By Karen Freifeld
(Reuters) -A leading Republican voice in Congress on China policy said on Wednesday that selling Nvidia’s best AI chip to China “would be akin (to) giving Iran weapons grade uranium”, as experts argued it would shrink the American advantage in artificial intelligence.
House Select Committee on China Chairman John Moolenaar said in a post on X that he told the administration, “we cannot sell the latest advanced AI chips to our country’s primary adversary.”
His comment came after President Donald Trump opened the door on Wednesday to Nvidia selling a lesser version of its Blackwell AI chip to China.
U.S. trade experts said giving China the chips could effectively spell the end of U.S. chip export restrictions, which were put in place in 2022 to make sure Beijing’s military would not benefit from American technology, and to slow the development of China’s AI efforts.
“If we decide to export B30As, it would dramatically shrink the U.S.’s main advantage it currently has over China in AI,” said Tim Fist, co-author of an analysis of the impact of allowing China the B30A chip, a downgraded version of Nvidia’s state-of-the-art Blackwell chip.
A spokesperson from Nvidia disagreed, saying the U.S. policy of controlling chips under President Joe Biden did not give America an advantage in AI. “Winning the support of mainstream developers everywhere enhances America’s economic and national security,” the spokesperson said in a statement.
“China has more than enough domestic chips for all of its military applications and has no reason to use our products for that purpose,” the statement said.
Senate Democratic leader Chuck Schumer and 11 Democratic senators also urged Trump on Wednesday to not lift restrictions on AI chips and American technology in pursuit of a trade deal.
Moolenaar said, “these chips should instead go to the U.S. companies that are building American AI dominance for years to come — not the future of the Chinese military.”
TRUMP MAY DISCUSS ‘SUPER-DUPER’ CHIP
Trump said on Wednesday he might speak to Chinese President Xi Jinping about Nvidia’s “super-duper” Blackwell chip at their Thursday meeting. The comments echoed those he made in August suggesting he might allow a 30% or 50% scaled-down version of Nvidia’s top chip to China.
But, Fist said, the B30A is a version of the best Nvidia chip in different packaging: China could buy twice as many and get the same result, likely at the same price.
A spokesperson for Nvidia declined comment on the details about the chip.
In the paper, published on Saturday, Fist and his co-authors analyzed nine scenarios covering a range of export strategies the administration might take for a downgraded Blackwell chip.
BEST- AND WORST-CASE SCENARIOS
In the best scenario, where no powerful chips are exported to China next year, the U.S. would have 30 times the AI computing power of China.
In the worst, where the U.S. allows the export of the B30A and comparable chips from other U.S. companies, China could surpass the U.S. in terms of how much AI computing power they gain in 2026.
Even in a median scenario, where a small amount of the chips is exported, the U.S. advantage shrinks to four times China’s computing power, the analysis found.
“If any meaningful quantities are allowed, it’s a huge change,” said Fist, director of emerging technology policy at the Institute for Progress, a Washington-based think tank. “It’s functionally ending the export control regime that we have today.”
Chris McGuire, a national security and technology expert who served in the U.S. State Department until last summer, agreed.
“If this chip is allowed to go, there are effectively no AI chip export controls anymore,” McGuire said. “The reason we have a big advantage on AI is because we have big advantages in computing power and in chips. If we give that away, best case is, it’s like a tie. Worst case, we fall behind.”
“We would be trading China our most advanced technology for soybean purchases,” McGuire said.
(Reporting by Karen Freifeld; Editing by Chris Sanders, Franklin Paul, Daniel Wallis and Deepa Babington)
