The Bureau of Industry and Security of the US Department of Commerce has officially announced that the export review of Nvidia's H200 chips to China will be changed from "presumed denial" to case-by-case review. This has resolved the long-standing and unresolved controversy over the export of H200 artificial intelligence chips to China.
 

On January 13th local time, the Bureau of Industry and Security of the US Department of Commerce officially adjusted its export control policies towards China, announcing that the review of export of Nvidia's H200 chips to China would be changed from "presumed denial" to case-by-case review. This also clarified the long-standing dispute over the export of H200 artificial intelligence chips to China. 
 

In December 2025, US President Trump was the first to disclose through social media his intention to allow the sale of H200 chips to China, while also proposing an additional condition of charging a 25% transaction fee. This move broke away from the conventional framework of international trade rules and immediately sparked internal divisions within the US government: the hard-line faction towards China was concerned that this would weaken the US's technological advantage in high-end chips, while the other camp believed that through technology transfer, they could curb the independent R&D process of China's domestic chip industry. After a month of multi-departmental consultations and negotiations, the final policy framework determined by the BIS focused on establishing a "approval control + economic regulation" dual mechanism. That is, exports would need to undergo joint compliance reviews by multiple departments such as the State Council and the Ministry of Defense, and the Chinese domestic purchasing entities would need to provide commitments for non-military use and complete safety management procedures. At the same time, it required NVIDIA to implement a strict customer due diligence mechanism, while the US would achieve the regulation of relevant market revenues through a 25% transaction commission. 
 

As the flagship product of Nvidia's Hopper architecture, the H200 chip's performance positioning determines its short-term supplementary value for China's artificial intelligence industry. The chip's FP8 precision computing power reaches 4 petaflops, and it can support single-card inference for models with 72 billion parameters. Although its performance is only one-fifth of Nvidia's latest Blackwell architecture B200 chip, it is still a key supplementary option for the current available high-end computing resources. According to industry compliance information disclosure, Nvidia plans to deliver the first batch of 5,000-10,000 sets of 8-card modules in mid-February 2026 (before the Lunar New Year), and set transaction conditions such as full prepayment and non-modifiable orders. This inventory arrangement indirectly confirms the rigid demand for high-end computing resources in the Chinese market. It is important to note that the export approval has clear quantitative control: the export volume shall not exceed 50% of its sales volume in the US domestic market, and the products must pass performance compliance verification by a third-party authoritative laboratory. 

From a short-term perspective, the compliant introduction of the H200 chip will effectively alleviate the shortage of high-end computing power supply in the domestic market, providing computing power support for the iterative development of large models and the large-scale construction of intelligent computing centers. This will further drive the release of demand in upstream and downstream industrial chain links such as AI servers, liquid cooling systems, 800G/1.6T optical modules, and HBM storage. Domestic leading cloud service enterprises and artificial intelligence research institutions will obtain a stable supply channel of high-end computing power, reduce reliance on informal supply chains, and effectively avoid supply chain compliance risks. From a long-term perspective, the 25% special transaction fees will significantly increase procurement costs. This cost pressure may be released through two paths: reducing the profit margin of enterprises or passing it on to downstream service pricing. This will form a substantive cost constraint. A more concealed risk lies in the fact that such practical computing power supplements may weaken the R&D drive of some enterprises and delay the process of local high-end chips being replaced. This logic is consistent with the market binding effect of the H20 chip supply to China before. 
 

As of now, the Chinese authorities have not issued a specific response to this policy adjustment. However, market attention to issues such as import approval standards and compliance boundaries for application scenarios continues to rise. This adjustment in the H200 chip export policy to China is essentially a specific manifestation of the new stage of the technological competition between China and the United States - the US attempts to achieve dual goals of economic gains and technology control through the export of non-core high-end technologies; while the Chinese industry is firmly promoting the process of core technology self-reliance and controllability while reasonably utilizing external resources.