A proposed worldwide approach by the federal government on January 13 aims to restrict the dissemination of potent GPUs made in the U.S., with the goal of fostering U.S. superiority over China in the AI chip sector.
Emphasizing the importance of preserving U.S. national security and economic strength, the announcement from the White House states that it is crucial not to offshore this crucial technology and ensure that American rails support the world’s AI.
The proposal by the White House was pushed forward despite opposition from various tech giants, including NVIDIA and Oracle. Once the rule is enacted, it will be applicable after a 120-day period following a commenting phase. The upcoming administration will be responsible for deciding on the implementation of the rule.
Proposed rule categorizes nations into allied, restricted, or limited segments
In the Interim Final Rule on Artificial Intelligence Diffusion, there are suggested limitations on the distribution of U.S. chips to safeguard the U.S. market share against competitive technologies from China. Put forth under the Commerce Department’s Bureau of Industry and Security, this rule addresses national security concerns exacerbated by generative AI, which includes cyber assaults.
The rule comprises six major aspects:
- 18 U.S. allies will enjoy unrestricted access to chip sales.
- The White House stated that chip orders with collective computing power of up to approximately 1,700 advanced GPUs are exempt. This implies that most universities, medical institutions, and research centers can place chip orders without impediments.
- Trusted partners referred to as “Universal Verified End Users” possess U.S. authorization to deploy a small portion of their worldwide AI computational capacity.
- Countries not on the U.S. “country of concern” list or recognized as close allies that meet specific security prerequisites can be classified as “National Verified End Users.” This designation enables them to procure 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” include China and its Special Administrative Regions, Hong Kong and Macau.)
- Foreign governments, healthcare providers, and enterprises beyond the U.S. or its close allies not designated as National Verified End Users can purchase up to 50,000 advanced GPUs from the U.S. per country.
- Governments that engage in agreements with the U.S. specifying export control, clean energy, and technology security endeavors can double their chip allotment.
China will face restrictions on acquiring advanced U.S. technology in general and AI foundational model weights. Meanwhile, Russia will continue to be prohibited from procuring advanced chips from the U.S.
Which countries are granted unrestricted access to U.S. AI chips?
The 18 allied nations with unrestrained access to U.S. AI chips as per 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 disclosed by CNN, the classifications of countries with limitations on U.S. chip technology access could serve to impede China’s acquisition of U.S.-made chips or its expansion in advanced tech realms in those countries. Israel and Mexico are among the nations with restricted access.
Depending on their business scale, companies outside the U.S. may encounter supply chain delays or limited opportunities for incorporating AI capabilities into their products due to the policy.
“While administrative decisions and export controls may offer a narrow, temporary advantage, they could lead to a broader, long-term setback in American technological leadership,” remarked Benjamin Lee, a professor of engineering and computer science at the University of Pennsylvania, in a communication with TechRepublic.
“Initially, export controls might hinder certain countries’ deployment of cutting-edge GPUs and large AI data centers. However, in the long run, these controls may steer other countries towards developing their unique hardware designs or software frameworks.”
NVIDIA, Oracle vehemently object to the action
Ned Finkle, NVIDIA’s Vice President of Government Affairs, disputes the new regulations on two primary grounds. He argues that the distribution of robust AI chips is crucial for national security and that restricting AI purchases in certain nations will be advantageous for the U.S.
Addressing the issue, Finkle expressed, “Although presented as an ‘anti-China’ measure, these regulations will not bolster U.S. security. They would extend control globally, encompassing technology already widely available in mainstream gaming PCs and consumer devices.”
SEE: The EU endorsed a significant amalgamation between Ansys, a simulation software firm, and Synopsys, a chip design software provider.
While Ken Glueck, Oracle’s Executive Vice President, acknowledged the necessity of certain restrictions surrounding AI usage in scenarios like weapons of mass destruction and super-intelligent AI, he opposes the new White House policy. In a blog post published on January 5, he labeled the rule as “highly intricate and overly broad” on this page.
“In a singular perplexing move, the BIS retroactively governs global cloud GPU deployments; shrinks the global market for U.S. cloud and chip suppliers; enforces volume limitations; dictates that 20 nations can only be deemed trustworthy if they consent to new unilaterally imposed conditions—including certification and semi-annual reporting mandates—and likely steers the rest of the globe towards Chinese technology,” critiqued Glueck.
Conversely, as disclosed in a statement to the Associated Press, Microsoft President Brad Smith portrayed a less critical stance: “We are confident in our ability to fully comply with the stringent security standards outlined in this regulation and cater to the technological requirements of nations and clients globally who depend on us.”
About Author
