inference computing power, the gap between China and U.S. suppliers remains enormous. After a surge in demand, Moonshot had to turn away some customers, highlighting the tightness of computing power supply. Nvidia currently has about 500,000 H200 chips in stock, but how many China can obtain is still subject to policy restrictions.
This article is primarily based on a report from the Financial Times, as relayed by THE DECODER. Specific figures (such as 10,000 each, 100,000 cap, 500,000 inventory) come from that report and have not been independently verified; they should be regarded as source claims rather than confirmed facts.
China's approval of small-batch H200 imports is a pragmatic choice under U.S. export controls, aimed at alleviating the computing power bottleneck for domestic AI companies. However, this measure cannot fundamentally change the generational gap in high-end computing power between China and the U.S., and China's long-term competitiveness in AI still depends on breakthroughs in its domestic chip industry.
Primary report
Primary source