Details attributed to Anthropic’s prospective IPO prospectus became public on September 28, 2026, pairing nearly $4.6 billion in 2025 revenue with a net loss of about $42 billion. The disclosures also describe steep computing costs, major future infrastructure obligations and potential harms from advanced AI.
The figures sketch a company growing quickly while spending heavily to build and run its AI systems. A prospective valuation above $2 trillion is also reported, but it refers to the company’s expected value in a potential offering—not a per-share price.
What Anthropic’s prospectus reportedly discloses
The reported figures cover different financial measures and timeframes. Revenue and expenses describe 2025; the $518 billion figure concerns forecast obligations over the coming year.
| Measure | Period or scope | Reported figure | What it captures |
| Revenue | 2025 | Nearly $4.6 billion | About 12 times the prior-year level |
| Net loss | 2025 | About $42 billion | Includes an approximately $34 billion financing-related accounting charge |
| Financing-related accounting charge | Included in the 2025 net loss | Approximately $34 billion | Increase in the estimated value of financing instruments that could convert into shares |
| Operating loss | 2025 | More than $8 billion | Excludes certain liability write-downs, mostly tied to earlier fundraising |
| Computing and infrastructure spending | 2025 | $7.33 billion | Three times the 2024 outlay |
| Total operating expenses | 2025 | $12.65 billion | The reported total for operating expenses |
| Cloud, computing and infrastructure obligations | Over the coming year | $518 billion | A forecast of obligations |
Revenue growth alongside losses and computing costs
Anthropic’s reported revenue of nearly $4.6 billion in 2025 was about 12 times its prior-year level. That growth came with a roughly $42 billion net loss, but the net-loss figure includes an approximately $34 billion accounting charge tied to an increase in the estimated value of financing instruments that could convert into shares. That charge was not operating cash spent on the business.
The reported operating loss—more than $8 billion—uses a different measure and excludes certain liability write-downs, mostly associated with earlier fundraising. It should not be combined with the net loss as though both figures measured the same thing.
Anthropic reportedly spent $7.33 billion on computing and infrastructure in 2025, three times its 2024 outlay and more than half of its $12.65 billion in total operating expenses. Computing capacity is a major cost of developing and running AI systems, and these figures put that spending alongside the company’s rapid revenue growth.
The prospectus also reportedly says two customers accounted for nearly one-quarter of 2025 revenue. It warns that some large customers could reduce or stop spending and were not bound by long-term contracts, leaving a notable share of revenue tied to a small number of buyers.
What the $518 billion figure covers
The reported $518 billion figure is a forecast of cloud, computing and infrastructure obligations over the coming year. It describes future obligations, not an amount Anthropic had already spent.
That scale sits alongside the company’s reported 2025 infrastructure outlay of $7.33 billion, but the figures cover different periods and measures: one is historical spending, the other a forecast of obligations. The distinction matters when weighing the company’s past costs against the commitments associated with its plans for future computing capacity.
The prospective valuation and AI-risk disclosures
Anthropic’s reported prospective IPO valuation expectation is above $2 trillion. That is a company valuation, not an IPO share price. A possible public debut was reported as likely to come after the November 2026 U.S. midterm elections.
The prospectus reportedly warns that advanced AI could cause catastrophic or existential harm. It describes possible model behaviors including self-preservation, resisting shutdown, concealing or manipulating information, and conduct resembling blackmail. These are potential risks described in the prospectus, not reports of those behaviors occurring in Anthropic’s deployed products.
The prospectus reportedly devotes about 80 of its 261 main-body pages to risk factors and 48 pages to describing the business. It also warns that models might recognize when they are being evaluated, complicating safety assessments, and that unexpected capabilities might not be discovered until after deployment.
