Biden administration officials have discussed capping sales of advanced AI chips from Nvidia and other American companies on a country-specific basis, people familiar with the matter said, a move that would limit some nations’ Artificial Intelligence (AI) capabilities.

The new approach would set a ceiling on export licenses for certain countries in the interest of national security, according to the people, who described the private discussions on condition of anonymity. Officials are focused on Persian Gulf countries that have a growing appetite for AI data centers and the deep pockets to fund them, the people said.

Deliberations are in early stages and remain fluid, the people said, noting that the idea has gained traction in recent weeks. The policy would build on a new framework to ease the licensing process for AI chip shipments to data centers in places like the United Arab Emirates and Saudi Arabia. Commerce Department officials unveiled those regulations last month and said there are more rules coming. 

The agency’s Bureau of Industry and Security, which oversees export controls, declined to comment. Nvidia, the market leader for AI chips, also declined to comment, as did Advanced Micro Devices. A representative for Intel, which also makes such processors, didn’t respond to a request for comment.

A spokesperson for the White House National Security Council declined to comment on the talks but pointed to a recent joint statement by the US and UAE on artificial intelligence. In it, the two countries acknowledged the “tremendous potential of AI for good,” as well as the “challenges and risks of this emerging technology and the vital importance of safeguards.”

Setting country-based caps would tighten restrictions that originally targeted China’s ambitions in Artificial Intelligence (AI), as Washington considers the security risks of AI development around the world. Already, the Biden administration has restricted AI chip shipments by companies like Nvidia and AMD to more than 40 countries across the Middle East, Africa and Asia over fears their products could be diverted to China. 

At the same time, some US officials have come to view semiconductor export licenses, particularly for Nvidia chips, as a point of leverage to achieve broader diplomatic goals. That could include asking key companies to reduce ties with China to gain access to US technology — but the concerns extend beyond Beijing. 

“We will have to have a conversation with countries around the world about how they plan to use these capabilities,” Tarun Chhabra, senior director of technology at the National Security Council, said at a forum in June without naming specific nations. “If you’re talking about countries that have a really robust internal surveillance apparatus, then we have to think about: How exactly will they use these capabilities to supercharge that kind of surveillance, and what will that look like?” 

There’s also the question of how global AI development could affect American intelligence operations, said Maher Bitar, another NSC official. “What are the risks not just on human rights grounds, but also in terms of the security and the counterintelligence risks to our personnel around the world?” Bitar said at the same event.  

It’s unclear how leading AI chipmakers would react to additional US restrictions. When the Biden administration first issued sweeping chip regulations for China, Nvidia redesigned its AI offerings to ensure it can keep selling into that market.

If the administration moves forward with country-based caps, it may prove difficult to deliver a comprehensive new policy in the final months of President Joe Biden’s term. Such rules could be challenging to enforce and would be a major test of US diplomatic relationships. 

Governments around the world are in a quest for so-called sovereign AI — the ability to build and run their own AI systems — and that pursuit has become a key driver of demand for advanced processors, according to Nvidia Chief Executive Officer Jensen Huang. Nvidia’s chips are the gold standard for data-center operators, making the company the world’s most valuable chipmaker and the top beneficiary of the AI boom.

China, meanwhile, is working to develop its own advanced semiconductors, though they still trail the best American chips. Still, there’s concern among US officials that if Huawei Technologies Co. or another foreign maker one day offers a viable alternative to Nvidia chips — presumably with fewer strings attached — that could weaken the US ability to shape the global AI landscape.

Some US officials argue that’s only a distant possibility, and that the Washington should adopt a more restrictive approach to global AI chip exports given its current negotiating position. Others warn against making it too difficult for other countries to buy American technology, in the event China gains ground and captures those customers. 

While officials have debated the best approach, they’ve slowed high-volume AI chip license approvals to the Middle East and elsewhere. But there are signs things could get moving soon: Under the new rules for shipments to data centers, US officials will vet and preapprove specific customers based on security commitments from both the companies and their national governments, paving the way for easier licensing down the road.

© 2024 Bloomberg LP

Intel, which is cutting thousands of jobs as it struggles to stay relevant in the chip industry, sold its 1.18 million share stake in British chip firm Arm Holdings in the second quarter, a regulatory filing showed on Tuesday.

Intel would have raised about $146.7 million (roughly Rs. 1,231 crore) from the sale, based on the average price of Arm’s stock between April and June, according to Reuters calculations.

The chipmaker said earlier this month that it would cut more than 15 percent of its workforce and suspend its dividend amid a pullback in spending on traditional data center semiconductors and a shift towards AI chips, where it lags rivals such as Nvidia.

Intel has said it is focused on developing advanced AI chips and building out its for-hire manufacturing capabilities, as it aims to recoup the technological edge lost to Taiwan’s TSMC, the world’s largest contract chipmaker.

The push to energise that contracting foundry business under CEO Pat Gelsinger has increased Intel’s costs and pressured profit margins, forcing it to seek cost cuts.

Intel and ARM both declined to comment on Tuesday when contacted by Reuters about the share sale.

“This looks to be consistent with the restructuring plan and the renewed focus on liquidity and efficiency that Gelsinger laid out from the last conference call,” said Benchmark Co analyst Cody Acree.

Santa Clara, California-based Intel had cash and cash equivalents of $11.29 billion (roughly Rs. 94,769 crore), and total current liabilities of about $32 billion (roughly Rs. 2,68,610 crore), as of end June.

Intel stock has lost more than 59 percent of its value so far this year, slumping 26 percent on August 2 after the company suspended its dividend. It was nearly flat in extended trading on Tuesday.

© Thomson Reuters 2024

(This story has not been edited by NDTV staff and is auto-generated from a syndicated feed.)