OTWopentechwire
Tech Intelligence, Openly Wired
Policy

Nvidia Exits China Market as Acting CEO Reverses Huang-Era Chip Deal

Colette Kress announces end to H200 sales amid tightened export controls and new board oversight of national security compliance

WZ
Wei Zhang
China Tech Correspondent · Hangzhou
Nov 20, 2026
6 min read
Nvidia Exits China Market as Acting CEO Reverses Huang-Era Chip Deal
Credit: Josh Edelson / Getty Images

A Sharp Policy Turn

Nvidia will stop selling H200 chips to China, acting CEO Colette Kress announced on Tuesday, closing off a trade pathway that had survived two years of Washington debate over technology competition with Beijing. The decision marks the company's withdrawal from what critics had termed a loophole in US restrictions on advanced semiconductor exports, a pathway that former CEO Jensen Huang had defended as essential to American competitiveness.

The disclosure arrived during Nvidia's first quarterly earnings call since Huang stepped down last week. On the same day, the company announced the formation of a board-level committee tasked with overseeing national security policies, export control compliance, and risks tied to global trade. The timing underscores how quickly the chipmaker is recalibrating its approach to geopolitics after a leadership change that followed revelations about Huang's personal conduct and questions over the company's exposure to smuggling investigations.

The H200 arrangement had been brokered directly between Huang and President Donald Trump. Under its terms, the chips travelled through US territory, allowing the government to collect a 25 per cent tariff on each shipment. The structure satisfied the White House's revenue goals but frustrated officials who wanted to deny Chinese laboratories access to the computing power needed for frontier artificial intelligence research. Huang had argued publicly that ceding the Chinese market would harm US technology leadership. In an April interview with podcaster Dwarkesh Patel, he said abandoning that market would be "a disservice to our country."

The Backdrop to Huang's Exit

Huang's departure came after reports surfaced linking him to an affair with the chief executive of a company under investigation for smuggling chips into China. The circumstances have not been fully disclosed, but the timing of his exit and the immediate policy shifts suggest the board moved to contain both reputational and regulatory risk. Kress, who had served as chief financial officer, assumed the acting CEO role and has quickly signalled a different stance on China exposure.

The policy environment shifted again on Monday, when US authorities issued new rules barring advanced American chips from foreign data centres that serve Chinese customers remotely. The regulations close a route that had allowed Chinese AI labs to access prohibited hardware by renting capacity in facilities outside the country's borders. Investment bank Jefferies estimated the rules could reduce Nvidia's data centre revenue by between 44 billion and 62 billion US dollars next year.

Kress told analysts that strong demand for the company's newest architecture, Vera Rubin, means the export controls will not affect headline revenue in the near term. The backlog of orders from customers in markets where sales remain unrestricted is large enough to absorb the lost volume, she said. That assessment, however, does not account for the longer-term erosion of Nvidia's addressable market in Southeast Asia, where data centre operators have built business models around serving clients across borders.

Regional Implications for Data Centre Hubs

Singapore, Malaysia, Thailand, and Indonesia have emerged as key nodes in the Asia-Pacific cloud infrastructure landscape. Operators in these countries have invested heavily in facilities designed to serve customers throughout the region, including those in China. The new US rules require any large data centre purchasing Nvidia chips to certify that it does not serve Chinese customers. Export control attorney Dan Fisher-Owens noted that facilities unable to provide that certification will be forced to consider alternative chip suppliers, further fragmenting Nvidia's market share.

Singapore's role is particularly sensitive. Nvidia's subsidiary there reported revenue of 104 billion US dollars in 2025, a figure the company stopped disclosing shortly afterward. Critics have long argued that as much as 20 per cent of Nvidia's total revenue is linked to demand originating in China, either directly or through third-party channels. The opacity around these flows has made it difficult to assess the true scale of the company's exposure, but the decision to halt H200 sales and the new certification requirements together suggest that exposure is now being unwound.

