Lawmakers are questioning President Donald Trump’s decision to approve sales of Nvidia H200 chips to China—artificial intelligence (AI) chips more powerful than anything U.S. export controls currently allow to China.

Sen. Elizabeth Warren (D-Mass.), ranking member of the Senate Banking, Housing, and Urban Affairs Committee, and Rep. Gregory Meeks (D-N.Y.), ranking member of the House Committee on Foreign Affairs, wrote a letter to Jeffrey Kessler, under secretary for industry and security, on Dec. 22, requesting information regarding the decision in which the president reversed position on AI chip export controls to China. The committees have oversight authority to review export control decisions.

Trump had announced on Truth Social on Dec. 8 that he would approve Nvidia H200 chips for sale to China with a 25 percent cut paid to the government and that other U.S. chip companies would have similar allowances. This would be the first time Nvidia’s Hopper line of AI chips would be allowed for sale to Chinese customers without modification.

The lawmakers said that this was a stark reversal from Trump’s April decision to halt sales of Nvidia H20 chips to China, a degraded version of the H200 that the U.S. company designed to meet export controls.

“The H200 is one of the most advanced AI chips on the market, and it is currently used to produce frontier AI systems with military applications,” the lawmakers wrote.

They said that they are concerned the decision “undercuts” national security.

They are requesting details of export control rule changes, any government-to-government agreement signed, the Commerce Department’s assessment of China’s most advanced chips and chipmaking capacity, and information on approved licenses within 48 hours of approval.

Reuters reported on Dec. 18 that agencies are already reviewing applications for H200 chip sales to China.

Several lawmakers have criticized the decision to allow H200 chip sales to China, saying it undermines U.S. export controls meant to rein in Chinese military modernization.

The H200 chip is more advanced than anything China currently has the ability to manufacture domestically, especially given the export controls that block Chinese companies from being able to purchase the most advanced semiconductor manufacturing equipment.

Analysts have compared Chinese and U.S. computing power in the AI race based on whether China has the ability to access H200 chips, which the Biden administration had banned from sale to China. Researchers largely estimate that the gap will narrow to a fine point, if not erode completely, in just a few years if sales are approved.

The Institute for Progress had published one such analysis just before Trump announced the approval of sales, estimating that without the H200, the United States would have 21 to 49 times more AI computing power than China in 2026. The researchers estimated this advance will shrink down to 1.3 to four times if H200 sales to China are approved but rise to 2.1 to 6.7 times if U.S. companies switch over to the latest generation Blackwell chips.

Published shortly after Trump’s announcement, an analysis by Chris McGuire, senior fellow for China and emerging technologies at the Council on Foreign Relations, showed that H200s could up to triple the trajectory of Chinese AI computing power.

McGuire wrote that the tech executives’ theory that U.S. export controls would only spur Chinese AI development was contrary to Chinese hyperscalers’ own estimates, which predicted a decrease in computing power over the next two years if they could no longer access Nvidia chips.

Chinese companies such as Huawei and DeepSeek have touted the strategy of using higher chip quantities when more advanced chips are not available. McGuire found that even with this strategy, the gap between the United States and China would widen exponentially with each passing year, as Chinese manufacturing quantity cannot catch up with increasingly advanced chips being manufactured at scale outside China.

Comparing Huawei and Nvidia AI capacity, he found that by 2027, even if Huawei were to increase production “a hundredfold,” it could not reach half of Nvidia’s output.

“Even under very aggressive assumptions about Huawei’s AI chip production capacity—that it will produce 800,000 AI chips in 2025 (double the highest public estimates), two million AI chips in 2026, and four million in 2027—it will not be enough,” he wrote. “Huawei would still produce only about 5 percent of Nvidia’s aggregate AI computing power in 2025, falling to 4 percent in 2026 and 2 percent in 2027. Huawei is not a threat that justifies loosening controls; it is evidence that the controls are working.”