Report says China may allow top AI firms to buy a limited number of NVIDIA H200 chips

NVIDIA shares are in focus after a report said China is planning to allow some of its best-known artificial intelligence companies to buy limited quantities of NVIDIA Corporation’s H200 chips, a high-end data-center processor used for training and running large AI models.

The claim comes from a chain of reporting: Reuters referenced a separate industry outlet’s reporting, which cited two people familiar with the matter. In that account, China’s approach would not be a broad reopening of access, but instead a selective allowance tied to how many chips are available and who is permitted to purchase them.

The H200 is Nvidia’s flagship data-center AI accelerator line, designed to deliver high throughput for large-scale model workloads. In practice, access to those devices is closely watched because the AI chip supply needed for training and deployment has become a competitive bottleneck, especially for enterprises that must build or scale capabilities quickly.

If China moves toward a limited, prioritized sales channel, it would likely shape near-term purchase timing and the mix of customer orders Nvidia receives. Such arrangements can also influence how quickly cloud providers and large AI labs can ramp compute capacity for new projects, depending on whether allocations are enough to meet demand.

For Nvidia, export-policy constraints have been a recurring theme for the broader data-center business. When regulators restrict which buyers can receive which models, companies often see uneven demand across geographies and product tiers, with some customers constrained even when overall AI spending remains strong.

Sector-wide, the possibility of selective access in China reflects how governments and regulators increasingly treat advanced AI hardware as both an industrial capability and a strategic resource. Rather than a single global “on-off” decision, the industry is now operating in a more granular environment where permissions, quotas, and allowable configurations can matter as much as headline chip availability.

What remains unclear is the scope of any prospective plan. The reporting does not provide details here on the size of the allocations, which specific “top AI firms” would qualify, how long any restriction would last, or whether the channel would include other Nvidia models in the same product family.

Investors and industry watchers will be looking next for indicates around whether any policy becomes operational, including changes Nvidia discloses about customer shipments, product-specific demand patterns, or regulatory commentary that clarifies how controlled access would work.