The tech industry is shocked by Donald Trump’s unexpected agreement with Nvidia. Trump recently announced a deal allowing the company to sell its H20 chips to China, securing 15 percent of the revenue. “The H20 is outdated, but it still has market demand,” Trump stated during a Monday press conference. “So we struck a small deal.”
This unusual and legally questionable arrangement marks a significant shift from the Trump administration, which had previously prohibited all H20 sales to China earlier this year. The change reportedly occurred after Trump met with Nvidia CEO Jensen Huang, who has asserted that allowing Chinese entities access to H20 chips does not threaten US national security.
On the surface, the situation involves a powerful executive influencing a president to act in his company’s interests. However, there is a deeper and more intricate story beneath the surface. Last year, Nvidia released the H20 following a US government ban on selling a more advanced chip, the H800, to China. This action was part of a strategic initiative by Biden administration officials who felt it was necessary to stop China from being the first to develop advanced artificial intelligence.
For months, I collaborated with Stanford University researcher Graham Webster to explore the reasoning behind the Biden administration’s belief that restricting China’s access to advanced semiconductors was necessary. Today, WIRED is publishing Graham’s definitive account, based on interviews with over 10 former US officials and policy experts, some of whom requested anonymity.
“I pursued this story because it was evident that the official military and human rights justifications for the controls were not the sole reasons,” Graham explained. “AI was clearly a factor, and I sought to understand the deeper reasoning.”
Graham reports that several key Biden administration officials at the White House and Commerce Department were convinced that AI was reaching several inflection points that could offer substantial military and economic advantages to a nation. Some thought that an artificial general intelligence, capable of self-improvement, could be imminent. The risk of China achieving these milestones first was deemed too significant to disregard.
In response, the Biden administration decided to act. In the fall of 2022, they introduced comprehensive export controls to stop China from accessing the most sophisticated chips necessary for training powerful AI systems, along with specialized equipment Beijing required for advancing its domestic chip industry.
The move was the beginning of a long-term project that “intended to reshape the dynamics between the two global superpowers and influence the path of potentially one of the most consequential technologies of several generations,” Graham writes.
What I found noteworthy in Graham’s story is the number of individuals involved in Biden’s export control initiatives who have moved into influential roles in AI, computing, and national security. Jason Matheny, who oversaw the White House technology and national security policy, is now president and CEO of RAND, a significant think tank frequently engaged by government clients. Additionally, Tarun Chhabra, a National Security Council member, now leads national security policy at Anthropic.