Why A.I. Tycoons Cannot Put America First
The fight over a national AI infrastructure reveals that even when they get their way, A.I. companies can't get out of their own way.
No one can ruin a good time like tech people.
On Tuesday afternoon, the heads of three vast tech firms stood with President Trump to announce plans to invest $500 billion in building data centers and power plants for the now entirely unregulated A.I. boom. It’s the dream scenario for any tech billionaire: White House approval, a national stage, blank-check permission. And yet almost immediately knives were out, and we were reminded yet again that just because modern tech companies have the resources of entire nations, they don’t really know how to behave in the national interest.
OpenAI CEO Sam Altman, SoftBank CEO Masayoshi Son and Oracle Chairman Larry Ellison are the heads of this new venture, and at the event they seemed happy to obey the number-one rule of appearing anywhere with Trump, which is to treat him as the main character. “We wouldn’t be able to do this without you, Mr. President,” Altman said from the lectern. Supplication is a small price to pay for carte blanche: Trump has already revoked Biden’s 2023 executive order that forced companies like Altman’s to report risks to consumers, national security, or public health to the government. And Biden’s FTC Chair Lina Khan, who sued tech companies left and right, has been replaced by Trump’s choice Andrew Ferguson, who in his pitch memo called Khan “anti-business.”
Tech leaders were happy with democratic administrations when they stayed out of the way (while confiding to reporters like me that people outside the industry weren’t smart enough to understand something like A.I. anyway), but Biden’s was the first administration to get involved, and as I wrote a few weeks ago, that was it for his support from Silicon Valley. So in spite of the awkward need to endlessly praise their host, this was a solid win for Altman, Son, and Ellison.
The new venture is being called The Stargate Project, in yet another moment where a business leader names something after a movie he clearly hasn’t watched fully. Elon Musk is fond of describing his personality in terms of movie characters. He once mentioned in a CNBC interview his admiration for Inigo Montoya’s single-minded determination to avenge his father in The Princess Bride, a misreading of a character so lopsided and fanatical that he cannot function once he’s achieved his life’s work. (Mandy Patinkin tweeted back “I do not think it means what you think it means,” God bless him.) Altman famously approached Scarlett Johansson to voice OpenAI’s chatbot, presumably thinking that the voice from the movie “Her” was the perfect one for his product, but also evidently not considering that it’s a movie about the social emptiness of life with technology. She refused, he did it anyway, and she sued.
Stargate is an amazingly awful choice for a name. The movie and subsequent TV series is about the discovery of a portal placed on Earth by extraterrestrials that can carry us to alien worlds. Amazing! Interplanetary travel made possible in an instant! What better inspiration could there be for the mind-bending potential of A.I.?
Well, if you watched the show — bear with me here — you’d know that in fact the portals are used by the insufferable Goa'uld race as a pipeline for slave labor from across the universe. See, the Goa’uld consider themselves gods, and the rest of us chattel. The humans in the show are under constant threat from these arrogant overlords and their overwhelming technological superiority, and spend 213 episodes of pretty entertaining television fighting back. All of this should have been brought up on the branding call.
Anyway, let’s talk about what this thing is supposed to accomplish, and why it brought out the daggers.
The big, foundational A.I. models — the ones run by OpenAI, Anthropic, Meta, Google — require outrageous amounts of expensive data processing to run. That, in turn, requires an outrageous amount of electrical power, along with attendant resources like water to cool the processors. The world of software and social media has created the illusion of a free and infinite supply of computing resources and human creativity, but A.I. is incredibly costly and deeply extractive. One A.I. executive told me that each request for an image from his A.I. company required 47 individual CPUs to spin up and complete the task. And as the AI Now Institute has been doggedly reporting for years, the climate goals of the biggest tech companies are now being actively cast aside as they build vast data centers that will inevitably produce huge amounts of carbon emissions.
In fact, in some cases the emissions of many of the firms that have made such pledges actually grew: Microsoft’s emissions increased from 11.6m metric tons of CO2 in 2020 to 13m metric tons in 2021. Instead of taking on the harder work of reducing their carbon demand, firms are pouring hundreds of millions of dollars into carbon-removal technology and carbon offsets, even as their emissions continue to grow. Such measures don’t have a good track record of effectively permanently removing carbon dioxide from the atmosphere and aren’t a substitute for reducing emissions overall.
This is why I hear from so many in tech that it’s time to resurrect nuclear power in the United States. Sam Altman and others have been pressing Washington on this since at least last year. In 2024 I traveled to Idaho to see a long-dormant nuclear plant that Oklo, a company backed by Altman, plans to restart, and since then several companies have joined the fray, hoping to capitalize on the desperate desire for more power from Silicon Valley. Regulatory hurdles have been the difficulty, however. In 2022 the Nuclear Regulatory Commission denied Oklo’s license application for lack of information. The NRC is supposed to be an independent and autonomous agency, but the president appoints its commissioners. Presumably Tuesday’s Stargate announcement, and the good will Altman and his two partners have won from President Trump, will help smooth the application process in future. In the meantime, Stargate’s founders revealed that the first of several enormous data centers is already under construction in Abilene, Texas. (Abilene Mayor Weldon Hurt told the AP that he’s excited about the project, but that “we didn’t know it was going to be quite this big. We thought it was going to be about a third of this size.”)
