The Trump administration has announced a decision to formally approve exports of Nvidia’s H200 artificial intelligence chips to China, reopening one of the most sensitive fronts in U.S.–China relations.
How far Washington should go in restricting advanced technology without undermining its own industrial champions has been a critical issue in its recent trade policies.
Under new rules unveiled Tuesday, Nvidia can resume China-bound sales of its second most powerful AI chip, but only under a tightly controlled framework that reflects the competing priorities at play. Each shipment of H200 chips must be vetted by an independent third-party testing lab to confirm technical specifications, while exports to China cannot exceed 50% of the total volume sold to U.S. customers. Nvidia must also certify that sufficient supply remains in the United States, and Chinese buyers are required to demonstrate “sufficient security procedures” and pledge not to use the chips for military purposes.
Those guardrails did not exist previously and mark a shift from the Biden-era approach, which broadly barred sales of advanced AI chips to China. In a statement, Nvidia welcomed the move, saying President Donald Trump’s decision “strikes a thoughtful balance that is great for America” and allows U.S. companies to compete globally rather than ceding ground to foreign rivals already under sanctions.
“The administration’s critics are unintentionally promoting the interests of foreign competitors on U.S. entity lists,” Nvidia said, arguing that participation in “vetted and approved commercial business” supports American jobs and technological leadership.
Chinese technology companies have reportedly placed orders for more than 2 million H200 chips, priced at roughly $27,000 each, far exceeding Nvidia’s current inventory of about 700,000 units. Nvidia CEO Jensen Huang said last week at the Consumer Electronics Show in Las Vegas that the company is ramping up production amid surging global demand, with competition for H200 access already driving up cloud-computing rental prices.
Yet even as Washington clears the way, Beijing appears to be slamming on the brakes. Reuters reported, citing people familiar with the matter, that Chinese customs authorities have told agents that H200 chips are not permitted to enter the country, and domestic technology firms were summoned to meetings this week where officials instructed them not to purchase the chips unless absolutely necessary.
“The wording from the officials is so severe that it is basically a ban for now,” one source said, though they cautioned the stance could change as negotiations evolve.
Authorities have not clarified whether the directives apply to existing orders or only new purchases, and Chinese regulators have offered no public explanation.
That ambiguity has fueled speculation about Beijing’s motives. Analysts say China may be weighing whether to block the H200 outright to give domestic chipmakers more breathing room, or whether the restrictions are a tactical move to extract concessions from Washington ahead of President Trump’s planned April visit to Beijing for talks with Xi Jinping.
“Beijing is pushing to see what bigger concessions they can get to dismantle U.S.-led tech controls,” said Reva Goujon, a geopolitical strategist at Rhodium Group.
From Washington’s perspective, the decision has already drawn sharp criticism from China hawks. Saif Khan, who served as director of technology and national security on the White House National Security Council under former President Joe Biden, warned that the rule could dramatically boost China’s AI capabilities.
“The rule would allow about two million advanced AI chips like the H200 to China, an amount equal to the compute owned today by a typical U.S. frontier AI company,” Khan said, adding that enforcing customer vetting and preventing misuse by Chinese cloud providers would be extremely challenging.
The H200 sits at the center of this debate because of its performance. It delivers roughly six times the capability of the H20 chip, a weaker product that Trump banned and later allowed last year, only for Beijing to effectively block its import by August. That episode led Huang to say Nvidia’s share of China’s AI chip market had fallen to zero.
While Chinese firms such as Huawei have rolled out alternatives like the Ascend 910C, industry experts say Nvidia’s H200 remains far more efficient for training large, advanced AI models at scale. That efficiency is precisely what alarms U.S. lawmakers concerned about military and surveillance applications, even as the Trump administration argues that controlled exports could slow China’s drive to build fully indigenous replacements.
There is also a financial dimension. Re-entering the Chinese market would generate billions of dollars in revenue for Nvidia and significant income for the U.S. government, which is set to collect a 25% fee on approved chip sales. White House AI czar David Sacks and others contend that keeping China dependent on U.S. technology is preferable to pushing it into accelerating domestic alternatives beyond Washington’s reach.
Still, with Chinese officials signaling resistance and U.S. critics warning of strategic fallout, the fate of the H200 remains uncertain.
