Sam Altman’s $1 trillion AI gamble

Sam Altman, CEO of OpenAI, delivers remarks at the Federal Reserve on July 22, 2025. (Photo by Andrew Harnik/Getty Images)

Last week, OpenAI presented its plans to build $1 trillion of artificial intelligence infrastructure, including sprawling data centers across the U.S. that will require immense amounts of power and freshwater. OpenAI’s announcement and a flurry of deals that followed have driven the stock market to record highs.

But tech writer Ed Zitron, a fervent skeptic of the AI industry, believes that OpenAI’s spending commitments are not credible.

Zitron also argues that the vast sums of money being raised and spent by AI companies are not justified by the current or future capability of the technology. “There’s just a disconnection between what the models can do and what they will do, and what people are saying,” he said. “And it’s truly obscene because it’s allowed several people to get very rich, and it will fuck our economy and our markets.”

In an interview with Oligarch Watch, Zitron, the author of the Where’s Your Ed At newsletter and host of the podcast Better Offline, predicted that the current AI “boom” will result in “the worst bubble of all time.”

This interview has been edited for length and clarity.

Oligarch Watch: Could you walk me through your case for an AI bubble?

Ed Zitron: For at least a year, it’s been obvious there isn’t enough revenue in AI. When you add it all up—software, AI compute, GPUs for training and running models—it’s about $55 billion. That’s everyone combined: Microsoft, OpenAI, Anthropic, Meta, Amazon. Meanwhile, the “Magnificent Seven” alone are planning $400 billion in capital expenditures this year. Private investors have already sunk at least $500 billion into data centers. That’s half a trillion dollars invested in an industry with little real demand.

Nobody is making money. OpenAI loses billions, Anthropic loses billions, Microsoft loses billions. It’s not as if there’s huge revenue offsetting losses. There’s just not much revenue, period, and massive losses. It’s insane.

Oligarch Watch: What about the circular financing—OpenAI spending heavily on Nvidia products, while Nvidia invests back into OpenAI?

Zitron: Nvidia’s role is bizarre. OpenAI doesn’t have the cash to buy GPUs outright. CNBC has reported Nvidia will lease GPUs to them. If that’s true, it means Nvidia essentially buys its own GPUs and then rents them back to OpenAI.

It gets stranger with CoreWeave, a so-called “neocloud” company that sells nothing but AI compute. Nvidia invested in CoreWeave, sold it GPUs, propped up its IPO, and is also one of its four biggest customers. CoreWeave then uses that revenue and those GPUs to raise debt—so it can buy even more Nvidia GPUs. The same cycle exists with Lambda, another AI cloud startup Nvidia invested in and supplies.

Nvidia is the largest company in the stock market, yet when you describe these loops out loud, it’s maddening. None of this is emotional; it’s just objectively insane.

Oligarch Watch: OpenAI has floated $1 trillion in AI infrastructure projects. Is that remotely feasible?

Zitron: No. The top ten private equity firms combined—Blackstone, Carlyle, KKR, the lot—control about $477 billion. U.S. venture capital has maybe $164 billion left this year. There just isn’t enough money unless the entire world funds OpenAI and only OpenAI.

Altman might be right that OpenAI “needs” a trillion dollars. But consider: the U.S. military budget is around $900 billion. People don’t seem to grasp the scale. It’s shocking—and irresponsible—how few call it out. You can’t just shrug and say, “They’ll figure it out.” These amounts are ludicrous, impossible.

Oligarch Watch: Sam Altman and Mark Zuckerberg recently discussed the possibility of an AI bubble. What did you make of that?

Zitron: Zuckerberg compared it to the dot-com bubble, which is hilarious. What happened to those companies? Many collapsed. And the overbuild back then at least left us with useful fiber networks and cheap hardware. With AI, the overbuild is GPUs. AI GPUs are good at parallel processing—punching numbers—but not much else. They depreciate quickly.

So it’s worse than the dot-com bubble. AI spending promises are currently propping up the U.S. economy. In the first half of the year, data center expenditures contributed more to economic growth than all U.S. consumer spending combined. That’s unsustainable. And unlike past bubbles, there are no distressed assets to bail out. You can’t salvage specialized GPUs once demand collapses.

Oligarch Watch: Why are Silicon Valley, Washington, and Wall Street so intent on pushing AI into every corner of American life?

Zitron: Tech ran out of hyper-growth ideas. There’s no new cloud computing, no next smartphone. They needed AI to be the thing. Software-as-a-service is slowing, so they tried to make AI the next consumer revolution.

But I also think we’re in the era of the business idiot. Many executives have no connection to labor, no respect for the work that makes them rich. To them, large language models seem magical. They don’t know enough to see the limits, so they spout ridiculous claims. Over the past year I’ve read so many blatant lies about AI it makes you sick. But because they see AI as a way to replace labor, they’re thrilled.

The media hasn’t helped. Too many reporters won’t do the basic work of understanding the technology. What we’re left with is a confluence of ignorance and desperation—fueling what I believe will be the worst bubble of all time.

Oligarch Watch: What parts of the hype are most disconnected from reality?

Zitron: The casual way people say, “AI has replaced coders.” It hasn’t. There’s just a disconnection between what the models can do and what they will do, and what people are saying,” he said. “And it’s truly obscene because it’s allowed several people to get very rich, and it will fuck our economy and our markets.

OpenAI bears responsibility. They could have tempered expectations by being honest about model capabilities. Instead, they let the myths spread. By not correcting the hype, they allowed the narrative that “AI can do anything” to take root. What AI is actually good at is benchmarks—benchmarks designed for it.

Oligarch Watch: Are investors getting more cautious?

Zitron: We’re in the pornographic stage of the bubble. Money is still doubling and tripling down. Some private-equity people are nervous—they know what’s happening. But most are trying to “catch a knife,” hoping they can time the crash and escape before it all unravels.

It’ll take a big event to truly scare them. They’re uneasy, but not uneasy enough.

Oligarch Watch: Do you expect a wave of AI company failures?

Zitron: Absolutely. Nearly every company plugging a large language model into its product is unprofitable. Eventually, one of the big ones will run out of money, fail to raise more, and hit the wall. The moment that happens, panic will set in. Investors will realize, “Wait, my company loses money all the time too,” and the whole charade collapses.