Anthropic's IPO filing reveals $518 billion in AI infrastructure commitments; majority cannot be cancelled.

Anthropic is wagering that AI will transform the global economy more profoundly than industrialization, electricity, and the internet combined, according to its IPO prospectus seen by Reuters. The cost to realize this vision is staggering: the company posted a net loss of US$42-billion in 2025 and plans to spend US$518-billion on cloud computing and infrastructure obligations in coming years.

The company has grown sharply while posting wider losses. Revenue surged 12-fold in 2025 to nearly US$4.6-billion, while operating losses exceeded US$8-billion when excluding roughly US$34-billion in accounting charges tied to financing that could eventually convert into equity rather than cash spent on operations. Compute and infrastructure spending reached US$7.33-billion last year—a threefold increase from 2024—accounting for more than half of total operating expenses of US$12.65-billion.

The public offering, expected to value Anthropic at more than US$2-trillion, would capitalize on the five-year-old AI startup's rapid rise and establish a valuation benchmark for the sector's leading players, including rival OpenAI. The prospectus reveals the company held cash, cash equivalents and short-term investments totaling US$20.28-billion as of December 31.

The company now faces a central contradiction. Anthropic's own research shows increasingly autonomous AI models can behave unexpectedly and harmfully—sabotaging code, assisting fraud, and manipulating information in controlled tests. These findings have entered public discourse, sparking concerns about whether companies can maintain control over increasingly powerful systems deployed commercially. CEO Dario Amodei has publicly called for the global AI community to slow the pace of capability releases to address these concerns. Yet last week, Anthropic rolled out its new Opus 5.5 model to counter OpenAI's momentum following GPT-6 Astra's launch, as the company races ahead of its expected IPO.

Anthropic warned investors that nearly a quarter of revenue came from two customers, and that many of its largest clients lacked long-term contracts and could reduce or cease spending. The company declined to comment on the filing.

Anthropic's expected valuation has more than doubled since May, when it was valued at US$965-billion. Reuters previously reported the IPO would likely occur after the November U.S. midterm elections, bringing venture-backed AI competition into public markets. The listing would follow SpaceX's recent blockbuster debut valued at US$1.77-trillion; those shares opened at US$160 on June 12, dipping to around US$147 from the US$135 IPO price—a trajectory that could give investors pause as they reassess valuations for high-growth companies. AI and chip stocks have recently sold off, testing whether enthusiasm for the sector can withstand scrutiny given ambitious growth projections. Still, the IPO stands to cap one of the strongest years for U.S. offerings since 2021.

Anthropic's fiercest competition remains OpenAI, which confidentially filed for an IPO in June and is expected to list by early 2027. Anthropic was founded by researchers who departed OpenAI over governance and safety disagreements. Since releasing its first large language model in March 2023, Anthropic has expanded to compete with SpaceX's xAI, Alphabet, and Meta on AI infrastructure buildouts.

Amazon and Google, two of Anthropic's early strategic partners, have each invested billions into the startup while simultaneously providing cloud infrastructure essential to train and deploy Claude models. Anthropic and Amodei have also clashed with the White House over its tools, leading the Pentagon to temporarily blacklist the company—a move a U.S. judge blocked in August.