Context: The U.S. government announced in December that it planned to issue a new regulation (“Export Control Framework for Artificial Intelligence Diffusion”) that would tighten control over the sales of global AI chips, including by mandating that companies report key information to the U.S. government, and that they block Chinese access to AI chips. A number of companies and industry groups, including Oracle (January 5, 2025 Oracle blog post), the Semiconductor Industry Association (SIA) (January 6, 2025 SIA statement) and the Information Technology Industry Council (ITI), which represents companies such as Amazon, Microsoft, Meta, Google, Ericsson, and IBM (January 8, 2025 ai fray article), urged the government not to issue the rule.
What’s new: A final draft of the rules was officially released yesterday and the industry has 120 days to submit comments (January 13, 2025 Bureau of Industry and Security interim final rule). Nvidia’s vice president of government affairs Ned Finkle was one of the first to speak out after its release (January 13, 2025 Nvidia blog post). In a critical blog post about the incoming rule, he wrote: “Built on American technology, the adoption of AI around the world fuels growth and opportunity for industries at home and abroad. That global progress is now in jeopardy.”
Direct impact and wider ramifications: Not everyone will be heavily impacted by the rule, with 18 countries (“allies”) exempt, and some foreign governments having the option to sign agreements in exchange for looser restrictions. But this does not take away from the harsh criticism brought by the world’s largest tech players. Mr. Finkle believes it is “already undercutting U.S. interests” and, while referring back to the first Trump Administration, says Nvidia looks forward to a “return to policies that strengthen American leadership, bolster our economy and preserve our competitive edge in AI and beyond”. Donald Trump is due to take office on January 20. The 120-day comment period therefore also means the Trump Administration will have final say over the rule.
The six key mechanisms under the rule are:
- Allies receive no restrictions: this includes Australia, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Japan, the Netherlands, New Zealand, Norway, South Korea, Spain, Sweden, Taiwan and the United Kingdom.
- No caps on orders for chips that have a collective power of less than 1,700 advanced graphics processing units (GPU): this will apply to the majority of chip orders, such as those placed by universities, medical institutions, and research organizations for clearly harmless purposes.
- Companies with “Universal Verified End User” status can place up to 7% of their global AI computational capacity in countries around the world: this status is earned by meeting high security and trust standards – and companies must be headquartered in close allies and partners. This 7% could amount to as much as hundreds of thousands of chips.
- Those with a “National Verified End User” status can purchase computational power equivalent to up to 320,000 advanced GPUs over the next two years: this status will also require entities to meet the same security requirements and be headquartered in any destination that is not a country of concern.
- Entities that are located outside of close allies and do not have a special status can only purchase up to 50,000 advanced GPUs per country: this means U.S. technology will still be able to service foreign governments, healthcare providers, and other local businesses if need be.
- Governments can propose arrangements that align with U.S. needs to double their chip caps: those that align their export control, clean energy, and technology security efforts with the U.S. can raise their caps by up to 100,000 of today’s advanced GPUs.
In a statement yesterday, the U.S. government noted that “in the wrong hands, powerful AI systems have the potential to exacerbate significant national security risks, including by enabling the development of weapons of mass destruction, supporting powerful offensive cyber operations, and aiding human rights abuses, such as mass surveillance” (January 13, 2025 White House statement). This new “AI Diffusion” rule is meant to remove or at least diminish those risks by thwarting smuggling, closing other loopholes, and raising AI security standards, it said.
The statement added:
“To enhance U.S. national security and economic strength, it is essential that we do not offshore this critical technology and that the world’s AI runs on American rails.”
However, in his blog post yesterday, Nvidia’s Mr. Finkle wrote that the Biden Administration is seeking to use its last days in office to “undermine” America’s leadership with a “200+ page regulatory morass, drafted in secret and without proper legislative review”.
Mr. Finkle recalled the first Trump Administration, claiming that it “laid the foundation for America’s current strength and success in AI… as a result, mainstream AI has become an integral part of every new application driving economic growth, promoting U.S. interests and ensuring American leadership in cutting-edge technology”.
That administration demonstrated that “America wins through innovation, competition and by sharing [its technologies] with the world”, he said, adding that the adoption of AI is built on American technology and fuels growth and opportunity for industries at home and abroad.
But, he wrote, “that global progress is now in jeopardy”.
The rules are an attempt to rig market outcomes and stifle competition, threatening to squander America’s hard-won tech advantage, he said.
Mr. Finkle also emphasised that the rule is “cloaked in the guise of an ‘anti-China’ measure” but rather than mitigate threats will only “weaken America’s global competitiveness”.
In a separate statement yesterday, EU officials emphasised that selling advanced AI chips to EU members represents an economic opportunity for the U.S. and “not a security risk” (January 13, 2025 European Commission statement). They stated:
“We have already shared our concerns with the current U.S. administration and we are looking forward to engaging constructively with the next US administration. We are confident that we can find a way to maintain a secure transatlantic supply chain on AI technology and supercomputers, for the benefit of our companies and citizens on both sides of the Atlantic.”
Those concerns join those voiced by Oracle, SIA and the Information Technology Industry Council last week. The ITI wrote in a letter to U.S. Commerce Secretary Gina Raimond that the rule would “cede the global market to U.S. competitors” and that “should the U.S. lose its advantage in the global AI ecosystem, it will be difficult, if not impossible, to regain in the future”.