A $42 billion loss would sink most companies. For Anthropic, it is the opening line of an IPO pitch. The maker of Claude is reportedly preparing to go public, according to Reuters, which reviewed a prospectus not yet filed as a formal regulatory document. If it clears the finish line, this offering would hand public markets their first major benchmark for what AI data centers actually costs to build and run.
A $42 Billion Loss That Isn’t Quite What It Sounds Like
The headline number is real, but roughly $34 billion of it reflects an accounting charge rather than cash spent running the business.
According to Reuters, the vast majority of Anthropic’s reported net loss stems from an accounting charge tied to financing instruments that could eventually convert into company shares. That is a change in estimated value on paper, not cash burned on servers or salaries.
Strip that out and you arrive at the figure that more directly describes the business: an operating loss exceeding $8 billion in 2025. If you are sizing up this stock, those are two very different numbers, and conflating them obscures the actual financial picture.
The Real Cost of Running a Frontier AI
Anthropic’s computing bill alone exceeded half of its total operating expenses in 2025, even as revenue surged.
Anthropic reportedly spent approximately $7.33 billion on computing and infrastructure in 2025, roughly three times what it spent the year before, as reported by Reuters. That single line item represented more than half of its $12.65 billion in total operating expenses.
Revenue tells a compelling counter-story: nearly $4.6 billion in 2025, a twelvefold increase over the prior year, according to Reuters. Demand is real. The unit economics are still being worked out.
The prospectus reportedly also discloses approximately $518 billion in planned cloud, computing, and infrastructure commitments over coming years. That figure reflects future contractual exposure, not money already spent, and whether those obligations are fully incurred or eventually renegotiated remains unconfirmed.
What Public Markets Are Actually Being Asked to Believe
Anthropic is asking investors to look past near-term losses and accept a long-horizon argument about AI’s economic transformation.
Anthropic’s prospectus reportedly frames AI as a civilizational shift comparable to electrification and the internet, according to Reuters. That is Anthropic’s own characterization, not an independent assessment, and it carries the weight of the entire investment case because the near-term numbers do not close on their own.
The reported target valuation exceeds $2 trillion, though final terms, share price, exchange listing, and ticker remain unconfirmed pending a formal regulatory filing. Timing is reportedly planned for after the November U.S. midterm elections, subject to market conditions.
The core tension for prospective investors is straightforward. Revenue grew twelvefold, but compute spending grew roughly threefold in a single year on a much larger base. Rapid growth does not automatically trace a path to profitability when the infrastructure bill scales alongside it.
Public markets are built on quarterly scrutiny. Anthropic is asking them to extend the time horizon considerably. The argument: whoever builds the best frontier AI now collects the returns later.
An Anthropic debut would set the public-market price for that bet, and the reception from investors could influence how the broader AI sector approaches spending and fundraising for years to come.
