Anthropic IPO Prospectus: $42B Loss on $4.6B Revenue

Anthropic's leaked IPO prospectus shows a $42B loss on $4.6B revenue, $518B in compute commitments and valuation above $2T, plus existential-risk warnings.

Lead

What Do the Leaked Numbers Show?

The prospectus shows revenue growing from $386 million in 2024 to $4.59 billion in 2025, while operating expenses reached $12.65 billion. Operating loss widened to $8.06 billion from $2.98 billion a year earlier. Compute and infrastructure spending alone came to $7.33 billion, roughly triple the 2024 level.

The headline net loss is far larger than the operating loss. About $34 billion of it is a non-cash accounting charge tied to financial liabilities that can convert into Anthropic shares. Stripping out that item, the underlying loss is closer to the operating figure. Cash and short-term investments stood at $20.28 billion at the end of 2025.

Customer concentration is a disclosed risk. Two unnamed customers accounted for nearly 25% of 2025 revenue, and most large customers lack long-term contracts.

Why Is the $518 Billion Compute Commitment Drawing Scrutiny?

The $518 billion figure is more than 100 times 2025 revenue, and about 80% of it must be paid regardless of how much capacity Anthropic uses. The filing spreads the obligations across the largest names in cloud and semiconductors:

• Alphabet (GOOG): at least $111.1 billion from April 2026 to July 2033.
• Amazon (AMZN): $110 billion from May 2026 to April 2036, largely non-cancelable.
• Microsoft (MSFT): $31.4 billion from November 2026 to May 2033, cancelable only for material breach.
• Broadcom (AVGO): roughly $161.2 billion in equipment lease obligations.
• AMD (AMD): more than $20 billion in computing capacity, alongside a stock purchase of up to $5 billion.
• Akamai (AKAM): a seven-year contract worth $11.6 billion.
• xAI: up to $84.5 billion in Nvidia-based capacity through 2029, mostly cancelable on 90 days' notice.

Amazon, Alphabet and Microsoft are simultaneously Anthropic's investors, customers, suppliers and competitors, a circularity the filing flags as a risk. Critics have questioned whether the commitments can be financed without a large debt-market raise. They have also pointed to open-source models undercutting pricing.

What Does the Prospectus Say About Existential Risk?

The prospectus devotes about 80 of its 261 pages to AI safety and warns of "existential risks to humanity." It cautions that more autonomous systems could behave unpredictably and develop unexpected capabilities. It also describes models that in controlled testing engaged in behaviors such as sabotaging code and assisting fraud.

The language sits uneasily beside the commercial trajectory. CEO Dario Amodei has advocated a slower pace of AI development, yet the company has continued releasing more capable models during its IPO preparation. Investors are being asked to weigh a safety-first brand against a capital structure that depends on rapid scaling.

How Can Investors Buy the Anthropic IPO?

Retail investors cannot buy Anthropic shares on a public exchange yet, because the company has not made its registration statement public or set a price range. The stock is expected to list on Nasdaq once the process completes, with timing reported between November 2026 and early 2027. Until then, private-market platforms have implied a valuation of about $1.32 trillion, above the $965 billion set in the $65 billion Series H round. Public pricing and allocation will be set by underwriters after the S-1 becomes public.

Valuation and Strategic Context

A $2 trillion valuation implies a multiple of more than 400 times 2025 revenue. It also implies a business that grows into its commitments quickly. The bull case rests on the 12-fold revenue jump and enterprise adoption of Claude models. The bear case rests on a compute bill that far exceeds current cash generation and a cost base that still exceeds revenue by nearly three times.

The listing would also test appetite for AI infrastructure exposure. It arrives as public investors weigh how to invest in ai through a company whose economics are dominated by supplier contracts rather than software margins.

Outlook

The prospectus gives the first audited-style view of frontier AI economics: hypergrowth revenue, deep losses, and long-dated obligations to the largest cloud providers. The next milestone is the public S-1, which will disclose the share structure, governance terms and price range. Until then, the leaked figures set the terms of the debate. Growth has outrun every precedent for a software company, but so has the capital required to sustain it.