Analysis
Anthropic's leaked IPO prospectus reportedly shows $4.6 billion in 2025 revenue against a $42 billion net loss, according to a Crunchbase News analysis of the filing. The company's IPO has already slipped from October to November.
Anthropic is still targeting a valuation of up to $2 trillion -- more than double the $965 billion private mark Pulse has tracked from this past May.
The prospectus also reportedly commits Anthropic to at least $518 billion in infrastructure spending over the next decade across six partners, with roughly 80% of that spending non-cancelable regardless of demand.
“A quarter of that 2025 revenue reportedly came from just two unnamed enterprise customers, a concentration level IPO investors typically price at a discount.”
## The revenue run-rate doesn't match the loss Noncash charges reportedly account for $34 billion of the $42 billion net loss, leaving roughly $8 billion in cash losses against $4.6 billion of revenue.
Separately reported Q3 2026 figures put quarterly revenue at $16.8 billion, up 45% sequentially, for an annualized run-rate near $67 billion -- a figure that's hard to reconcile with the $4.6 billion full-year 2025 number without unusually fast acceleration between periods.
A quarter of that 2025 revenue reportedly came from just two unnamed enterprise customers, a concentration level IPO investors typically price at a discount.
## OpenAI is pulling the opposite lever Where Anthropic presses ahead toward a bigger IPO target, OpenAI has gone the other way. CEO Sam Altman has called a 2026 listing "ill-advised" given AI-safety concerns, and the company confirmed it will not go public this year, instead closing a $30 billion private round.
That round itself was repriced down to a $1.25 trillion pre-money valuation from an earlier $1.4 trillion report about a month earlier -- the opposite trajectory from Anthropic's rising IPO target.
The bear case: analysis firm New Constructs has called the Anthropic offering "an unprecedented test of investor gullibility," estimating a $2 trillion valuation requires roughly $358 billion in annual after-tax operating profit by 2035.
Gartner, by contrast, forecasts AI data-center revenue topping out near $3.6 trillion through 2030, against an estimated $8.5 trillion needed to cover planned investment, financing costs and a reasonable return. Open-weight pricing adds another headwind: DeepSeek's R1 model reportedly prices around 7 cents per million input tokens, roughly 18 times cheaper than Anthropic's comparable offering, by one independent analysis.
None of that guarantees the IPO stumbles -- a $16.8 billion quarterly run-rate is real revenue. But the distance between a $965 billion private mark in May and a $2 trillion public target six months later is a gap that public investors, unlike VCs marking paper gains, will have to price with real money on day one.