The Biden administration is introducing a new framework aimed at regulating the export of advanced computer chips used in the development of Artificial Intelligence (AI).
The proposal seeks to strike a balance between safeguarding national security and maintaining the economic interests of chip manufacturers and international partners.
However, the plan, unveiled on Monday, has prompted concerns among industry executives who warn that the new rules could restrict access to existing chips, particularly those used in video gaming, and limit the availability of chips essential for data centres and AI products in 120 countries. Nations such as Mexico, Portugal, Israel, and Switzerland could face limitations under the new regulations.
Commerce Secretary Gina Raimondo, in a briefing to the press, emphasised that it is vital to preserve the United States’ leadership in AI and the development of AI-related semiconductor technologies. AI’s rapid evolution promises significant societal changes, including advancements in scientific research, autonomous driving, and economic transformation. However, as the capabilities of AI grow, so do the risks to national security. Raimondo stated that the framework aims to protect the most advanced AI technologies from adversarial nations while ensuring their broader benefits are shared with partner countries.
White House National Security Adviser Jake Sullivan stressed that the framework intends to ensure that cutting-edge AI technologies are developed within the US and its closest allies, preventing the offshoring of vital technologies like those seen in the battery and renewable energy sectors.
Industry groups, such as the Information Technology Industry Council, cautioned that the swift implementation of the proposed rules could disrupt global supply chains and disadvantage US firms. The Semiconductor Industry Association also expressed disappointment that the policy was being fast-tracked before the presidential transition. SIA President John Neuffer warned that the new regulations could damage the US economy and global competitiveness, potentially ceding strategic markets to rivals.
A senior industry executive, who requested anonymity, indicated that the restrictions would limit access to chips already in use for video games, despite government assurances to the contrary. Furthermore, the framework could restrict which companies are permitted to build data centres abroad.
The proposal includes a 120-day comment period, meaning the incoming Republican administration, under President-elect Donald Trump, could ultimately decide the future of these export controls. This creates a complex situation where Trump would need to navigate between economic interests and national security concerns.
Officials expressed urgency in implementing the new framework to preserve what is perceived to be a critical advantage of six to 18 months over competitors like China in AI development, a lead that could diminish if rival nations stockpile chips.
Ned Finkle, Nvidia’s Vice President of External Affairs, pointed out that the previous Trump administration had helped lay the groundwork for AI’s growth. He argued that the proposed rules could stifle innovation without effectively bolstering US security. Finkle further asserted that the new regulations, while framed as an “anti-China” measure, would not enhance US security and could lead to global control over widely available technologies.
Under the new framework, around 20 key allies and partners, including countries like the UK, France, Germany, Japan, and Canada, would face no restrictions on chip access. Other nations would face limits on chip imports, with caps set at 50,000 graphics processing units (GPUs) per country. Additionally, government-to-government agreements could raise the cap to 100,000 units for countries aligned with US renewable energy and technological security goals.
Certain institutions in these countries could also apply for special legal status, allowing them to purchase up to 320,000 GPUs over two years. However, there will be limits on the amount of AI computational capacity that can be deployed abroad. Orders of up to 1,700 advanced GPUs would not require a licence, aimed at meeting the needs of universities and medical institutions rather than large-scale data centres.
The new rules are not expected to hinder the expansion of AI-driven data centres for major cloud providers like Amazon, Google, and Microsoft, as exemptions will apply for trusted companies seeking large AI chip clusters. Brad Smith, Microsoft’s president, expressed confidence that the company could meet the new security standards while fulfilling the technological needs of global customers.
Microsoft faced bipartisan scrutiny last year after announcing a significant investment in a UAE-based technology firm, G42, which operates data centres in the Middle East and is known for developing leading Arabic-language AI models. G42 also maintains ties with China, prompting calls from some US lawmakers for increased scrutiny of the AI partnership and stronger national security safeguards to prevent the transfer of critical US technologies to China.
—AriseNews
