Anthropic races for record IPO while OpenAI delays
Two AI labs, the same week, two opposite bets on the public market. OpenAI CEO Sam Altman said on the record that going public now would be foolish, keeping his company private. Anthropic is racing, according to Reuters, toward the largest IPO ever, with chipmaker Nvidia as anchor investor. If you’re paying attention, there’s more here than two isolated news items.
Keep one distinction sharp, because the entire post hinges on it: one story is on the record, with a name attached. The other is reporting, solid but unconfirmed. Don’t weigh them equally.
Altman: now is an unwise moment
The hard facts first. In an interview with Fortune, recorded by Editor-in-Chief Alyson Shontell and published September 12, 2026, Altman said OpenAI will not go public this year. “We’re not rushing into an IPO”, and to the direct question of whether it happens in 2026: “I would say not 2026, yeah. We’ve got a lot of stuff to do.” His reasoning centers on safety and timing: going public only “when we’re ready, which is when the business is ready.” TechCrunch carried the remarks. OpenAI has filed a confidential IPO registration, but no valuation has been disclosed.
“Right now would be an ill-advised moment to go public.”
An unwise moment to go public, says the man leading one of the world’s most expensive AI companies. That is a striking sentence from someone already waiting in the IPO queue.
| Lab | IPO now? | What is said or reported | Source status |
|---|---|---|---|
| OpenAI | No, not in 2026 | Altman: “an ill-advised moment”; confidential filing submitted | on-record interview |
| Anthropic | Racing toward it | up to $100 billion, valuation around $2 trillion, Nvidia anchor for up to $10 billion | reporting (Reuters) |
Anthropic heads the other direction
Enter the counterpoint. According to an exclusive Reuters report (Krystal Hu and Milana Vinn), widely picked up by other outlets, Anthropic is in talks to bring Nvidia on as anchor investor: the heavyweight name that commits early and pulls other backers across the finish line. Nvidia would invest up to $10 billion. Anthropic aims to raise up to $100 billion at a valuation around $2 trillion, by the same reporting (some reports cite up to $2.3 trillion). One trillion is one thousand billion. It would be the largest IPO on record, expected before the November 2026 U.S. midterms, though one report suggests a possible shift to mid-October.
Red-pencil this heavily: this is reporting, not a signed deal. Nothing has been filed, we are talking about discussions, and both Anthropic and Nvidia declined to comment. The Decoder summarizes the Reuters story. Treat every number above as solid journalism, not established fact.
What explains the appetite comes from Anthropic itself, via Reuters: the revenue run rate, annualized if you extend the current pace one full year, rose from roughly $9 billion at the end of 2025 to more than $65 billion by late July 2026. Numbers like that make a record fundraise sound less absurd than it reads. And absurd it reads, because the previous record-holder cannot touch it.
The current IPO record belongs to Saudi oil giant Saudi Aramco in 2019, raising $29.4 billion. The Reuters target of $100 billion sits, for comparison, roughly 3.4 times higher.
There is one more piquant layer. Nvidia supplies Anthropic with the chips powering its compute, and would now also supply the capital for the IPO. The anchor investor is, in other words, the very supplier Anthropic depends on. You might call it a “compute landlord” loop: your infrastructure landlord becomes your shareholder. Money cycles in a tight, closed ring.
Watch the money flow, not the manifesto
Here is where it gets interesting. The same week this IPO news broke, Anthropic CEO Dario Amodei called for speed limits on AI (widely reported via TechCrunch, The Verge, and The Decoder on September 12, 2026). In July 2026, Anthropic signed onto a call for governments to slow AI development. At the same time, the same company is, per Reuters, racing through the largest capital round ever.
We have seen this pattern before. We wrote that Anthropic signed up for roughly half a trillion dollars in compute power, and yesterday that OpenAI wants to coordinate slowdown and asked Congress for antitrust guidance. The same split every time: the brake in words, the accelerator in numbers.
Our take: watch the money flow, not the manifesto. When a company asks for slowdown in the same week it races for the largest capital raise ever, its actions tell you what words conceal. A $100 billion IPO is not a pause button.
Now for the honest counterargument, because it truly exists. The Anthropic numbers are Reuters reporting, not a signed contract; they can shift or fall apart. An early IPO need not be cynical either: raising capital when the market is hot is simply smart timing, and someone has to fund compute for safe research. And Altman’s caution cuts both ways. It may be genuine, or it may simply be that OpenAI is not ready for the market yet.
Either way, the test stays the same: judge an AI lab by what it signs, not by what it writes. The manifesto is free. The signature on a prospectus costs, if Reuters is right, $2 trillion.
