Anthropic’s IPO prospectus, reviewed by Reuters and reported by CNBC, shows a public offering that could value the company at more than $2 trillion, more than double the $965 billion valuation it carried in a private round just four months ago. The same filing discloses a $42 billion net loss for 2025 and a plan to commit $518 billion to cloud, computing and infrastructure obligations over the coming years.
Most of that loss is not cash burn. Roughly $34 billion of it is an accounting charge tied to financing instruments that could eventually convert into Anthropic shares, separate from money the company actually spent running the business. Strip that out and the operating loss was still more than $8 billion, on revenue that grew twelvefold in 2025 to nearly $4.6 billion, according to the filing.
The filing is also the first place Anthropic has had to say, in writing and under legal exposure, how concentrated its customer base is. Nearly a quarter of last year’s revenue came from just two customers. Anthropic’s own risk-factor disclosure adds that most large clients could walk away at any time, since few of them are bound by multi-year agreements. That customer profile carries more risk than a typical enterprise software business with diversified, contracted revenue, especially for a company asking public shareholders to help fund a $518 billion infrastructure commitment.
Compute is where the money actually went. The $7.33 billion Anthropic put into compute and infrastructure during 2025 was three times what it spent in 2024, and it swallowed more than half of the company’s $12.65 billion operating-expense total for the year. Cash and short-term investments stood at $20.28 billion as of December 31, a cushion against a spending plan roughly 25 times that size.
The timing is not accidental. Anthropic shipped Opus 5.5 last week to answer OpenAI’s GPT-6 Astra, even as CEO Dario Amodei has separately urged the wider industry to ease off how quickly it ships new capabilities. That call sits awkwardly next to his own company’s race toward the same public markets OpenAI is chasing. OpenAI filed confidentially for its IPO in June, and media reports point to a listing sometime in early 2027. Whichever lab prices first sets the valuation benchmark the other has to justify itself against, which is why Anthropic’s listing, expected to land after the November midterms according to earlier Reuters reporting, matters beyond its own balance sheet.
A recent precedent is worth watching here. SpaceX’s June IPO valued Elon Musk’s company at $1.77 trillion, and its shares jumped 19 percent on debut to $160. They now trade around $147, still above the $135 IPO price but well off the opening pop. If Anthropic’s debut follows a similar arc, the gap between first-day enthusiasm and where the stock settles will tell investors more about durable AI valuations than the $2 trillion figure itself.
Anthropic declined to comment on the prospectus. The company has also been sparring with the White House: the Pentagon temporarily blacklisted Anthropic’s tools over a policy dispute, a move a US judge blocked in August, a reminder that the company’s biggest customer risk isn’t only concentrated commercial accounts, it also includes its own government relationships.
For enterprise buyers who signed Claude contracts this year, the prospectus is the first real look at whether the vendor behind those tools is pricing them sustainably or subsidizing them with investor cash ahead of a public offering.
Reported by Reuters, via CNBC, on 28 September 2026.
