WASHINGTON — Nvidia can build the chips that power the world’s most sophisticated artificial intelligence, but building a way back into a market it once dominated has proved considerably more complicated.

Because China, once a $7 billion revenue stream for the company, remains stubbornly out of reach. American export restrictions have complicated Nvidia’s efforts to sell there, and Beijing has put up barriers of its own. Now, CEO Jensen Huang is trying to find his way back into a market he once dominated — and convince both governments that letting him in would be in their best interests.

“[China has] made it very clear that they don’t want Nvidia to be there right now,” Huang told reporters in Washington on Tuesday. “I hope that will change in the future.” He added a statistic that doubles as a warning: Half of the world’s AI researchers are Chinese, and his worry is that those researchers will keep developing AI with or without Nvidia — and if they build on someone else’s technology, America could lose its influence over the industry’s future and the engine of its own innovation.

Before Washington tightened its chip restrictions and Beijing began pushing domestic alternatives, Nvidia commanded about 95% of China’s advanced-AI chip market. “We went from 95% market share to 0%,” Huang said, confirming earlier remarks that the company is “100% out of China.” The numbers are clinical, but the stakes are existential: Analysts estimate that the China exclusion could cost the company as much as $15 billion in annual sales, while depriving the U.S. Treasury of roughly $3 billion in tax revenue.

Now, the plot thickens. As Huang heads to South Korea later this week for a summit of global tech executives and planned meetings with President Donald Trump — who is in the country for a trade summit with President Xi Jinping — Huang was asked: Are you also going to be meeting with the Chinese leader?  If not, who are you going to be meeting with? Will you talk with Trump about selling Blackwell chips in China? His response: “No comment. I’ve got a busy schedule coming up.”

He added, “There are a lot of meetings to be had. We have a lot of announcements to make there.”

Made in America — but missing a market

In Washington, Huang sounded like he was presenting the most patriotic version of a plea. Nvidia has spent the past year making its case in terms the Trump White House responds well to: more chips manufactured in the U.S. and advanced packaging brought back onto American soil (as well as some new supercomputers for the Department of Energy). “Nine months after [Trump’s] administration, here we are manufacturing Blackwell here in America,” Huang said, making Nvidia’s manufacturing progress part of Trump’s political success story.

But Huang had a less flattering message for Washington about the consequences of its China policy. “A policy that causes America to lose half the world’s AI developers … hurts us more,” he warned. His concern is that Chinese developers who can’t use Nvidia’s chips and CUDA software will learn to build on competing systems instead — and once that happens, it will be nearly impossible for Nvidia to win those customers back.

Huang made a similar-sounding argument in Taipei last spring when he called U.S. export controls “a failure” and told The Guardian that the restrictions “gave [China’s companies] the spirit, the energy, and the government support to accelerate their development.” He continues to argue that the restrictions intended to preserve America’s technological advantage are giving Chinese chipmakers every incentive to develop alternatives. Every door that closes for Nvidia opens a factory for someone else.

The power paradox

Chinese chipmakers have capitalized on Nvidia’s absence. Domestic GPU startups have tripled in valuation as state-backed projects are increasingly turning to local silicon. Washington’s restrictions may have been intended to protect America’s AI lead, but they’re giving Chinese companies a rather powerful incentive to develop their own AI alternatives. The longer Nvidia stays out of the country’s market, the more capable its competitors become.

But Huang isn’t ready to surrender the market. In a CNBC interview earlier this year, he described China as “a very large market … a $50 billion opportunity within two or three years.” He’s trying to convince Washington that allowing Nvidia to compete there would benefit American interests, too. But time is working against him. Every quarter that Chinese engineers spend developing alternatives to Nvidia’s technology just makes those alternatives more competitive.

Wall Street doesn’t seem to care — yet. Nvidia continues to post blockbuster results even without access to the Chinese market, thanks to the insatiable demand for its AI chips elsewhere. Investors have been happy to reward that success, but the longer Nvidia remains outside China, the harder it may become to reclaim its former position, and the more worried markets might become.

Huang has spent years building an ecosystem around Nvidia’s tech, making the company at the center of how much of the world is developing its AI prowess. China remains an “important market” for Nvidia, as Huang has said, but its developers are increasingly being pushed toward other technologies.

The tech CEO knows how to build the chips that power the AI boom. But building a bridge back to China will take something even he can’t manufacture: permission from two competing governments and enough customers waiting on the other side to make the crossing worthwhile.

This article first appeared here.

Privacy Preference Center