·NvidiaNvidia still isn’t sure if it can sell chips to China, threatening access to what CEO Jensen Huang calls a $50 billion opportunityBy Nicholas GordonBy Nicholas GordonAsia EditorNicholas GordonAsia EditorNicholas Gordon is an Asia editor based in Hong Kong, where he helps to drive Fortune’s coverage of Asian and economics news.SEE FULL BIO Jensen Huang, Co-Founder and CEO, NVIDIA speaks onstage at the All-In and Hill & Valley Forum "Winning The AI Race" at Andrew W.
Mellon Auditorium on July 23, 2025 in Washington, DC. Roy Rochlin—Getty Images for Hill & Valley ForumIt may take some time for Nvidia to see an upside from U.S.
President Donald Trump’s unprecedented offer to allow AI chip sales to China in exchange for a 15% cut.
The deal could shake up an export control regime designed to maintain the U.S.’s edge in strategic nologies.
Officials Treasury Secretary Scott Bessent are already positioning the arrangement as a model for other companies hoping to sell sensitive nology to China.
But during Nvidia’s earnings call on Wednesday, CFO Colette Kress noted that while the U.S.
government has expressed an “expectation” that it’d get 15% of Nvidia’s H20 revenue, it “has not published a regulation codifying such requirement.” In a later exchange filing, Nvidia warned that “any request for a percentage of the revenue by the USG may subject us to litigation, increase our costs, and harm our competitive position and benefit competitors that are not subject to such arrangements.” The U.S.-China war casts a shadow on Nvidia’s otherwise surging .
The company dropped China sales from its forecast, and its 10-Q filing includes a litany of warnings regulatory scrutiny from both Washington and Beijing.
And that’s a worry for Nvidia and its access to what CEO Jensen Huang described Wednesday as a $50 billion opportunity in the “second-largest computing market in the world.” Nvidia’s chips may be the world’s best AI cessors—but Chinese competitors are catching up.
Nvidia reported $46.7 billion in quarterly revenue, a 56% jump year-on-year, even as Nvidia revealed that it didn’t ship any H20s to Chinese customers last quarter.
Quarterly fit also jumped by 59% to hit $26.4 billion. The company forecast $54 billion in revenue for the next quarter, which would represent a more than 50% jump from the same period a year earlier.
Nvidia did not include possible H20 sales to China as part of its forecast, citing “geopolitical issues”.
Yet in her remarks, Kress said the company could see as much as $5 billion in H20 revenue in the coming quarter if geopolitical issues recede. “Every licensed sale we make will benefit the U.S.
economy,” she said. Sales to customers that use China as a billing location dropped to $2.8 billion, down from $3.7 billion a year ago.
Revenue from customers that invoice in Singapore rose 80% to $10.1 billion.
In its stock filing, Nvidia clarified that revenue booked to Singapore involves ducts that are “almost always shipped elsewhere,” and added that 99% of its “controlled Data Center compute revenue” billed to Singapore came from U.S.
based customers last quarter. Nvidia has been barred from selling its leading AI cessors to China since 2022. Huang has been a longtime critic of U.S.
export controls, complaining that they will instead encourage the growth of domestic Chinese alternatives and freeze U.S. companies out of the market.
In April, Nvidia revealed that it would need a license to sell the H20 cessor, its attempt to make a cessor that complies with U.S. law, to China.
Then, in late July, the Trump administration reversed course: As part of its trade war truce with China, the U.S. signaled that it might allow H20 sales in China.
s slump Despite the more-than-50% jump in revenue, Nvidia s dropped by 3.1% in ext trading, as data center revenue, at $41 billion, missed some of the more bullish estimates from analysts.
Asian had a mixed reaction to Nvidia’s earnings. Taiwan Semiconductor Manufacturing Company fell by 2.5%. Foxconn, which makes AI services, dropped by 1.0%.
Yet SK Hynix, a Korean supplier of high-bandwidth memory to Nvidia, jumped by 3.3%. Tokyo Electron, a major manufacturer of chipmaking equipment, rose by 1.9%. Chinese companies performed poorly.
Alibaba plunged by 4.4%, though negative sentiment may have been due to concerns over a fierce and expensive fight over food dery market with Meituan and JD.com.
Baidu, another leading Chinese AI developer, fell by 1.3%. But Chinese chipmakers, increasingly viewed as real competitors to Nvidia’s AI chips–at least in China–performed well in Thursday trading.
Semiconductor Manufacturing International Corporation, China’s national champion chipmaker, rose by 10.4%.
Cambricon, another local AI chipmaker, surged by 15.7%, and surpassed Kweichou Moutai as China’s most expensive stock.
Greater competition In its stock filing, Nvidia gave another stark warning its competitive position in China going forward.
“We may be unable to create a competitive duct for China’s data center market that receives apval from the USG,” it warned.
“In that event, we would effectively be foreclosed from competing in China’s data center computing/compute market.” Nvidia’s ducts—including its “fourth-best” H20 cessor—are still ahead of what Chinese companies can duce.
Yet domestic chipmakers, Huawei and its Ascend chips, are starting to catch up, and, importantly, may soon offer better ducts than what Nvidia or its peers are allowed to sell in China.
Qilai Shen—Bloomberg via Getty Images Beijing would be quite happy for Chinese companies to buy local alternatives.
Earlier this month, both Bloomberg and the Financial Times reported that Chinese officials were pushing local companies, particularly those affiliated with the government, to halt their purchases of Nvidia chips.
Officials have questioned whether Nvidia’s China chips pose a security risk, ing U.S. discussions of including backdoors and kill switches into Nvidia ducts to combat chip smudging.
Nvidia strongly denies including any such backdoors into its duct, and Huang has reiterated that Nvidia’s chips do not pose a threat to Chinese national security.
In its stock filing, Nvidia pointed out that China could exert its own regulatory pressure on the company, pointing to an antitrust investigation launched last year into its 2020 acquisition of Mellanox.
(Beijing has often used antitrust measures as a way to retaliate to U.S. export controls). Beijing officials are also asking whether complying with U.S.
export controls constitutes discrimination against Chinese customers.
China’s $50 billion opportunity On Wednesday, Huang suggested that China could offer Nvidia a $50 billion opportunity for the year, “if we were able to address it with competitive ducts.” U.S.
s have recovered since January’s DeepSeek shock, but China’s open-source models from companies Alibaba, Moonshot AI, Z.ai and DeepSeek have continued to impress outside observers.
And it’s helped boost Chinese s: Hong Kong’s Hang Seng Index is up by more than 25% for the year, compared to 10% for the S&P 500. China’s embrace of open-source models is also changing how U.S.
developers apach AI.
In recent weeks, both OpenAI and xAI have released open-source versions of their prietary models; OpenAI CEO Sam Altman even pointed to China as the reason behind his company’s return to open source.
“The vast majority of the leading open source are created in China, and so it’s fairly important, I think, for the American nology companies to be able to address that market,” Huang said Wednesday.
“The open source models that have come out of China are really excellent.” Earlier this month, Trump suggested that he was open to Nvidia selling a version of its powerful Blackwell cessor for the Chinese market.
CEO Jensen Huang previously confirmed that discussions were taking place, and in Wednesday’s earnings call, said “the opportunity for us to bring Blackwell to the China market is a real possibility.” “We just have to keep advocating for the sensibility of and the importance of American companies to be able to lead and win the AI race and help make the American stack the global standard,” Huang said.
Nvidia may have less time than Huang hopes. Last week, DeepSeek unveiled V3.1, the version of its AI model—with a special feature that optimizes performance on Chinese-made chips.