Anthropic, the maker of the Claude AI model, reported a loss of approximately $42 billion in 2025 on revenue of $4.59 billion, according to its IPO prospectus. The company has confidentially submitted a draft registration with regulators but has not publicly filed for an initial public offering.
The filing also shows Anthropic has signed up for $518 billion in computing commitments, a staggering figure that underscores the capital intensity of training and running frontier AI models.
The scale of the loss
A $42 billion loss on $4.59 billion in revenue is not a typical pre-IPO financial profile. For context, the loss is roughly nine times the company's annual revenue. That gap reflects the enormous cost of compute — the chips, data centers, and energy required to build and operate models like Claude.
The prospectus doesn't break down exactly where the $42 billion went, but the $518 billion in computing commitments offers a clue. Those commitments are contractual obligations to purchase computing capacity, likely spread over multiple years. They represent a bet that future revenue will grow fast enough to justify the spending.
Confidential filing, public silence
Anthropic has not publicly filed its S-1, the document that would give outside investors a detailed look at its finances, risks, and growth strategy. Instead, it submitted a draft confidentially, a route that lets companies test the waters with regulators before committing to a public process.
That means the $42 billion loss and the $518 billion commitment figure are among the few hard numbers to emerge so far. The company hasn't said when it plans to go public, how much it hopes to raise, or how it would use the proceeds. It also hasn't commented publicly on the prospectus details.
What the computing commitments mean
The $518 billion in computing commitments is the kind of number that raises immediate questions about financing. Anthropic doesn't have $518 billion in cash. The commitments are likely multi-year obligations tied to cloud providers and chip suppliers, and they'll need to be serviced whether or not revenue scales as hoped.
For a company that lost $42 billion in a single year, that's a heavy load. It also explains why an IPO might be attractive: public markets could provide a new source of capital to fund those commitments, and a public currency — stock — could be used for acquisitions or partnerships.
The Claude factor
Anthropic's core product is Claude, a family of AI models that compete with offerings from other major AI labs. The company has positioned Claude as a safer, more reliable alternative for enterprise customers, but the market is crowded and pricing pressure is intense.
Revenue of $4.59 billion shows there is real demand. The loss shows that demand isn't yet profitable at this scale. Whether Anthropic can close that gap depends on whether it can grow revenue faster than its computing costs — and whether the $518 billion in commitments turns out to be a smart bet or an anchor.
For now, the confidential filing means the public won't see the full picture until Anthropic decides to make it public. Until then, the prospectus numbers stand as the clearest signal yet of just how expensive it is to compete at the frontier of AI.
