Anthropic's draft IPO prospectus, first reported Monday and dissected across financial media on Tuesday, gives investors their first detailed look at the finances of the artificial intelligence developer ahead of a possible public listing. The filing, which is not yet public, shows 2025 revenue of $4.59 billion, up roughly 12-fold, or 1,088%, from about $386 million in 2024. Enterprise adoption of its Claude models drove the surge.
Growth came with steep costs. The operating loss widened to $8.06 billion from $2.98 billion a year earlier, and the accounting net loss reached nearly $42 billion once writedowns of various liabilities, mostly tied to earlier fundraising, are included. Compute and infrastructure spending totaled $7.33 billion, or roughly $1.60 for every $1 of revenue.
The forward commitments are the headline risk. The prospectus outlines about $518 billion in future cloud, computing and infrastructure obligations. Customer concentration is another flag: two of the largest direct customers each generated 12% of 2025 revenue, and the company warned that many large clients are not locked into long-term contracts and could cut or stop spending. Chief Executive Dario Amodei reportedly devoted more than a third of the document to risk factors, including warnings about the potential for advanced models to behave unpredictably.
On valuation, the company is reportedly targeting more than $2 trillion in an offering that, after several delays, could come after the US midterm elections in November. Morgan Stanley (NYSE: MS) and Goldman Sachs (NYSE: GS) are reportedly in pole position to lead the deal. Anthropic said in April that annualized revenue had passed $30 billion, up from roughly $9 billion at the end of 2025.
The disclosures land amid heightened scrutiny of AI economics. Goldman Sachs analysts estimate that the major AI hyperscalers will need roughly $300 billion in revenue to cover their data center and chip spending, while noting that Amazon (NASDAQ: AMZN), Alphabet (NASDAQ: GOOGL) and Microsoft (NASDAQ: MSFT) have announced a combined revenue backlog of $1.7 trillion. If that backlog converts, the argument goes, the heavy investment is defensible. US equity futures wavered Tuesday as the leak and OpenAI's decision to shelve a next-generation model raised fresh questions about AI safety and spending.
For investors, the IPO is effectively a referendum on frontier AI economics: whether revenue growth can outrun compute commitments. Anthropic is private and cannot be traded, so the read-through falls on listed cloud and infrastructure providers, whose backlog and capital spending plans will be scrutinized more closely as the filing circulates. Traders should watch for an official public S-1, any revision to the timeline, and commentary on customer contracts.
Key uncertainties include the final valuation, the size of the raise, and whether the two largest customers renew. Until the filing becomes public, investors are working from leaked figures that may change before any listing.