Anthropic just took its first formal step toward going public. The AI model maker has filed confidentially for an IPO, and co-founder Daniela Amodei used the Bloomberg Tech conference on Thursday to explain why, according to TechCrunch AI. Her answer was blunt: building frontier AI costs a fortune, and the public market is where that kind of money lives.
“It’s a really big upfront cost to train the models and to serve inference on them,” Amodei said, as reported by TechCrunch AI. “My guess is that over time, the sort of core set of companies that are working to advance the frontier are just going to need access to capital, and I think the public market is very well suited to that.”
What stands out here is the timing. Anthropic isn’t going public because private money dried up. It’s doing it while private demand is red-hot.
The numbers behind the move
The context makes this filing more interesting, not less. Multiple investors told TechCrunch AI that Anthropic’s $65 billion raise at a $965 billion valuation, announced last week, was heavily oversubscribed. Investors are lining up.
And the growth is hard to argue with:
- Annualized revenue crossed $47 billion in May
- That’s up from roughly $9 billion at the end of 2025
- The $965 billion valuation puts it within striking distance of the trillion-dollar club
A jump from $9 billion to $47 billion in under half a year is the kind of curve that makes an IPO an easy sell. Filing confidentially lets Anthropic prep the paperwork and test the waters with regulators before committing to a public timeline.
The doubt she’s brushing off
Not everyone is convinced the AI spending boom pays off. Companies like Uber have said that while AI can deliver returns, not all of their AI budgets have proven productive. That’s the quiet worry hanging over the whole sector: if big customers decide their AI spend isn’t earning its keep, they pull back, and the growth story stalls.
Amodei isn’t buying it. She thinks businesses are still early in learning how to actually use these tools.
“The use cases today, I expect will continue to be the primary driver of efficiency or creativity, whether that’s coding, financial services, legal, [or] health care,” she said, per TechCrunch AI. “But as the business community gets more familiar with the tools, we’re all going to learn together.”
Her bet is that value realized goes up as companies get fluent, not down.
Why Anthropic won’t build its own data centers
Amodei also drew a clear line between Anthropic and rivals like OpenAI and Elon Musk’s xAI, both of which are pouring money into their own data centers. Anthropic is taking the opposite stance.
“Anthropic’s view has always been wanting to plan for the best outcome but not overextend ourselves such that we’re buying more compute than we could productively use,” she said. The company would rather have slightly more demand than it can serve than the reverse.
That caution has a price tag. Last month Anthropic surprised the industry by partnering with xAI for compute, a deal disclosed in SpaceX’s S-1 filing at a cost of $1.25 billion per month. So instead of sinking capital into concrete and chips, Anthropic is renting capacity, even from a competitor.
Why it matters
This is significant because Anthropic going public would hand investors a rare, direct way to bet on a pure-play frontier AI lab. OpenAI is still structured around its nonprofit parent. xAI is private and tangled up with Musk’s other ventures. A listed Anthropic becomes a public benchmark for whether the AI buildout actually returns cash.
For practitioners, watch two things. First, an S-1 eventually means real financials: gross margins, inference costs, customer concentration. That’s data the whole industry has been guessing at. Second, Anthropic’s rent-don’t-build compute strategy is now a public test case against the build-it-yourself camp.
A confidential filing isn’t a listing date. But the direction is set, and the disclosures that follow could reshape how everyone values AI companies. More details are available at the original TechCrunch AI report.