Nvidia is abandoning China after Jensen Huang’s ouster

Glass Epoch DeskOriginal digestSource: TechCrunch AI

AI-assisted digest, checked against the original source before publishing. How we work

Nvidia's interim CEO says the chipmaker is done selling H200s to China, as new export rules reshape where its accelerators can go.
Nvidia is abandoning China after Jensen Huang’s ouster
Photo: Unsplash · free stock license

Nvidia will stop selling its H200 chips in China, acting CEO Colette Kress said Tuesday, closing out the company's role in a contested workaround to the U.S. ban on advanced chip sales to the country. The disclosure came on the first earnings call since co-founder and CEO Jensen Huang stepped down last week, and the same day Nvidia said it would create a board-level committee to oversee national security policies, export control compliance, and global trade risks.

The China H200 arrangement had been brokered by Huang and President Donald Trump, and required chips to transit the United States so the government could collect a 25% tariff. The pathway drew objections from China hawks who wanted to limit Chinese labs' computing power and preserve the U.S. AI lead, while Huang argued the market was worth defending: in April he told podcaster Dwarkesh Patel that ceding it "to concede that market for the United States technology industry is a disservice to our country."

Huang's departure — which came in part over an affair with the CEO of a company under investigation for smuggling chips into China — coincided with new rules issued Monday that bar advanced U.S. chips from foreign data centers serving Chinese customers remotely. Investment bank Jefferies estimated those controls could trim $44 billion to $62 billion from Nvidia's data center revenue next year, though Kress said demand backlog for chips like Vera Rubin would leave top-line revenue unaffected.

Meanwhile, Chinese frontier labs have found routes to prohibited hardware: all of China's leading AI models show evidence of training on Nvidia Grace Blackwell chips that cannot be exported to China, likely at data centers outside the country. Sales to those facilities fall under Monday's rules, and large training-scale data centers buying Nvidia chips will have to certify they serve no Chinese customers, export control attorney Dan Fisher-Owens told TechCrunch. Absent that certification, buyers could be pushed toward alternative chips.

What to watch: whether the certification requirement actually reshapes the buyer pool for hyperscale data centers, how much of Nvidia's Southeast Asia pipeline proves China-linked, and whether Kress's claim that Vera Rubin backlog neutralizes the revenue hit holds up as orders clear. Also open is whether Chinese demand shifts decisively to domestic accelerators or continues routing through third countries.

Key facts

  • • Acting CEO Colette Kress announced the H200 China exit on Nvidia's first call since Huang's departure.
  • • The H200 China pathway routed chips through the U.S. to collect a 25% tariff.
  • • New Monday rules bar advanced U.S. chips from foreign data centers serving Chinese customers remotely.
  • • Jefferies projected a $44–62 billion hit to data center revenue next year.
  • • Nvidia's Singapore subsidiary booked $104 billion in revenue in 2025 before disclosure ended.

Why it matters

Builders selling into the AI supply chain should treat export-control compliance as a first-class product constraint: certification duties now attach to data center customers, not just chip vendors. Expect accelerated demand signaling for non-Nvidia accelerators in Southeast Asia and China, and more volatility in forward guidance as backlog claims are tested against actual orders.

Spotted an error? Send a correction.

Get the day's AI news in one email

The Daily Epoch · 60-second read · unsubscribe anytime

Related