Anthropic prices IPO at $1.3 trillion with $900 billion in compute obligations

Anthropic has set its initial public offering price at approximately $1.3 trillion to $1.4 trillion, a valuation that sits below early market expectations of $2 trillion.

The pricing proposal, released on November 10, places the maker of the Claude chatbot at the center of a high-stakes financial debate regarding the cost of artificial intelligence infrastructure.

The company's securities filings reveal approximately $900.9 billion in purchase commitments and lease agreements, primarily for computing hardware and facilities.

These obligations are not immediately due. Roughly $70.7 billion is payable within 12 months of September 30, 2026. Another $451.6 billion is expected to be incurred before the end of 2030, with the remainder extending through 2045.

Investors must weigh these long-term liabilities against the company's current financial performance. Anthropic reported an October annualized revenue run rate of approximately $81 billion, a figure that exceeds the $20 billion estimate often cited in industry discussions.

However, annualized revenue does not equal profit or free cash flow. The company must still cover costs for electricity, personnel, networking, and facilities before it can address its massive infrastructure bills.

The economics of serving different types of users vary widely. A single Nvidia RTX PRO 6000-class GPU, with a retail price near $20,000, might serve 15 to 20 light users or two intensive coding users.

If 20 light users pay $20 monthly, the total revenue is $400. Recovering the $20,000 hardware cost alone takes 50 months under that scenario, before accounting for power, maintenance, or engineering staff.

Intensive coding agents, which run long tasks and consume high token volumes, present a different cost structure. A $200 monthly subscriber who uses the service continuously may generate less contribution profit than a lighter user who shares hardware capacity more efficiently.

Nvidia holds a dominant position in the hardware market that supports these systems. The company announced an agreement in September 2026 to acquire Hugging Face for approximately $12.93 billion, with closing expected in 2027.

Hugging Face serves as a major distribution platform for open-source machine learning models. Nvidia already controls the CUDA software ecosystem that underpins much of GPU computing.

The acquisition combines hardware supply, software infrastructure, and model distribution under one corporate roof. Nvidia states that Hugging Face will remain open and that developers can still choose competing hardware, but regulatory observers are examining whether this structure limits competition for independent researchers and smaller cloud providers.