Anthropic has confidentially filed a draft initial public offering prospectus with the Securities and Exchange Commission, disclosing a net loss of nearly $42 billion for 2025 against revenue of $4.6 billion. The majority of this loss, approximately $34 billion, stems from non-cash accounting charges related to convertible financing whose value increased alongside the company's rising worth. Excluding these charges, the operating loss exceeded $8 billion, driven largely by compute and infrastructure spending that tripled to $7.33 billion. The filing reveals ambitious future commitments, including plans to spend about $518 billion on cloud computing and infrastructure in coming years, with roughly 80% of these contracts being non-cancelable. Major suppliers include Google, accounting for at least $111 billion, and Amazon for $110 billion, while SpaceX is contracted for $1.25 billion monthly through May 2029.
Despite holding $20.28 billion in cash at the end of 2025, Anthropic faces significant concentration risks, with nearly a quarter of its 2025 revenue derived from just two customers who lack long-term contracts. Revenue growth remains rapid, with second-quarter 2026 figures topping $11.5 billion, more than double the entire 2025 total. This follows a May fundraising round that valued the company at $965 billion, surpassing OpenAI’s last disclosed valuation. Backers now anticipate a public market debut after November’s U.S. midterm elections with a target valuation exceeding $2 trillion. The prospectus also includes extensive risk disclosures, warning that increasingly autonomous AI models could exhibit self-preserving behaviors, resist shutdowns, or assist in fraud, prompting CEO Dario Amodei to advocate for mandatory safety rules modeled on aviation oversight.
The disclosure highlights a stark divergence between accounting metrics and operational reality, where massive non-cash losses obscure an underlying business model heavily dependent on capital-intensive infrastructure commitments. While the $42 billion headline figure may alarm retail investors, the critical financial pressure lies in the $518 billion planned expenditure against a current cash position of $20.28 billion. This structural imbalance suggests that Anthropic’s path to profitability relies less on immediate revenue generation and more on securing sustained external capital or achieving exponential scale before contractual obligations mature. The reliance on non-cancelable cloud contracts with tech giants further entrenches the company within a specific ecosystem, potentially limiting flexibility if market dynamics shift or if regulatory pressures impact these partnerships.
From a regulatory and credibility standpoint, the explicit warnings regarding existential risks and model misbehavior represent a novel challenge for public market compliance. Unlike traditional tech firms, Anthropic is asking shareholders to accept liability for potential autonomous actions of its software, a concept without established precedent in securities law. The length of the risk section, nearly 80 pages compared to 48 for business operations, signals an attempt to preempt litigation by fully disclosing theoretical dangers. However, this transparency may also invite heightened scrutiny from regulators concerned about systemic risks posed by highly capable AI systems. Investors must weigh the promise of a $2 trillion valuation against the unproven nature of governing autonomous agents and the concentration risk inherent in a customer base where two entities drive a quarter of revenue.
