Anthropic goes public on Wednesday, and the stock market is expected to make the AI lab the newest trillion-dollar company — even as tech stocks slide, one generative AI ETF has shed 35 percent of its value since June, and bears at Vox Media’s reporting say the AI bubble has already burst.
The IPO lands amid a perfect storm of political, financial, and leadership shocks.
This was supposed to be the boom's victory lap. A seven-person startup becoming a world-conquering juggernaut in five years, then cashing out at a trillion on Wall Street? That’s the closing scene of the feel-good movie.
Instead, the credits are rolling over a market in retreat, an interim CEO cleaning up a scandal, and a Congress that just campaigned against the very data centers this IPO is meant to fund. Vox Media, which reported the story this week, frames it bluntly: the AI industry’s charmed childhood is over, and its troubled adolescence has begun.
IPOs mint trillion-dollar companies
- Anthropic’s Wednesday IPO is expected to mint a new trillion-dollar company despite a months-long tech sell-off.
- Nvidia CEO Jensen Huang has resigned amid a scandal over an extramarital affair with a potential Chinese intelligence asset.
- AI spending is projected at $581 billion this year — about 1.8 percent of US GDP — while sector revenues sit near $540 billion, and Bain & Company estimates $6 trillion in annual revenue is needed by 2031 to justify the buildout.
The $6 trillion question hanging over Claude’s IPO
Start with the math, because the math is the story. According to Goldman Sachs, US investment in AI infrastructure will hit $581 billion this year — roughly 1.8 percent of the entire economy. Bain & Company estimates the industry needs $6 trillion in annual revenues by 2031 to justify that spending. Current revenues, per an estimate from Grandview Research cited by Vox, are closer to $540 billion.
For comparison, Vox notes that at the peak of the dot-com boom in 2000, telecom and fiber spending totaled just 1.15 percent of GDP. The railroad buildout of the late 19th century is the closer analog — and both of those booms, the outlet reminds us, ended in tears. Investors weren’t wrong that trains and the internet would change the world. They were wrong about how soon the money would arrive.
But the gap between spending and revenue isn’t even the sharpest blade in this drawer. The real threat to trillion-dollar valuations is cheaper competition. Vox reports that the Chinese open-weight model Kimi K3.1 Pro costs $10 per million tokens of output, while OpenAI’s most advanced model, GPT-6 Astra, runs $50. When the discount option is “good enough,” the premium product has a problem.
Brent Donnelly, a former HSBC trader and president of Spectra Markets, put the shift in trader’s terms in an interview with Vox. “There was an all-you-can-eat buffet approach initially where token budgets weren’t really a huge consideration,” he said. “But now that it’s not free, people are looking for cheaper options.”
And it’s not just individual users pinching tokens. A September survey of 200 enterprises by Futurum Research found that 53 percent had partially substituted open-weight models for frontier ones. A 2026 NBER study adds a grimmer stat: just 2 percent of firms account for over half of AI enterprise spending, even as artificial intelligence has delivered large productivity gains in the software industry.
Chips fund the AI buildout
Here is the kicker. Nvidia has functioned as what Vox calls the AI sector’s mini-Federal Reserve — a funder and financial backstop for the whole buildout. As of this August, the chipmaker had invested $70 billion in AI companies and extended roughly $230 billion in financial guarantees supporting data center expansion, including a promise to buy up to $6.3 billion of unsold capacity from CoreWeave.
That machine ran beautifully until last week, when longtime CEO Jensen Huang abruptly resigned. Vox reports the affair was with a potential Chinese spy and proved politically fatal. Nvidia, which has separately extended roughly $230 billion in financial guarantees to support the data center expansion, is now led by Colette Kress, and with her at the helm the AI industry may find funding harder to come by.
“This is a classic kind of black swan event. A CEO has an affair with a potential Chinese spy and then all of a sudden people notice there were a lot of cockroaches underneath the floorboards,” Ryan Cummings, a former staff economist for the White House Council of Economic Advisers, told Vox.
Restrictions win data center votes
Meanwhile, the politics turned. In the 41 jurisdictions where data center development was on the ballot, the side favoring restrictions prevailed in 40, Vox reports.
AI correction hurts Hollywood
Hyperscalers like Amazon and Google are diversified tech giants that own the data centers.
One more wrinkle:
The boom survives — its honeymoon doesn’t
To be clear, and Vox is careful about this: “burst” is premature. The Nasdaq remains more than 4 percent higher than when the year began, and Anthropic’s models keep getting better at a startling clip.
But Donnelly’s warning hangs over Wednesday’s celebration: “Once the fever breaks, it never really comes back,” he told Vox. “People aren’t going to be as optimistic now that the cracks have begun to show.”
The trillion-dollar birthday party is still on. The question the guest list should be asking is who’s paying for the next round — and whether, this time, anyone believes the bill.
Related reading