Can Nvidia sell AI chips to China? Yes. But under tight conditions

Shares of Nvidia (NVDA) rose slightly after a report said the company plans to ship advanced AI chips to China before the Lunar New Year in mid-February.

At first glance, the news looks like a routine update on deliveries. In practice, it reflects how tightly controlled and uncertain the global AI chip market has become.

According to the report, Nvidia expects to deliver its H200 processor to Chinese customers in the coming months. The H200 is Nvidia’s second-most powerful AI chip and is mainly used to train large artificial intelligence models.

Sources familiar with the matter said the company aims to ship between 5,000 and 10,000 chip modules.

This would equal roughly 40,000 to 80,000 individual chips.

That volume is meaningful, but it is far from enough to reshape the balance of global supply. Instead, it points to a narrow and carefully managed opening rather than a broad return to normal trade between the United States and China.

The shipments follow approval by the U.S. government allowing Nvidia to sell the H200 to China, subject to a 25 percent surcharge. Similar permission would apply to other American chipmakers such as Intel and AMD.

Even so, people familiar with the plan stressed that nothing is final. Chinese authorities have not yet approved any purchases, and the schedule could change depending on government decisions.

One proposal being discussed would require each H200 purchase to be bundled with a certain amount of domestically produced chips.

That condition would allow Chinese firms to access foreign technology while also supporting local chipmakers, even if those domestic alternatives are less powerful.

The uncertainty highlights a central feature of today’s AI chip market. Commercial deals exist only as long as they align with political goals on both sides, and those goals can shift quickly.

The H200 itself helps explain why the issue is so sensitive.

The chip is designed by Nvidia and manufactured by Taiwan Semiconductor Manufacturing Company using its advanced 4-nanometer process.

It is estimated to deliver around six times the performance of Nvidia’s H20 chip, which was created to comply with earlier export restrictions.

That performance gap matters because it directly affects how fast and efficiently large AI models can be trained.

For Chinese technology companies such as Alibaba and ByteDance, access to even a limited number of H200 chips could shorten development timelines and reduce costs, even if supply remains constrained.

At the same time, the fact that Nvidia describes the H200 as its second-most powerful chip underscores the limits of the current policy environment. The company’s most advanced hardware remains restricted, and every product sold into China is shaped by regulatory boundaries.

The report also said Nvidia plans to expand production capacity for the H200 in the new year, with orders for additional capacity expected to open in the second quarter of 2026. That suggests Nvidia sees sustained global demand for the chip, regardless of how China policy evolves.

However, greater production does not guarantee broader access. Any increase in shipments to China would still depend on export rules, approvals, and ongoing negotiations between governments. That makes long-term planning difficult for both Nvidia and its customers.

China, meanwhile, is pushing to expand its own advanced chip production.

Emergency meetings reported earlier this month show Beijing weighing whether to allow the shipments and under what conditions. Requiring foreign chips to be paired with domestic alternatives would help China maintain momentum in AI development while reducing long-term reliance on Western suppliers.

For Nvidia and other Western chipmakers, this creates a difficult balance. The Chinese market remains large and important, but access increasingly comes with political and economic trade-offs.

Taken together, the planned shipments do not signal a return to normal trade in advanced AI hardware. They reflect a controlled flow of technology, closely watched and easily reversed.

Nvidia (NVDA) has a Disruption Score of 4.

Click here to learn how we calculate the Disruption Score.