What won’t make this process any smoother is rivalry among billionaires. Elon Musk, who has $50 billion invested in his own A.I. company, has a long history of conflict in this arena. He’s fought with his peers for years over the direction A.I. should take, whether it’s arguing about robot overlords with Larry Page over a campfire or suing OpenAI, his onetime company, for pursuing an A.I. “dictatorship.”
But Tuesday’s White House event, where Trump called Altman “by far the leading expert based on everything I read,” seems to have really pissed Musk off. Musk, who has worked his way so deep into Trump’s inner circle that he rents a cottage at Mar-a-Lago, began firing off posts. First he took a shot at Masayoshi Son. “They don’t actually have the money,” he wrote beneath OpenAI’s announcement of the deal, adding later that “SoftBank has well under $10 [billion] secured. I have that on good authority.” For his part, Satya Nadella, who formed a key partnership with OpenAI and is involved with Stargate but wasn’t part of the White House announcement, said from Davos “all I know is I’m good for my $80 billion.” The casual sniping at this monetary scale is pretty amazing.
Musk criticizing Son is also an amazing piece of screenwriting, by the way. Kudos to whoever’s coming up with this stuff. The SoftBank CEO has made famously terrible investment decisions before, including ignoring everyone around him and handing billions to WeWork, which went on to light that investment on fire. And for years he had the distinction of losing the greatest amount of money in the history of capitalism, thanks to his failed investments in the dot-com boom of the 2000s. But he’s still among the richest men in the world, and of course his Guinness-recorded distinction was erased recently by none other than Musk, who now himself holds the record for single greatest loss. The kettle, the pot, etc.
Then Musk and Altman started jawing. “i genuinely respect your accomplishments and think you are the most inspiring entrepreneur of our time,” Altman wrote. Musk wasn’t placated, and fired off a series of posts calling Altman a liar and pointing out Altman’s past anti-Trump rhetoric. Altman wrote back “just one more mean tweet and then maybe you’ll love yourself…” He also seems to have felt compelled to apologize for his past opinions about Trump:
Looking back, I supposed all of this was inevitable. Altman said of Musk at the DealBook conference this year that “it would be profoundly un-American to use political power, to the degree that Elon has it, to hurt your competitors and advantage your own businesses.” He assured the audience that Musk wasn’t likely to do so. And so wow is it amazing to read the first sentence of the OpenAI announcement: “The Stargate Project is a new company which intends to invest $500 billion over the next four years building new AI infrastructure for OpenAI in the United States.” (Emphasis mine.)
Now a handful of people leading a dubious industry built on the promise of scientific breakthroughs (likely) and endless free time (unlikely) has an open runway to launch with little or no government oversight, and to make good on the promise that an investment like this is needed to get the US ahead of the Chinese A.I. sector, for whatever that’s worth. The question is whether the personalities involved can see beyond their own reputations, balance sheets, and shareholders to actually build something in the national interest.
It doesn’t have to be this way, of course. In the 1970s and 1980s the Taiwanese government went to enormous lengths to create public-private research partnerships in technology development. Its Industrial Technology Research Institute (ITRI), a sort of DARPA for incubating technology there, managed to bring technology from RCA to the country, and then spun off the United Microelectronics Corporation, handing it the 4-inch wafer technology that would go on to become the backbone of the modern semiconductor industry. The government went so far as to create a national foundry that could be used by the growing number of private companies there, as well as by global companies who didn’t have their own manufacturing capabilities. Sun Yun-suan, the economics minister who created ITRI and set all this in motion, said this in a 1979 speech:
The greater the difficulties of the nation, the greater our ability to understand that we can save ourselves and our country only by strengthening and relying on ourselves. The government is determined to work together with all the people of the country in these difficult times. For now our fundamental tasks are to sustain economic prosperity, develop culture, education, science and technology, seek political improvement and promote unity and harmony.
Sun got it all off the ground by working to convince Taiwanese legislators to invest just $10 million in the effort. They hated the idea, and it took more than a year to scrape together just enough support to pass, but there were no billionaires scrapping for credit or permission. Today that model has given rise to the world’s most dominant semiconductor company, TSMC, and a global network of customers so reliant on Taiwan that the United States (and President Trump specifically) may someday have to physically defend the island nation from invasion by China in order to preserve our access to their crucial technology. That’s how you launch an industry by presidential authority.
really enjoyed reading your thoughts on these latest ai news and putting this is perspective with the Taiwan - TSMC backstory