News

Elon Musk believes the nation that controls advanced chipmaking capacity will win the AI race.

“In the next few years, I think America is likely to win,” Musk said in a March 17 interview with U.S. Senator Ted Cruz. “Then it will be a function of who controls the AI chip fabrication. The factories that make the AI chips, who controls them? If more of them are controlled by China, then China will win.”

Taiwan’s TSMC makes all the AI chips today, and that is a national security issue for the U.S., according to Musk.

Musk is making big bets on AI. Nvidia and Musk’s xAI joined a consortium backed by Microsoft, investment fund MGX and BlackRock to expand U.S. AI infrastructure in a race to dominate the new technology, Reuters reported on March 19. Musk’s Tesla is using TSMC in Taiwan to make AI chips for its Dojo supercomputer that is developing training models used for autonomous driving in the automaker’s cars.

“Currently, 100% of the advanced AI chips are made in Taiwan,” Musk told Cruz. “If [China] were to invade in the near term, the world would be cut off from advanced AI chips.”

China, which claims Taiwan as part of its sovereign territory, has escalated military exercises around Taiwan in recent months. The island nation is at the center of a tech war between China and the U.S. that could morph into a hot war.

The supply of AI chips is “essential for national security, and we’re not doing enough,” Musk said.

Musk is correct on U.S. national security, AI chips and Taiwan, SemiAnalysis analyst Jeff Koch told EE Times.

“Risk concentration in Taiwan is extreme for electronics generally and especially AI chips,” Koch said. “The problem is not just advanced logic but also high-bandwidth memory and advanced packaging, as well. All three capabilities are key, and the U.S. has little to no capacity—at least for commercially relevant processes like TSMC’s N3 and CoWoS.”

Musk should have the ear of U.S. President Donald Trump, who made Musk his right-hand man leading the newly created Department of Government Efficiency. Trump last week announced that TSMC will add $100 billion of investment to its Phoenix, Arizona, production facility, helping reduce exposure to Taiwan.

Still, TSMC did not provide a technology roadmap or timeframe for the new investment in Phoenix.

“Vague commitments with a large dollar value aren’t enough: they must specify timing, process technologies and capacity to be built,” Koch said. “These specifics are needed to give the U.S. confidence it will not face a gap in AI capabilities over the medium and long term.”

AI dependence

Nearly all advanced GPUs from Nvidia, as well as other AI ASICs from major players like Google, AWS and Meta, are made by TSMC in Taiwan, Paul Triolo told EE Times. Triolo advises global tech clients at Washington, D.C.-based Albright Stonebridge Group.

“This heavy dependence is not going to be reduced anytime soon, with no major dent before 2030,” he said. That is only if “everything goes smoothly with TSMC investments in its Arizona gigafab and Intel and Samsung become competitive in this space,” he added.

Delays in U.S. production starts by Intel and Samsung are signs that the CHIPS Act stimulus measures created under the Biden administration may not be meeting expectations. The two companies have so far failed to catch the AI wave.

“AI chip production will take years to onshore,” Koch said. “The U.S. government needs to continue incentivizing and pressuring chipmakers to onshore key capabilities, or we face the risk of not just losing our lead in AI chips but actually falling behind China over a five-year time span.”

Triolo sees risks in framing the AI competition as a race between China and the U.S.

“Assertions from some quarters that getting first to artificial general intelligence or advanced machine learning will give the winner a decisive strategic advantage is a dangerous formula,” he said. “This is because the hardware base for AGI/AML will continue to exist 100 miles off the coast of China, and further rhetoric around this, coupled with more U.S. controls on Chinese firms, cutting them off from Taiwan, substantially raises the risks around a conflict over Taiwan, with catastrophic consequences.”

Other analysts said Musk is correct to highlight an excessive reliance on Taiwan for AI chips.

“Musk is not saying anything different from what I have been preaching for over 10 years,” TechInsights vice chair Dan Hutcheson told EE Times. “The difference is that my flock is far smaller. Musk is simply coming from a realpolitik analytical perspective.”