We have officially entered the era of trillion-dollar math that doesn't add up.

On Wednesday, the U.S. military pulled up its tents and left Iraq after more than 23 years, 4,418 U.S. deaths, and $1.7 trillion in spending.

This week we also learned from Anthropic's long-anticipated initial-public-offering prospectus that the artificial-intelligence colossus is eyeing a $2 trillion valuation.

Add to that a new Brookings paper, which says the AI infrastructure buildout will total $10.3 trillion from 2025 to 2032.

All mind-blowing numbers. But is any of this worth it? (Especially the thousands lost in Iraq, which puts the rest in perspective.)

It would be hard to deny that AI mania has us, Icarus-like, flying too close to the sun. Nvidia, the poster child of AI exuberance, is up 1,249% since ChatGPT was launched on Nov. 30, 2022-even as AI models have gone rogue countless times. And still, Wall Street is shoveling money into these enterprises. All of which should give us pause.

But it hasn't, evident in light of the recent man-bites-doggery in which President Donald Trump pooh-poohs AI naysayers while AI CEOs practically beg for federal oversight. Mostly what came out of Big Tech's big sit-down at the White House this week, besides a "morally binding" agreement and the president saying AI is "going to be very, very safe," was a photo op reminiscent of a Soviet-era parade reviewing stand, replete with a POTUS-posted seating chart.

A majority of Americans think we need more. In a new poll, over half of those surveyed by The Wall Street Journal want safeguards on AI, and nearly two-thirds want a pause in the development of the technology.

"I'm particularly concerned about [AI companies] pushing the idea that they can regulate themselves," says Geoffrey Hinton, a Nobel laureate in physics who worked at Google parent Alphabet for a decade and has been called "the Godfather of AI." "Nobody thinks drug companies should be allowed to regulate themselves. The FDA forces them to prove their drugs are safe before their release. At the very least, we should have something like that for AI."

Though a number of dissenters from AI companies have resigned in protest recently, some from Silicon Valley were wary of tech running amok years earlier, such as Frances Haugen of Facebook parent Meta Platforms, and Tristan Harris, Timnit Gebru, and Hinton of Alphabet.

One of the first and loudest of the technorati to turn tech-wary was Roger McNamee, co-founder of Silver Lake and Elevation Partners, who worked closely with venture firm Kleiner Perkins and Meta CEO Mark Zuckerberg. McNamee began biting the hand that fed him a decade ago, warning about social media most pointedly in his book Zucked: Waking Up to the Facebook Catastrophe, published in 2019. True, Meta's stock has risen over 300% since then, but McNamee was calling out societal harm, which he argues hasn't diminished. (McNamee is an executive producer of a new movie about Facebook, The Social Reckoning , in which Bill Burr plays a character loosely based on him.)

McNamee has turned his attention to the business of AI, which he seems to think is a kind of delirium "balanced on the head of a pin": "The pin is this notion that people are going to deploy AI widely, they're going to pay a ton of money for it, and that companies are going to get the cost structure down so it generates enough cash flow to service their debt," he says. "That's a lot of dependencies."

Anthropic's IPO is "completely ludicrous" and "the triumph of Wall Street fees over critical thinking," he says. "Remember, there are 10 other companies selling this stuff. A $1 trillion or $2 trillion valuation for a company with this revenue is ridiculous." (Anthropic's revenue grew 12-fold in 2025 to nearly $4.6 billion.)

Ah yes, Anthropic's math. Reuters reports that the company a) lost $42 billion last year, b) expects to spend $518 billion on cloud computing and infrastructure obligations in coming years, and c) warns that its business could cause "catastrophic or existential risks to humanity."

Sheesh. Remember when we thought tobacco companies were the bad guys? More to the point, who in their right mind would dip their beak here? Sure, early investors like former Google CEO Eric Schmidt might like you to buy their pre-IPO Anthropic shares-Schmidt bought in May 2021 at a $461 million valuation, according to PitchBook, which, at that $2 trillion valuation, would give him a dizzying 4,338-fold gain-but why oblige him? As Barron's recently noted, "Anthropic's IPO looks like a risky bet, at best."

What about the AI bubble writ large?

"Ordinarily, you'd have expected the government to intercede. Now you're counting on the bond market," says McNamee, referring to the deteriorating credit ratings and bond prices of some AI players such as Oracle, Akamai Technologies, and CoreWeave.

"AI itself is not going to go bust," says Hinton. "It works, just like railways worked. But everybody thought you could get rich quick by investing in railways, and there was a bust. That often happens with new technologies. Will we see a bust? I don't think it's impossible. It is a large fraction of the U.S. stock market now. People are piling in."

To paraphrase Sen. Everett Dirksen's reputed observation: "A trillion here, a trillion there, and pretty soon you're talking real money."

The money here is real. The math, not so much.

Endnote: Thanks to all of you who wrote regarding my column from last week asking which stock, Apple, Berkshire Hathaway, or Walmart, would do best over the next five years. The tally was pretty much split three ways, though a few folks voted to own all three.