This article was featured in the Daily Prospect newsletter. Sign up for it here.
Anthropic set its initial public offering price at $83 a share on Thursday, and amid opposition protests in the streets, it opened with a big pop to $111.50. Bulls and bears whipsawed throughout the trading day, pushing the stock all the way under $93 before it rallied, subsided, and then closed at $97.17.
There was $500 billion in investor demand before the stock went on sale, and those with a prior relationship with Anthropic got first dibs. Indeed, every investor felt a pull to have a chunk of Anthropic in their portfolio. The Financial Times quoted some, who said uniformly that the stock is a “must own.” The logic is that joining everyone else by piling into something that underperforms is at least explainable, while missing out on the generational opportunity of buying into the first pure AI model maker to go public would be catastrophic. It’s the same logic animating the movement of Anthropic into market indexes and index funds.
The Prospect in your inbox
Analysis that goes beyond the noise with in-depth, progressive reporting.
No paywall, no programmatic ads — just the reporting. Unsubscribe anytime. Submitting your email signs you up for The Daily Prospect, our weekday morning newsletter.
This “there is no alternative” mindset means that all of the questions about the company—the enormous $855 billion in noncancelable commitments to buy compute and infrastructure, the fact that Anthropic’s creditors are also its suppliers, the revenues that have already slowed down when they need to explode to match these costs, the prices for its models becoming the cheapest they’ve ever been, the enormous litigation risk from copyright infringement and culpability in self-harm and suicide and rogue models breaking into computer systems and more, and that whole thing about destroying all humanity on Earth—must get ignored as more and more stock is hoarded.
As the American Economic Liberties Project has detailed, it’s appalling that the Securities and Exchange Commission has allowed thousands of people to get rich (every Anthropic employee is now a millionaire) when all of the above risks are hanging out there, to say nothing of the somewhat dodgy financing in its disclosures (Francine McKenna is absolutely the best read for that information) and the fact that insiders could under the agreement sell on the first day and fully exit before 90 days are up, a process that reeks of a pump-and-dump scheme. Democrats and public pension funds across the country tried to slow down the rush to IPO. But they were ignored, and once the course was set, the initial pop was going to happen.
Yet while Anthropic is now one of the top ten companies in the world by valuation, and its seven founders are now sitting on a stake worth over $24 billion each, the chart of the first day does not look like a no-brainer juggernaut, but a company that people are wary of rendering a verdict about. It’s normal first-day IPO territory, with a day-one gain of around 17 percent. It looks a lot like the SpaceX IPO, which was seen as successful until the fall from the initial pop just kept continuing: SpaceX is down more than 40 percent from its peak, a few days after its debut. And despite being the largest IPO in history, Anthropic didn’t raise all boats; the market still slid.
That implies something about the very different world this industry is now living in. Anthropic’s top brass kind of knew this, downshifting to a lower valuation from the lofty $2 trillion once imagined. As Bloomberg suggested, the Anthropic IPO could be the last gasp of the AI trade, a quick dopamine hit before it slides into crisis and (if we’re lucky) irrelevance. The founders and employees (and the bankers who set up the IPO) got their payday, a tidy sum based on extracting without compensation the collective knowledge of the world. But they may saddle the rest of us with the consequences.
You cannot talk about Anthropic’s uncertain opening without talking about the incredible drag on the AI trade from Nvidia CEO Jensen Huang’s spectacular fall, as the Prospect has covered relentlessly. An incredible stat: Half of the fall in the S&P 500 since its peak on October 6 can be attributed to Nvidia’s losses. And the company is—wait for it—about to invest $10 billion in Anthropic. This is part of Nvidia’s circular financing model that has made it the Bank of AI: It invests in companies that buy chips from itself.
In some ways, the losses at Nvidia aren’t where the bubble pop will come from. The companies that rely on Nvidia’s financing are at far more risk, as Gary Marcus notes. If Nvidia closes the checkbook, CoreWeave and other companies will fade away. A hot IPO is nice for distraction purposes, but it doesn’t eliminate this fundamental set of problems.
Neither does Anthropic’s big day wipe away its core problem, as I put it previously: The company doesn’t make enough money to cover its expenses. And that ties directly into how being a public company changes the corporate posture.
As Rep. Pat Ryan (D-NY) explained, a public Anthropic will feel the pressure to keep the valuation high. That means selling as many ads as possible, or entering partnerships with businesses to exploit consumers or replace workers, or cutting corners on safety and rushing out the next frontier model rather than pacing that frontier. Anthropic incorporated as a public benefit corporation, which is supposed to take all stakeholders into account. But that promise is in tension with executives whose net worth is explicitly tied to getting nods of approval from Wall Street. Public companies are under a microscope, the effects of which are seen every day in stock ticker information.
The Prospect in your inbox
Analysis that goes beyond the noise with in-depth, progressive reporting.
No paywall, no programmatic ads — just the reporting. Unsubscribe anytime. Submitting your email signs you up for The Daily Prospect, our weekday morning newsletter.
At some point, Anthropic may have to think about reducing expenses, and its biggest expense is chips. It can either stop buying so much from Nvidia or try to take the business in-house; either way, a stock that is already in freefall would suffer some more.
This is before getting into the changing regulatory environment given the election, what we might learn about potentially criminal smuggling of AI chips to China, and a hundred other factors. Investors can overlook a lot, especially if they’re getting rich in the process: Witness AI doomer Jacob Coxon, who famously warned about the end of humanity from model acceleration, pledging to keep his OpenAI equity in advance of a future IPO because “I like money, bro.” But as much as it seems like facts don’t matter on Wall Street and everyone with wealth and power muddles through, that’s true right up until the point that it isn’t. Think 1929 or 2008.
What lawmakers and regulators need to know is the extent of the tight coupling between AI companies, the commitments the major companies have with creditors, and the point at which the finances will begin to break. Preparations for a crisis moment will never go to waste, even if that day of reckoning is far off. And as part of that, those in charge of preventing an asset bubble from catastrophic harm should think about whether the people who laughed through the party until it all came to a halt should pay a price for what they have wrought.
