Anthropic's IPO prospectus reveals that its revenue grew twelvefold in 2025 to $4.6 billion, but its operating loss widened to $8.06 billion. The filing also warns that its AI technology could pose "existential risks to humanity." Backers are targeting a valuation above $2 trillion, and the IPO is expected to set a benchmark for the AI industry.
SynthePulse Insight · AI deep reading
Version 1 · 1 source
An S-1 prospectus that has not yet been made public lays out two of Anthropic's curves at once: revenue grew twelvefold in 2025 to nearly $4.6 billion, while operating losses widened from $2.98 billion to $8.06 billion; at the same time, investors are pointing to a target valuation above $2 trillion, even as the company itself warns investors that its technology could pose a "human existential risk."
Anthropic has sent an S-1 prospectus to a small group of partners, formally beginning preparations for a public listing. The document is not currently public; the figures the outside world is seeing come from the Financial Times and Reuters reviewing and relaying the prospectus, which THE DECODER then compiled. Therefore, the key figures cited in this article—revenue, losses, valuation, and so on—are all secondhand relayings, lacking direct corroboration from a first-party document.
The first set of facts the prospectus presents is that growth and losses are expanding at the same time. According to Reuters, Anthropic's 2025 revenue grew twelvefold to nearly $4.6 billion; over the same period, operating losses widened from $2.98 billion to $8.06 billion. The revenue curve points up, and the loss curve points up as well, with the latter's absolute increase being larger.
The second set of facts explains why the losses widened. Anthropic's spending on compute and infrastructure reached $7.33 billion, three times the previous year and more than half of total operating costs. In other words, the company put more than half of its operating costs into the underlying resources that support model capabilities, rather than conventional expenses that can be amortized linearly as revenue grows.
If you look only at the net loss, the figure is even more striking: about $42 billion. But according to Reuters, about $34 billion of that comes from an accounting charge, reflecting an increase in the valuation of financing that can later be converted into stock, and is not an actual operating cash outlay.
This distinction is crucial to understanding Anthropic's financial state. An accounting charge and an operating loss are two different things: the former is triggered by a rise in financing valuation, while the latter is driven by actual inputs such as compute and infrastructure. The prospectus discloses both lines at the same time, and readers need to look at them separately, rather than equating the $42 billion directly with "money burned in a year."
It should be noted that the specific composition and subsequent impact of the $34 billion accounting charge are not explained in detail in the available information. Whether it will change again as the valuation fluctuates also cannot be judged from the current materials.
The most unusual aspect of this prospectus is that Anthropic warns investors in the document that its own technology could pose a "human existential risk," and mentions that models could manipulate, extort, or act in unpredictable ways. According to the Financial Times, nearly one-third of this lengthy document concerns risk factors.
For a company to list extreme negative scenarios for its own products as a risk in a listing document is still formally a standard information-disclosure action, but the content points to the technology itself rather than to external competition or regulation. It also constitutes a kind of self-positioning: Anthropic is both selling model capabilities and reminding investors of the uncertainty of those capabilities.
The prospectus also states that Anthropic expects AI's reshaping of the global economy to be more profound than industrialization, electricity, and the internet. This statement is the company's own judgment, not a verifiable established fact, and should be regarded as a source-party claim.
On revenue structure, Anthropic warns that in 2025 just two customers contributed nearly a quarter of revenue, and that many large customers have not signed long-term contracts. This means revenue concentration is high, and the continuity of a considerable portion of revenue lacks contractual lock-in.
This risk stands in contrast to the rigidity of compute spending: on one side are total cloud, compute, and infrastructure commitment expenditures of $518 billion over the coming years, and on the other side are large customers who may adjust their purchases at any time. Committed spending is long-term and rigid, while the revenue side is relatively dispersed and lacks long-term contract guarantees.
The available materials do not disclose the specific identities of these two customers, nor do they explain their contract details and stability. Therefore, the extent to which revenue concentration will translate into operating risk can currently only be retained as an uncertainty.
According to the Financial Times, investors believe a valuation above $2 trillion is possible. This figure would be more than double the $965 billion valuation in May, and would also exceed SpaceX's market value of about $1.8 trillion at its June listing.
It must be made clear that $2 trillion is an investor expectation, not a confirmed offering price. Whether the final IPO valuation can reach this level cannot be answered by the available information.
The timeline is likewise uncertain. According to Reuters sources, Anthropic's listing may not take place until November, after the U.S. midterm elections. This timing comes from unnamed sources and may change due to market or regulatory factors. By comparison, competitor OpenAI confidentially filed for an IPO in June.
Beyond the prospectus, there is another set of figures pointing to the future. According to the Financial Times, Anthropic generated $11.5 billion in revenue in the second quarter of 2026 and is expected to achieve adjusted operating profitability for a second consecutive quarter.
If this figure holds, it would mean an order-of-magnitude jump in revenue scale compared with nearly $4.6 billion for all of 2025, and the loss structure may be improving. But two points need attention: first, this data comes from a Financial Times report and has not been officially confirmed by the company; second, "adjusted operating profitability" and operating profitability are not the same measure, and the adjustment items are not explained in the available materials.
Therefore, whether the second quarter of 2026 can become a true profitability turning point, and whether this adjusted profitability can be sustained, remain unclear at present.
According to Reuters, analysts expect this first AI company to go public to set a valuation benchmark for the entire industry. The logic behind this judgment is that, before Anthropic, leading AI companies lacked public-market pricing, and there was no direct reference between primary-market valuations and secondary-market pricing.
But "setting a benchmark" is itself an analyst expectation, not a fact that has already occurred. Whether an AI industry valuation benchmark will truly be established by this depends on Anthropic's final offering price, its post-listing performance, and the subsequent moves of competitors such as OpenAI.
Looking at confirmed facts, source-party claims, and inferences separately, what this prospectus currently provides is a portrait of a company with high growth, high investment, high concentration, and a proactive disclosure of extreme risks; as for whether it is worth $2 trillion, the materials themselves do not provide an answer.
All information in this article comes from a single source article (THE DECODER's relay of reports by the Financial Times and Reuters), and the original prospectus is not provided. Key figures such as revenue, losses, and valuation are all secondhand relayings, lacking direct corroboration from a first-party document; the Q2 2026 revenue and adjusted operating profitability are from a Financial Times report and have not been officially confirmed by the company; the November listing timing comes from unnamed Reuters sources; the target valuation above $2 trillion is an investor expectation, not a confirmed offering price. The article does not disclose the full prospectus, specific customer names, or contract details.
Anthropic's S-1 prospectus shows two things at once: the revenue of a leading AI company can grow twelvefold in a single year, but the compute investment supporting that growth is also multiplying, and revenue is highly concentrated among a few customers. Writing "human existential risk" into the risk factors is both a disclosure and a positioning. As for the $2 trillion valuation and the November listing timeline, both are currently expectations and unnamed sources, not confirmed facts.
Primary report
Primary source