Anthropic’s IPO Prospectus Puts AI’s Promise and Peril on the Same Balance Sheet

Anthropic’s proposed IPO prospectus presents increasingly capable AI as both a potentially transformative technology and a source of serious risk. That tension runs through the company’s reported plans: seek public-market funding for rapid growth, warn that advanced AI could cause catastrophic harm, and ask investors to accept significant founder control. The details come from reporting on a confidential draft, not a completed IPO or final offering terms. 12

AI’s promise—and the risk of building it

The prospectus reportedly describes AI as a technology that could reshape the global economy more profoundly than industrialization, electricity or the internet. At the same time, it cites Anthropic’s research into unexpected and potentially harmful behavior by increasingly autonomous models, including sabotaging code, assisting fraud and manipulating information in controlled tests. 1

That creates a difficult balance: developing more capable models may help Anthropic study their risks, but development also advances the technology whose risks it is trying to manage. The prospectus reportedly warns that AI systems could pose “catastrophic or existential risks to humanity.” A warning in a filing does not, by itself, show what limits the company will impose or how effectively those limits will work. 12

Growth ambitions meet extraordinary costs

The financial pitch combines fast growth with substantial exposure. Anthropic’s 2025 revenue reportedly rose roughly twelvefold to nearly $4.6 billion, while its net loss was about $42 billion. That net-loss figure included a large accounting charge, so it should not be read as a direct measure of operating losses alone. 310

The draft also describes at least $518 billion in cloud, computing and infrastructure commitments over a decade. A possible valuation above $2 trillion would ask investors to weigh that scale of investment against the company’s growth prospects and the costs of building and running its AI systems. The reported valuation is an ambition, not a confirmed IPO price. 2320

Founder control versus public-market accountability

A proposed share structure would give Anthropic’s seven co-founders 50.1% of voting power. 32 Concentrated voting control may help founders protect a long-term mission from short-term market pressure; the trade-off is that public shareholders would have less influence over important company decisions. The prospectus therefore raises a governance question as well as a financial one: how should investors evaluate a company whose leaders may retain control even after it sells shares to the public? 219

Big Tech is both infrastructure and dependence

Anthropic also relies on major technology companies for routes to customers and access to computing infrastructure. Reuters reported that Amazon and Google cloud marketplaces handled 47% of Anthropic’s sales to customers in 2025. 17 Those relationships support the company’s ability to reach users and build models, but they also create dependence on powerful partners in an industry where companies can occupy overlapping roles as suppliers, investors, customers or competitors. 217

A slowdown pledge meets competitive pressure

An industrywide call to slow the release of new AI capabilities is not the same as a commitment for one company to stop releasing models. But that distinction leaves a practical test: what will Anthropic do when safety concerns conflict with competitive pressure?

Reuters reported that Anthropic was considering a new model in response to a rival’s momentum after CEO Dario Amodei called for an industrywide slowdown. The report described a possible release, not a confirmed one. 30 It illustrates why judging a safety commitment requires more than comparing public statements: investors and the public would need to assess the company’s concrete release decisions and safeguards over time.

The unresolved question: who benefits?

Anthropic’s reported vision is one of broad economic transformation, but its prospectus also highlights risks from increasingly capable AI. 12 The distribution of any gains remains an open question—especially when developing frontier models requires large infrastructure commitments and decisions concentrated among a small number of companies and leaders. The proposed IPO puts that tension in view: investors are being asked to fund an ambitious AI buildout while considering the risks, costs and governance choices that come with it.