Making Sense of Sam Altman’s $7 Trillion AI Chips Gambit
Think of Sam Altman as the Taylor Swift of tech — he consistently finds ways to remain in the headlines for the most outrageous and glamorous reasons. His attempt to bank $7 trillion to build AI chip factories had many eyes rolling.
And most recently, his company OpenAI was in the news for the Sora text-to-video AI, which Microsoft CTO Mark Russinovich tweeted was “both amazing and frightening. I think it marks the end of reality as we know it.”
Altman’s ambitious $7 trillion fund to create a string of AI chip factories would make him the manager of the world’s third-largest economy.
“What is it with Sam Altman that people are tracking him like he’s Taylor Swift? Instead of Swifties, maybe we should call Altman’s fan base ‘Sammies,’” TechInsights analyst Dan Hutcheson wrote.
Why Sam Altman Wants to Invest $7 Trillion into AI
Altman’s $7 trillion play could be an ambitious plan to restructure the entire semiconductor industry, and the fund is around two to three times the size of the electronics industry, Hutcheson wrote when explaining a summary of the Chip Insider newsletter published by the analyst firm.
Altman’s big-picture idea could span over decades, and prepare the world for AI-disrupting electronics. Electronics will need a giant supply of AI chips in the coming decades.
His idea could also be to disrupt legacy chip supply models. Today, chip supply is in the hands of a few — effectively Intel and Taiwan Semiconductor Manufacturing Co. — who are thinking short term. Their unpreparedness led to the recent chip shortage that affected the automobile and electronics industries.
“Reverse engineering Altman’s tactical approach into his strategy… the funding scale difference reveals that the strategy is more likely to buy the companies he needs, rather than build his own,” Hutcheson wrote in the summary.
The Gamble
The $7 trillion stunner was originally reported by The Wall Street Journal earlier this month.
Altman’s $7 trillion plan covers everything needed to operate AI chip factories, including electricity and equipment. He talked to the UAE government’s venture investment firms and Softbank CEO Masayoshi Son to provide funding and has approached chip maker TSMC to manage the factories.
Altman doesn’t think small — he wants to be the king of chips, like he is of AI. When Altman let the AI genie out of the bottle, it was viewed as a world-changing technology much like the internet in the 1990s.
AI has two foundational legs driving its swift adoption. One is AI models, which has been replicating quickly in line with the cut-and-paste nature of software. The other is hardware such as GPUs, which made it possible for OpenAI put decades-old AI theory to practice.
But the hardware can’t keep up with the software advances. GPUs and other chips are scarce. Nvidia is the major player here, and top companies are taking whatever GPUs they can get their hands on. Rival companies AMD and Intel are just getting started with their own AI hardware, but thus far have no significant sales.
Nvidia has become rich, and the company’s market cap is approaching about $2 trillion. Nvidia has repeatedly said the production of GPUs can’t keep up with the demand, which is the problem Altman is looking to solve.
The Challenges
It is hard for chip experts to get their head around the $7 trillion number and the challenges surrounding it. The top concern: the semiconductor industry is volatile with up and down cycles every two to three years. Companies like Intel make a lot of money when demand is high, but also lose a lot of money when the going gets tough.
Altman is also digging his toes in geopolitical hell with technology policy and regulation debates squarely directed at AI and semiconductors, said Nathan Brookwood, principal analyst at Insight64.
Governments are weaponizing semiconductors to enforce trade policies. The US government is trying to choke China’s AI efforts by banning the exports of advanced AI chips such as Nvidia’s GPUs.
Altman’s goal to establish a global supply chain of AI chips will likely hit many roadblocks.
Researchers recently suggested regulating AI computing power. Some ideas included introducing a kill switch for chips running harmful AI, an international AI chip registry and required reporting of training compute usage from cloud providers and AI developers. The paper, which came out of Cambridge University, implied that AI chips were weapons, much like nuclear weapons back in the day.
“The ability to reliably exclude people from accessing these informational goods is much weaker than for physical goods, as evidenced by, for example, the history of nuclear technology,” wrote researchers in the paper.
The idea may be noble, but it is difficult to regulate hardware, Brookwood said, adding that hardware makers like Nvidia are already finding ways to work around US regulations to continue supplying its GPUs to China.
Tracking the Cash
Altman’s ask of $7 trillion exceeds every available number about the chip market today. The total chip market revenue was about $520 billion in 2023, a decline of about 10% from 2020, according to semiconductor industry consortium SEMI.
The numbers will rebound and grow by up to 16% in 2024, said Ajit Manocha, president and CEO of SEMI, in a blog entry.
“Like many others, I remain bullish on the industry reaching $1 trillion around the end of this decade,” Manocha wrote.
Rival chip makers are also using less money to set up the latest chip factories.
In 2022, Intel announced it was investing $20 billion to establish a new cutting-edge chip factory in Columbus, Ohio. Intel is investing €80 billion (US$86 billion in today’s numbers) to build out its Europe chip infrastructure, including a factory in Germany, research facilities, packaging facilities and other sites.
Even with a dozen new factories, Intel’s investments won’t surpass $1 trillion. TSMC this month said it would invest $20 billion for a new factory in Japan.
Rivals Strike Back
The United Arab Emirates (UAE) government insisted that the world needs the $7 trillion factories.
But that idea was shot down by Nvidia CEO Jensen Huang at the recent World Governments Summit in Dubai. Huang was asked a light-hearted question about how many GPUs $7 trillion could buy,
“Well apparently all the GPUs,” Jensen said. Nvidia’s top-line H100 GPUs retail at between $30,000 to $40,000. OpenAI’s GPT-4 models run on Nvidia’s GPUs
Ironically, Huang was interviewed by Omar Sultan Al Olama, who also interviewed Altman at the same summit. There were no questions about Altman’s AI chip plans.
The UAE clearly wants to be a larger player in the global chip market, and it already has a stake in GlobalFoundries, which is a top-10 chip manufacturer.
“In the UAE we don’t lack ambition,” Al Olama said.
A staggering lack of talent is also affecting the semiconductor industry, and whether Altman will find workers to fill up factories is an open question.
“The industry remains in the grip of a global talent shortage. We project the industry will need 1 million additional workers by 2030” SEMI’s Manocha said in the blog entry.
