Nvidia stops shipping H200 AI chips to China — ending the licensed channel Huang built
Interim CEO Colette Kress told investors Nvidia is no longer shipping licensed Hopper chips to China, closing the arrangement Jensen Huang brokered with the White House — and a new US rule now reaches the overseas data centers that served Chinese customers from abroad.
- The halt: “We are no longer shipping licensed Hopper data center products to China,” interim CEO Colette Kress said on the November 17 earnings call. No reason given, no resumption date.
- The scale: $2.4 billion in licensed Hopper sales to China in the quarter ended October 25 — about 2% of $115.8 billion in revenue. Fourth-quarter guidance of $130 billion assumes no China data center compute revenue.
- The channel: H200 exports were approved case by case after Washington eased its review in January 2026; chips shipped through the US with a 25% tariff, in an arrangement brokered by Huang and President Trump.
- The back door: a November 16 Commerce Department rule bars advanced US chips from foreign data centers serving Chinese customers remotely. Jefferies puts the revenue at risk at $44–62 billion next year; Kress says demand backlog means no top-line impact.
- The context: roughly 600,000 H200 chips were reported held in Hong Kong as Beijing withheld import permission, and China had largely refused delivery even before the halt.

Nvidia has stopped shipping its licensed Hopper AI chips to China, interim chief executive Colette Kress said on the company's November 17 earnings call — closing the one channel through which the most powerful American AI hardware was still legally reaching Chinese buyers. The halt lands less than a week after Jensen Huang resigned as chief executive, and on the same day Nvidia created a board-level committee to oversee national security policy, export controls and global trade risks.
The statement was blunt, and Kress offered no reason for it. “Our guidance does not have any licensed Hopper data center products included, and we are no longer shipping licensed Hopper data center products to China,” she said. She did not say whether the halt is permanent, whether customers had cancelled orders, or when — if ever — shipments might resume.
A $2.4 billion quarter, then nothing#
The timing is what makes the halt striking. In the quarter that ended October 25, Nvidia sold $2.4 billion worth of licensed Hopper data center products to China — about 2% of its $115.8 billion in quarterly revenue. The fourth-quarter outlook, by contrast, assumes no data center compute revenue from China at all. One quarter the channel was open and generating billions; the next, it is simply gone from the forecast.
The chips at issue are from Nvidia's Hopper family, including the H200, the accelerator Chinese labs use to train and run advanced AI models. They are manufactured for Nvidia in Taiwan by TSMC. The H200 was never a grey-market product: Washington approved its sale to Chinese buyers under a licensing arrangement, and the chips were shipped through the United States so the government could collect a 25% tariff.

The channel Huang built — and Beijing never embraced#
The arrangement was Jensen Huang's work, brokered with President Donald Trump after Washington eased its review approach in January 2026, letting H200 export applications be considered case by case under security conditions instead of facing a presumption of denial. Huang had argued for years that competing in China mattered more than ceding the market — in April he told podcaster Dwarkesh Patel that conceding the market would be “a disservice to our country.” China hawks in Washington saw it differently: a loophole that handed Chinese labs computing power the export bans were meant to deny.
Beijing, for its part, never embraced the channel either. Even before Nvidia stopped shipping, China had largely refused to take delivery of the chips, preferring to build demand for its domestic chip industry even though major AI compute users still want the superior Nvidia hardware, according to Jacob Feldgoise, a researcher at Georgetown's Center for Security and Emerging Technology. Reuters reported on November 5 that roughly 600,000 H200 chips were sitting in Hong Kong while Beijing withheld permission for large-scale mainland imports. The halt, in other words, formalizes a freeze that was already half in place.
The rule that closed the back door#
The halt arrived the day after a second, wider move. On November 16, the US Commerce Department's Bureau of Industry and Security issued a rule barring advanced American chips from foreign data centers that serve Chinese customers remotely — the workaround that had let Chinese labs rent compute in Singapore, Malaysia and elsewhere instead of importing the hardware.
The evidence for that route is now public: all of China's leading AI models show signs of having been trained on Nvidia Grace Blackwell chips that are prohibited from being exported to China, with the training likely taking place at data centers outside the country. Under the new rule, large data centers that want to keep buying Nvidia chips will have to certify they are not serving any Chinese customers, export control attorney Dan Fisher-Owens told TechCrunch — and if they cannot, they may have to look to other chip suppliers, eating further into Nvidia's market share.
The financial stakes are contested. Investment bank Jefferies has forecast that the new rules could trim between $44 billion and $62 billion from Nvidia's data center revenue next year. Kress pushed back on the call, saying the backlog of demand for chips such as Vera Rubin means the rules would have no impact on top-line revenue. Critics of Nvidia's China dependence make a broader claim: as much as a fifth of the company's revenue traces to demand linked to the country.

What it costs — and what to watch#
For Nvidia itself, the direct cost of the halt looks small. The $2.4 billion in licensed Hopper sales was roughly 2% of quarterly revenue, and the company's order book for its newest hardware is full. The larger cost is strategic: the end of the licensed channel removes Nvidia's last legal foothold in the world's second-largest AI market, and the remote-access rule now forces a choice on the data centers in Southeast Asia that had become the industry's workaround.
What to watch: whether Kress or the board ever explains the halt or names a date for resumption; what happens to the roughly 600,000 H200 chips reported to be held in Hong Kong; how strictly the new remote-access rule is enforced — and whether it pushes Chinese labs further toward domestic chips, which is exactly what Beijing has been encouraging. The board's new national-security committee will be the body answering those questions, and its first test is already here.
Sources: TechCrunch, “Nvidia is abandoning China after Jensen Huang's ouster”; BusinessWorld / Reuters, “Nvidia stops shipping US-approved AI chips to China”; Dave Finances, “Nvidia halts licensed Hopper chip shipments to China”.