The AI Industry Is Running Out of Time

Why the sudden rush, guys?

Hey, Alberto here! 👋 Each week, I publish long-form AI analysis covering culture, philosophy, and business for The Algorithmic Bridge. Paid subscribers also get Monday how-to guides and Friday news commentary. I publish occasional extra articles.

I have an IMPORTANT REMINDER for you.

I’m running a 50% discount offer on annual subs that ends on June 5th (tomorrow). Get yours now and enjoy the discount forever.

If you read the free edition and have been thinking about subscribing, this is the best time. I won’t do another offer anytime soon.

Vanité by Philippe de Champaigne, 1646

AGI is Silicon Valley’s favorite three-letter acronym, but there’s a rising contender. Earlier this week, Anthropic announced in a blog post that it has confidentially submitted a draft S-1 statement to the Securities and Exchange Commission. Once it’s reviewed, they will have the option to launch an IPO (initial public offering). Their reported annualized revenue run rate is nearing $50B, and they’re on track for the first profitable quarter ever. Pretty good. OpenAI is closely trailing them. Revenue is lower, and there’s no profitability in sight yet—the consumer market is less valuable than the enterprise market—but they’re pretty much equal. Indeed: they’re both bleeding cash.

Bleeding is not a sustainable business model at their size anymore. One can be unprofitable for a while, and no one will bat an eye—looking at you, Uber—but if you require a planet’s worth of energy and cash injections to run, then we will intensely bat our eyes. At some point, either you figure out how to survive or the economy dies. Well, they figured it out by transferring the risk to the economy. They did become “too big to fail.” Now if they die, we die with them. The timing—non-profitability, exorbitant valuations, huge spending commitments without enough revenue, huge CapEx without returns—raises questions. One might think OpenAI and Anthropic are rushing to go public—that the race was never toward AGI but against the clock.

The cynical view, which I don’t hold but Bank of America’s chief strategist Michael Hartnett does, is that this double IPO story—triple if you count SpaceX, which acquired xAI recently—is less a public-market opportunity than a way for early investors to hand years of accumulated risk to the public market. They’re racing, yes, but not one another; they’re racing against us. The IPO is how insiders will exit the AI bubble. “The pricking is the converting of wealth into money,” says Bridgewater founder Ray Dalio. You thought a bubble meant they’d lose? That’s not how it works: if no bubble, they’d do business as usual, getting rich eventually. If bubble, they’d exit, getting rich now. So even a bubble is bad for us: they’re passing AI’s excessive valuation on to pension funds, index funds, and the broader economy.

If the market realizes that the bubble was actually a bubble and not exaggerated anti-hype before the IPO, Anthropic and OpenAI’s chances to cash out at the peak get shattered. If it happens afterward, however, the problems are redistributed to the broader economy. Privatize the gains, socialize the losses, or something to that effect. If the gap between promise and delivery is fundamental rather than temporary, it’s you and me, normal people, who are screwed.

And the gap is becoming too obvious to ignore: enterprises are losing trust. Just this month, Microsoft announced it’s cancelling Claude Code licenses in favor of an internal tool due to unsustainable costs. Uber is “limiting all employees to $1,500 in monthly token spending per AI coding tool.” Starbucks is retiring its AI tool across North America after 9 months due to reliability problems. Even Sam Altman has admitted that the AI costs have become “a huge issue.” We’re about to get inundated with stories like these because the problem is universal: there’s no clear relationship between AI spending and AI returns. Their IPO window isn’t closing because AGI is almost here, but because the story about “AGI almost here” has a fleeting shelf life.

Was this their plan all along? I don’t think so. I think Silicon Valley is echo-chambered enough that these guys’ beliefs about “superintelligence for 2027” are actually genuine. But so what? They are nonetheless discharging the risks onto the rest of us either way—we have all the reason to be mad.

But risk doesn’t mean danger and it’s only fair to acknowledge what happens if AI does work out. In that case, the IPO would extend what is now a private gain into a public gain. That’s how the world prospers, how the tide lifts all boats, even those that didn’t do any work at all. Unfortunately, as neat as it is that story about the wonders of capitalism, it brings me no consolation: I don’t trust that a technology this unreliable—we can say for sure that AI models won’t stop hallucinating—will pass the test of time as the internet or electricity did before.

This is, to me, the core issue here. They win in any case, either as enablers of a technological revolution or as wealthy shareholders. We only win if AI works out. It’s paradoxical to put it this way, but it’s us—you and me, normal people—whose future is intimately tied to the future of AI, not theirs.

But why won’t AI pass the test of time? Figuring innovation out into productivity and economic growth takes time, so why is AI different? To answer this, I will ask another question: what does a non-figure-out-able technology look like? I do know.