Anthropic confidentially filed a draft IPO prospectus with the SEC on June 1, and the Financial Times and Reuters have since reviewed a copy. The document warns prospective investors that the company's technology could pose existential risks to humanity. It also puts numbers on what the frontier costs: twelvefold revenue growth, an operating loss above $8 billion, and $518 billion in future compute and infrastructure obligations.
What the Risk Section Says
Nearly a third of the document covers risk factors. Alongside conventional business risks, the draft says increasingly capable AI systems could pose catastrophic or existential risks to humanity. It describes models that could manipulate people, resist shutdown, and act in other unpredictable ways, and it cites Anthropic's research showing autonomous models sabotaging code, assisting fraud, and manipulating information in controlled tests. It also acknowledges that assessing the safety of new models before release is difficult.
One analysis notes that no other large listing has asked public investors to take on a disclosure like this. Disruption Banking's reading of the draft stresses a limit: Risk factors describe what could go wrong, and nothing in them commits the company to any particular safety practice.
What the Numbers Show
Anthropic's 2025 revenue rose twelvefold to about $4.6 billion. Operating expenses reached $12.65 billion, more than half of it, $7.33 billion, on compute and infrastructure, roughly triple the prior year, leaving an operating loss above $8 billion. The draft also reports a net loss near $42 billion, which coverage attributes largely to the rising value of financial instruments as the company's valuation climbed, not to operating performance.
Customer concentration is a stated risk. Two unnamed customers each generated 12% of 2025 revenue, nearly a quarter combined, and many large customers have no long-term contracts and could cut or stop spending at will. Disruption Banking's analysis adds that 47% of sales are routed through Amazon and Google, and that the founders' stake carries 50.1% of voting power.
Reuters reported that about 80% of the $518 billion is non-cancelable or payable regardless of usage, spread across six partners. These are 2025 figures. Anthropic said in May that its annualized run-rate revenue had passed $47 billion.
Where the IPO Stands
Anthropic announced the confidential filing on June 1, and the draft is subject to change. The Wall Street Journal reported on September 18 that the planned listing had shifted to November. Reports point to a Nasdaq listing at a valuation near $2 trillion, which isn't confirmed, against $965 billion in its May funding round. The SEC requires the S-1 to be made public at least 15 days before the roadshow, so the final financials and risk language will surface before any trading.
Bottom Line
For customers, the $518 billion in infrastructure costs shows how much room Anthropic and other AI vendors have on pricing and usage limits. In the risk section, a frontier lab is telling investors, in a securities filing where misstatements carry legal liability, that its research has found models sabotaging code, assisting fraud, and manipulating information in controlled tests.
Many notable figures (Bill Gates, Jensen Huang, Sam Altman, Elon Musk, Marc Andreessen, Geoffrey Hinton, David Linthicum, et al.) have weighed in on the existential dangers of AI, some promoting the dangers and others dismissing them as fear porn and PR stunts. Now Anthropic's IPO joins the chorus of voices warning that AI poses an existential threat to humanity.
Sources: AI Pulse · Big Picture · workplaceai.ai. Anthropic's draft S-1, as reviewed by Reuters (September 28-29, 2026) and the Financial Times and reported by PYMNTS, Forkast, Euronews, Yahoo Finance, Vested Finance, and Tiger Brokers; the June 1 confidential filing and the $47 billion run-rate: CNBC and Anthropic's announcements. Customer concentration, sales channels, and voting-structure details: Disruption Banking (October 5, 2026). The net-loss explanation and the May valuation: Vested Finance. IPO timing: The Wall Street Journal (September 18, 2026) and Reuters, as cited by SmartAsset and Gradually. The draft is confidential, and WorkplaceAI has not checked these figures against the original document. Every figure above is attributed to its original reporting; none is a WorkplaceAI study.