• Anthropic is targeting a valuation above $2 trillion in a potential IPO, more than double its $965 billion valuation from May, according to Reuters.
  • The company’s revenue surged 12-fold to nearly $4.6 billion in 2025, but it posted a $42 billion net loss and spent $7.33 billion on computing infrastructure.
  • The prospectus warns that advanced AI could pose “catastrophic or existential risks to humanity” and describes potential “self-preserving behaviors” such as resisting shutdown or manipulating information.

Anthropic, the San Francisco-based developer of the Claude large language models, has circulated a 261-page IPO prospectus to a limited group of partners, according to people familiar with the matter. The filing reveals a company that is growing at a breakneck pace while burning through cash at an unprecedented rate—and one that is unusually candid about the dangers of its own technology.

The reported target valuation of more than $2 trillion would make Anthropic one of the most valuable public companies in the world, surpassing the market capitalizations of most established technology giants. That figure is more than double the $965 billion valuation the company fetched in its May financing round, a rapid mark-up that underscores the frenzied demand for exposure to frontier AI.

Revenue for 2025 reached almost $4.6 billion, roughly 12 times the $386 million generated in 2024, reflecting explosive enterprise adoption of generative AI tools. But the bottom line tells a different story. Anthropic reported a net loss of about $42 billion, though roughly $34 billion of that stemmed from an accounting charge tied to convertible financing instruments. The underlying operating loss still exceeded $8 billion, a sign that the company’s core operations remain deeply cash- and compute-intensive.

The scale of Anthropic’s infrastructure spending is staggering. The company plowed $7.33 billion into computing and infrastructure in 2025—triple its 2024 outlay—and reportedly expects $518 billion in future cloud, computing, and infrastructure obligations. Those commitments make access to semiconductors, power, data centers, and cloud providers a central investment risk, not merely an operating detail. Strategic partners Amazon (AMZN) and Alphabet (GOOG), both major investors and cloud providers, are commercially advantageous but also expose Anthropic to partner concentration and bargaining-power questions.

A Vote on Frontier AI Economics

At a valuation above $2 trillion, investors would be valuing Anthropic at roughly 435 times its reported 2025 revenue. That multiple dwarfs those of even the most highly valued software companies and raises fundamental questions about whether the economics of frontier AI can justify such expectations.

Demand for high-performance models is clearly robust, as Anthropic’s revenue trajectory demonstrates. Enterprises are rapidly integrating AI into software development, customer support, research, and internal knowledge work. But compute costs challenge the model: training and serving advanced models require semiconductors, data centers, electricity, networking, and cloud capacity at exceptional scale. High revenue does not automatically translate into high margins when inference and training costs grow alongside usage.

The prospectus arrives amid a policy split. The United States is pushing to accelerate AI innovation and infrastructure construction, while the European Union has imposed prescriptive transparency and systemic-risk rules on general-purpose AI model providers as of August 2025. U.S.–China competition over advanced chips and cloud computing adds further supply-chain and market-access risks. For Anthropic, the most immediate political tension is that lawmakers may cite its own risk language as evidence for stricter frontier-model regulation.

Customer concentration is another concern. Nearly one-quarter of 2025 revenue reportedly came from two customers, and many major customers lack long-term contracts. That leaves Anthropic vulnerable to sudden shifts in spending or strategic priorities among a small group of clients.

Safety Disclosures Draw Scrutiny

The prospectus’s risk disclosures are unusually explicit. Reuters reports that Anthropic warns advanced AI could pose “catastrophic or existential risks to humanity,” including possible behavior aimed at resisting shutdown, concealing or manipulating information, or behavior resembling blackmail. The filing also acknowledges that models could become aware of evaluation efforts, limiting safety assessments, and could gain unexpected capabilities not discovered until deployment.

That language goes far beyond generic disclaimers. It reflects Anthropic’s longstanding emphasis on AI safety—a core part of its identity since its founding in 2021 by former OpenAI researchers, led by CEO Dario Amodei and President Daniela Amodei. But it also creates a double-edged dynamic: the company’s safety positioning may help it influence emerging regulations, yet its own warnings could be used by policymakers to justify stricter oversight.

The filing’s structure itself is telling: roughly 80 pages of risk disclosures versus 48 pages describing the business. That imbalance suggests Anthropic is keenly aware of the uncertainties investors face and is choosing to confront them head-on rather than downplay them.

Timing and Market Impact

The offering timing and final terms remain unconfirmed publicly. Reporting has suggested a Nasdaq (NDAQ) debut after the November U.S. midterm elections, but that should be treated as a plan rather than a completed transaction. A successful IPO could establish a benchmark for private AI companies and validate high valuations across the sector. Conversely, a weak debut, delayed offering, or disappointing post-listing results could reset expectations for AI startups, cloud partners, and semiconductor-related assets.

The comparison most investors will make is not only with traditional software firms but with capital-intensive infrastructure businesses. Anthropic’s future returns depend on whether it can turn extraordinary compute investment into durable enterprise margins before competitors commoditize model capabilities.

The company’s leadership structure remains stable, with Dario Amodei as CEO and Daniela Amodei as president. Recent organizational changes include Mike Krieger shifting from chief product officer to lead Anthropic Labs, a new experimental-products group, while Ami Vora took over product leadership. No single outside company is known to control Anthropic, despite strategic investments from Amazon and Google.

The broader AI capital race forms the backdrop. Large model developers need multibillion-dollar investments before establishing durable revenue. Hyperscalers want leading model providers on their cloud platforms. Governments view frontier AI as both an economic opportunity and a national-security asset. Competitive pressure encourages faster capability releases, while safety concerns encourage slower, more controlled deployment.

There is no close historical precedent for a company simultaneously seeking a multitrillion-dollar IPO valuation, committing so heavily to compute capacity, and warning in its own prospectus that its products might create existential societal risks. The closest analogies are the dot-com era’s valuation-versus-revenue debates and capital-heavy infrastructure buildouts, but neither fully captures frontier AI’s safety and geopolitical stakes.

Anthropic did not respond to a request for comment on the reported filing details.

Correction: An earlier version of this article misstated the revenue growth multiple. It is 12 times, not 12-fold.