The federal government on Jan. 13 proposed a global policy to limit the distribution of powerful, U.S.-made GPUs, an effort to promote U.S. dominance over China in the AI chip industry.

“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,” the White House’s announcement reads.

The White House moved forward with the proposal despite pushback from several tech giants, including NVIDIA and Oracle. If the rule is implemented, it will be enforceable after 120 days, after a comment period. The incoming administration will determine whether the rule will go into effect.

Proposed rule sorts countries into allied, restricted, or limited categories

The Interim Final Rule on Artificial Intelligence Diffusion proposes restrictions on U.S. chip distribution to secure U.S. market share against competing technologies in China. Issued under the Commerce Department’s Bureau of Industry and Security, the rule mitigates national security risks exacerbated by generative AI, including cyberattacks.

The rule has six main parts:

• 18 U.S. allies will have unlimited access to chip sales.
• “Chip orders with collective computation power up to roughly 1,700 advanced GPUs” are exempt, the White House said. Therefore, most universities and medical and research institutions can place chip orders without roadblocks.
• Trusted partners known as “Universal Verified End Users” have U.S. permission to place a small percentage of their global AI computational capacity globally.
• Countries on neither the U.S. “country of concern” nor close ally list that meet certain security requirements can be listed as “National Verified End Users.” This designation allows them to purchase computational power equivalent to up to 320,000 advanced GPUs from the U.S. for use within neutral countries over the next two years. (The “countries of concern” are China and its Special Administrative Regions, Hong Kong and Macau.)
• Foreign governments, health care providers, and businesses outside of the U.S. or its close allies not designated National Verified End Users can purchase up to the equivalent of 50,000 advanced GPUs from the U.S. per country.
• Governments that sign arrangements with the U.S. to specify export control, clean energy, and technology security efforts can double their allotted number of chips.

China will be restricted from advanced U.S. technology in general and AI foundation model weights. Russia also remains blocked from purchasing advanced chips from the U.S.

Which countries have unlimited access to U.S. AI chips?

The 18 ally countries with unlimited access to U.S. AI chips under the policy are:

• Australia.
• Belgium.
• Canada.
• Denmark.
• Finland.
• France.
• Germany.
• Ireland.
• Italy.
• Japan.
• Netherlands.
• New Zealand.
• Norway.
• Republic of Korea (South Korea).
• Spain.
• Sweden.
• Taiwan.
• United Kingdom.

As CNN points out, the categories of countries with caps on how much U.S. chip technology they can access could help prevent China from acquiring U.S.-made chips or getting its foothold in the advanced tech industry in those countries. Israel and Mexico are among the countries with limited access.

Depending on the size of their operation, companies outside the U.S. could see delays in their supply chains or limited opportunities to add AI features to their products due to the policy.

“Although administration rulings and export controls produce a narrow, short-term advantage, they may produce a broader, long-term loss to American technological leadership,” Benjamin Lee, University of Pennsylvania professor of engineering and computer science, told TechRepublic in an email.

“In the short-term, export controls will slow some country’s deployment of the most advanced GPUs and the largest AI data centers. But in the long-term, export controls will cause other countries to develop their own hardware architectures or software models.”

NVIDIA, Oracle strongly oppose the move

NVIDIA Vice President of Government Affairs Ned Finkle disagrees with the new rules on two major points. He says that distributing powerful AI chips is a matter of national security and that restricting AI purchasing in some countries will benefit the U.S.

“While cloaked in the guise of an ‘anti-China’ measure, these rules would do nothing to enhance U.S. security,” Finkle wrote. “The new rules would control technology worldwide, including technology that is already widely available in mainstream gaming PCs and consumer hardware.”

SEE: The EU approved a major merger between simulation software company Ansys and chip design software provider Synopsys.

Ken Glueck, executive vice president of Oracle, acknowledged the importance of some restrictions around AI use in cases such as weapons of mass destruction and super-intelligent AI. However, he also opposes the new White House policy, calling the rule “highly complex and wildly overbroad” in a blog post on Jan. 5.

“In a single confusing action, the BIS retroactively regulates global cloud GPU deployments; shrinks the global market for U.S. cloud and chip suppliers; establishes volume restrictions; tells 20 countries they can be trusted only if they agree to new unilaterally imposed terms—including certification and semi-annual reporting requirements—and likely pushes the rest of the world to Chinese technology,” wrote Glueck.

However, in a statement to the Associated Press, Microsoft President Brad Smith was less critical: “We’re confident we can comply fully with this rule’s high-security standards and meet the technology needs of countries and customers around the world that rely on us.”