The US greenlights Nvidia H200 chip sales to ten Chinese firms as the CEO targets a market breakthrough.
In a significant policy shift, the US Department of Commerce has approved export licenses allowing Nvidia to sell its H200 Tensor Core GPUs to ten Chinese companies. This decision marks a notable easing of AI chip export restrictions that have been in place since October 2024, and comes as Nvidia CEO Jensen Huang seeks to regain access to what was once the company's second-largest market.
The Policy Shift
The approved licenses cover sales of Nvidia's H200, the company's most advanced AI accelerator available for commercial deployment. Unlike the heavily restricted H100 and A100 chips, the H200 falls into a category that US officials have deemed suitable for controlled export under specific compliance frameworks.
According to sources familiar with the matter, the ten approved companies include a mix of cloud service providers, autonomous vehicle developers, and AI research institutions. All recipients have undergone enhanced due diligence processes and agreed to strict end-use monitoring requirements.
What Makes the H200 Different
The H200 represents Nvidia's latest iteration of data center AI acceleration technology, featuring:
- 141GB of HBM3e memory, up from 80GB in the H100
- Enhanced FP8 Transformer Engine for large language model training
- 4.8 TB/s memory bandwidth
- Optimized for inference workloads with significantly improved cost-per-token economics
The chip's architecture makes it particularly well-suited for deploying large foundation models at scale, which aligns with the use cases of the approved Chinese companies.
Nvidia's Strategic Calculus
CEO Jensen Huang has been vocal about the importance of the Chinese market to Nvidia's long-term growth trajectory. In recent earnings calls, he emphasized that AI development is a global phenomenon and that overly restrictive export policies could ultimately push Chinese companies to develop entirely separate technology stacks.
"When you cut off access to leading technology, you don't slow down innovation—you just redirect it," Huang stated in a March 2026 interview. "The question for policymakers is whether it's better to maintain influence through controlled engagement or lose all visibility through complete isolation."
Nvidia has also worked closely with US regulators to design compliance frameworks that address national security concerns while preserving commercial opportunities. These frameworks include enhanced supply chain transparency, restricted resale provisions, and ongoing audit rights.
Market Reaction and Implications
Financial markets responded positively to the announcement, with Nvidia shares rising 3.7% in after-hours trading. Analysts view the approval as a signal that US-China technology policy may be entering a more pragmatic phase, balancing security concerns with economic interests.
For the approved Chinese companies, access to H200 hardware represents a significant competitive advantage. While domestic Chinese AI chips have made progress, they still lag behind Nvidia's offerings in raw performance and software ecosystem maturity. The ability to deploy H200-based infrastructure could accelerate AI development timelines by 12-18 months compared to relying solely on domestic alternatives.
Geopolitical Context
This development arrives amid broader discussions about the future of AI governance and technology transfer. Some US lawmakers have expressed concern that even limited exports could inadvertently support capabilities with dual-use potential. However, the Biden administration has indicated that complete technological decoupling would be neither feasible nor strategically advantageous.
The Commerce Department emphasized that the approved licenses include "unprecedented monitoring and verification mechanisms" designed to ensure chips are used only for declared purposes. Companies found in violation of end-use restrictions face severe penalties, including permanent export prohibition and potential criminal charges.
What's Next
The initial allocation of H200 chips to the ten approved companies is expected to begin in Q2 2026, with delivery timelines extending through the end of the year. Nvidia has indicated that additional license applications from other Chinese entities are under review, though approval is far from guaranteed.
Industry observers will be closely watching how this policy evolves. If the initial deployment proceeds without security incidents, it could pave the way for broader access. Conversely, any violations or concerns could trigger immediate policy reversals.
For now, the ten approved companies represent a carefully calibrated experiment in managed technology transfer—one that attempts to balance commercial interests, technological leadership, and national security considerations in an increasingly complex geopolitical landscape.