Anthropic publicly filed for a US stock market listing on Monday, setting out plans to trade on the Nasdaq under the symbol ANT in what The Guardian reports is slated to be one of the largest debuts in US history. The filing, first submitted confidentially in June, gives investors their first audited view of the company behind the Claude chatbot, which analysts and investors have valued at more than $1.5tn, according to The Guardian.
The same document carries an unusual warning for a company asking the public for capital. According to BBC News, Anthropic said advanced AI systems "could, if not developed, deployed, and operated with appropriate safeguards, oversight, and security measures, pose catastrophic or existential risks to humanity, particularly if such systems were misused, replicated, weaponised, or allowed to operate with a high degree of autonomy". The company has not yet disclosed the number of shares to be released, a price range or a listing date, the BBC adds.
The revenue line and the loss beneath it
The financial disclosures show a business scaling quickly and spending faster. The Guardian reports revenue of $41m in 2023, $386m in 2024 and $4.585bn in 2025, an increase of more than 1,000%, followed by $16.286bn in the first six months of 2026 alone. The BBC puts that first-half figure at $16.3bn, with losses of more than $80.7bn. The Guardian reports a net loss of $80bn for the same period and notes that, as of the end of June, the company had raised $141.1bn. Anthropic's commercial model differs from that of its main rival. The Guardian notes that Anthropic has focused on businesses as its primary paying customers, unlike OpenAI, which has courted everyday consumers for ChatGPT. As of 30 September, Anthropic said it had more than 14,000 enterprise customers, of which 11,900 generated $100,000 or more in revenue, more than 1,900 generated $1m or more and about 250 generated $10m or more. That mix carries concentration risk. The Guardian reports that a quarter of Anthropic's revenue comes from its top five customers, though no single customer represents more than 10% of the business. The filing warns that a "substantial portion" of revenue comes from a limited number of customers and that losing any of them could put a huge dent in revenue.What the prospectus leaves blank
Wired reports that the S-1, though several hundred pages long, omits notable financial figures, largely around the most recent quarter ended 30 September and lease agreements tied to compute investments. It quotes the filing's own redacted sentence: "Claude has experienced the fastest enterprise adoption of any technology in history, as demonstrated by revenue growing from $0 at commercial launch in the first quarter of 2023 to $_____ for the three months ended September 30, 2026." Michael Dambra, an accounting and law professor at the University at Buffalo School of Management, told Wired there are "practical reasons" for the third-quarter numbers to be absent, since closing the books for a company of this size takes time, and that SEC rules would require an update with third-quarter financials if Anthropic lists later than about mid-November. Wired also reports Dambra's view that companies with redactions in their S-1s were more likely to underprice their shares at IPO. The compute obligations are where the redactions matter most. Wired reports that Anthropic describes compute as the "most significant operating cost in our business" and that the filing discloses future minimum payment obligations of approximately $417bn in aggregate on compute and infrastructure leases started or amended since 31 December 2025, an additional $245bn added since 30 June, and a further $125bn to $150bn tied to commercial agreements the company caveats "have not yet been finalized". Including short-term agreements with minimum non-cancelable payments of $21bn, Wired calculates a minimum of an estimated $808bn owed if Anthropic follows through on all of it.Safety language as a disclosure item
The risk section runs to more than 90 pages, according to both The Guardian and the BBC. The Guardian reports that Anthropic's executives set out four categories of potential risk: misuse by bad actors, loss of control, authoritarian abuse and economic disruption. The filing states that the company's most capable models have "capabilities that could provide meaningful assistance to malicious actors seeking to develop chemical, biological, or other dangerous weapons", and that its safeguards "are incomplete, may be outdated in light of evolving threat models, have been and may in the future again be circumvented, or may fail to operate as intended". The filing also discloses a specific security incident. The Guardian reports that in September 2025 staff documented what they believe was the first large-scale cyber-attack that did not rely on significant human intervention, with Anthropic assessing with high confidence that a Chinese state-sponsored hacking group manipulated its Claude Code tool into attempting infiltration into roughly 30 global targets, including large technology companies, financial institutions, chemical manufacturing companies and government agencies. There are commercial and political exposures beyond safety. The BBC reports that Anthropic warned it relies on a cluster of cloud computing providers, Taiwanese chipmakers already facing huge demand and data centres concentrated in the US. The Guardian reports that the company has incurred $6.2bn in debt as it built out infrastructure and spends billions yearly on computing power from third parties, and that it has been designated a supply chain risk by the Trump administration, a first for an American company, preventing any military agency from contracting with it. Anthropic has challenged the blacklisting.Why it matters
Anthropic will be the first fully AI-focused company to go public, the BBC reports, making the listing a test case for the technology and for the financial viability of the multitrillion-dollar AI boom. For enterprise buyers, the S-1 is the clearest signal yet of how the company's contracts are structured and how dependent its revenue is on a small group of large accounts. For investors, the gap between $16.3bn of first-half revenue and an $80.7bn loss defines the central question: whether forward compute commitments of the scale Wired describes can be converted into durable margin. Matt Orton, chief market strategist at Raymond James Investment, told the BBC that investors expected rapid growth alongside rapid financial losses, and that the filing would help provide visibility because "we need to know where its revenue growth is coming from". Peter Andersen, founder of Andersen Capital Management, told Reuters, as reported by the BBC, that the listing was "a significant milestone in the development of artificial intelligence" and that its outcome would be "fairly indicative of the sentiment" among investors towards continued AI spending.Risks and unknowns
The share count, price range and listing date remain undisclosed. Third-quarter financials and the terms of compute leases are redacted, and Wired reports that experts expect many details to be filled in before the company goes public. The BBC notes that OpenAI is expected to list early next year, according to a person familiar with its plans, which would test investor appetite across the sector rather than at a single company. The BBC also reports that on the day of the filing the owners of the rights to Elvis Presley's music sued Anthropic, accusing it of using hundreds of the star's songs to train Claude.FY Outlook
The next milestone is an amended filing with third-quarter figures and, in all likelihood, a price range. Wired's reporting suggests the redactions may themselves influence the opening price, since companies that withhold information often need to discount to attract buyers. The more consequential disclosure for operators will be the final terms of compute contracts running to 2030 and, in some cases, 2045. Until those are visible, the market is pricing a business whose revenue trajectory is clear and whose cost base is not.Sources and References
- The Guardian — theguardian.com
- BBC News — bbc.co.uk
- Wired — wired.com