The shift places Southeast Asian data centre operators in a difficult position. Many have built their growth strategies around serving a pan-Asian customer base, and the requirement to exclude Chinese clients will force them to choose between access to Nvidia's hardware and a significant portion of their potential market. Some may opt for chips from AMD or emerging suppliers, while others may seek to partition their infrastructure to maintain compliance. Either path introduces complexity and cost.

China's Response and Domestic Alternatives

Even before Nvidia's announcement, China had largely stopped accepting deliveries of H200 chips. Jacob Feldgoise, a researcher at Georgetown's Centre for Security and Emerging Technology, said the Chinese government prefers to stimulate demand for domestically produced semiconductors, even though major AI compute users still regard Nvidia's products as superior. The policy reflects Beijing's broader push for self-sufficiency in critical technologies, a campaign that has accelerated as US export restrictions have tightened.

Yet Chinese frontier AI laboratories have found ways to access Nvidia's hardware. Evidence suggests that all of the country's leading AI models have been trained on Grace Blackwell chips, which are prohibited from export to China. That training likely took place at data centres outside Chinese territory, a workaround the new US rules are designed to eliminate. The requirement for certification means that facilities hosting such training runs will now have to choose between Nvidia chips and Chinese customers.

The dynamic creates a paradox for US policymakers. The goal of export controls is to slow China's progress in AI by restricting access to the most capable hardware. But as long as Chinese labs can rent capacity abroad, the controls are porous. The new rules attempt to close that gap, but enforcement will depend on the willingness of data centre operators to comply and the ability of US authorities to verify their certifications. The risk is that some facilities will find ways to obscure the origin of their customers, or that Chinese entities will route their demand through intermediaries in jurisdictions with lighter oversight.

Nvidia's Strategic Recalibration

The creation of a board-level committee to oversee national security and export compliance signals that Nvidia is preparing for a more complex regulatory environment. The committee's mandate includes monitoring global trade risks, a recognition that the company's growth over the past five years has made it a focal point in US-China technology competition. The board's decision to establish this oversight mechanism immediately after Huang's departure suggests that directors believe the company's previous approach carried unacceptable risk.

Kress's statement that the new export controls will have no impact on top-line revenue is carefully hedged. It applies only to the near term and relies on the assumption that demand for Vera Rubin will absorb the volume previously destined for China-linked customers. Over a longer horizon, the loss of potential customers in Southeast Asia and the risk of further tightening in US export policy introduce uncertainty. Nvidia's dominance in AI accelerators is not in immediate jeopardy, but the company is now navigating a landscape where geopolitical considerations constrain its commercial strategy in ways that were less binding under Huang's leadership.

The shift also raises questions about how other US semiconductor companies will respond. If Nvidia's withdrawal from China-linked sales becomes the industry standard, competitors may follow to avoid regulatory scrutiny. Alternatively, some may see an opportunity to capture market share by offering products that fall below the performance thresholds targeted by export controls, or by structuring sales in ways that satisfy compliance requirements while maintaining access to contested markets.

What Comes Next

The immediate effect of Nvidia's decision is to align the company more closely with the preferences of US officials who have argued for stricter limits on China's access to advanced computing. The longer-term consequences will depend on how quickly Chinese domestic chip production improves, how effectively US authorities can enforce the new data centre certification rules, and whether Nvidia's competitors see an opening to expand their presence in markets the company is now exiting.

For data centre operators in Southeast Asia, the next few months will be a test of their ability to adapt. Those that can credibly certify they do not serve Chinese customers will retain access to Nvidia's hardware. Those that cannot will need to find alternative suppliers or restructure their operations. The outcome will shape the region's role in the global AI infrastructure landscape and determine whether US export controls succeed in their stated goal of maintaining a technology lead over China.

Read next
Policy

Nvidia Installs Board Committee for Export Controls After Huang Departure

Kenji Watanabe · 6 min
Policy

Nvidia Creates Board Committee to Police Chip Exports After CEO Exit

Hana Park · 6 min
Policy

Southeast Asia’s electric car sales rise 75% in first half of 2026, IEA says

Valerie Nguyen · 2 min
Spot something wrong? Email corrections@opentechwire.com. We log every correction publicly.