President Trump’s announcement that the American chip maker Nvidia would be able to sell semiconductors to China appeared to roll back years of policies aimed at blocking the country’s access to advanced technologies.
At stake, officials in Washington have argued, is China’s ability to use American technology to gain military and economic advantages, particularly through artificial intelligence. By cutting China off, lawmakers wanted to preserve the lead the United States has achieved in A.I. and slow China’s efforts to catch up.
On the other side, Nvidia’s chief executive, Jensen Huang, spent months trying to persuade Mr. Trump to reverse course. He argued that limiting China’s access to American technology has only spurred Chinese companies to improve faster.
Mr. Trump’s decision on Monday to allow Nvidia to sell its second-most-powerful chip, known as the H200, to Chinese commercial customers vetted by the U.S. Commerce Department strikes “a thoughtful balance that is great for America,” an Nvidia spokesman said.
The impact of Mr. Trump’s decision is unclear. But the Chinese government has already spent billions in effort to become an A.I. superpower. The country’s chip makers are racing to make advanced chips on their own, and Chinese A.I. companies have become more efficient at using the chips they have access to.
“The push for domestic production will continue,” said Jiang Tianjiao, an associate professor at Fudan University. “Beijing’s strategy is clear.”
Investor bets that a Chinese company could soon compete with Nvidia have driven huge gains in the country’s tech stocks this year. Shares of the partly state-backed chip designer Cambricon have risen more than 120 percent, compared with a year ago. Last week, Moore Threads, a start-up chip maker founded by a former Nvidia executive, went public in one of the Shanghai Star Market’s largest offerings of the year.
But even Huawei, the telecommunications giant leading China’s race to compete with Nvidia, remains at least two years behind the Silicon Valley company, experts said.
Chinese chips still lag behind Nvidia’s in performance, and the companies face major constraints on their output. Chinese chips can produce, at most, 2 percent as much computing power as foreign rivals, said Tim Fist, a senior adjunct fellow at the Center for a New American Security, a think tank.
Computing power is crucial in A.I. development. Advanced A.I. systems are powered by hundreds of thousands of chips known as GPUs, which companies pack into buildings known as data centers to create giant supercomputers that gobble electricity. OpenAI, the A.I. start-up that created ChatGPT, is planning to build five such facilities in the United States that would together consume more electricity than the three million households in Massachusetts.
Chinese demand for cutting edge chips is also high. The biggest tech companies, including Alibaba, Tencent and ByteDance, TikTok’s parent company, have poured money into A.I. efforts and infrastructure like data centers.
For Chinese companies, “gaining access to higher-end chips like the H200 does indeed come at a critical moment,” said Frank Kung, an analyst at TrendForce, a market research firm in Taipei, Taiwan.
Some officials who worked for former President Joseph R. Biden Jr. have warned that access to the Nvidia chips could help China’s A.I. companies buy time while the country’s chip makers improve the supply and performance of their own offerings.
Mr. Trump’s decision to allow Nvidia to sell advanced chips in China has also renewed a debate about how much future advancements in A.I. technology will depend on vast stockpiles of powerful chips.
Washington’s constraints have pushed Chinese companies to be more efficient. This year, the Chinese start-up DeepSeek rocked the tech industry with its claim that it had created powerful A.I. systems that required fewer chips and were significantly cheaper to build than those of better-funded American rivals.
The less powerful China-made chips require more electricity. But industrial electricity costs in China are far lower than in the United States, said Lian Jye Su, a chief analyst at Omdia, a tech research firm. The abundant cheap power has helped companies create sophisticated A.I. systems despite less powerful chips, Mr. Su said.
The Chinese government has pushed companies there to buy domestic chips and warned Nvidia’s chips may carry risks. In July, China’s internet regulator said that it had summoned Nvidia to explain security risks associated with one of its chips developed for sale in China, saying the chips could be shut down remotely or used to track a user’s location.
Many Chinese tech companies are already using a mix of domestic and foreign chips, said Paul Triolo, a partner at DGA-Albright Stonebridge Group. “Just because some Chinese companies will choose to purchase some number of H200 does not mean domestic development of advanced A.I. hardware will stop,” he added.
Mr. Trump’s willingness to widen China’s access to advanced American chips has drawn condemnation from members of Congress and former Biden administration officials.
This summer, Mr. Trump proposed the idea of the government taking a cut of A.I. chip sales to China. On Monday, he reiterated that, writing that 25 percent “will be paid to the United States of America.”
Mr. Trump’s transactional approach to export controls on chips has led to some paradoxical outcomes. Hours before he said the Commerce Department was finalizing the details to allow sales of the H200 in China, the Department of Justice announced it had detained two people for selling those chips to the country.
Xinyun Wu contributed reporting from Taipei.
Meaghan Tobin covers business and tech stories in Asia with a focus on China and is based in Taipei.
The post How Much Could China Gain From Access to Powerful Nvidia Chips? appeared first on New York Times.
