$61.5B → $965B
The valuation ran 15.7x in 14 months
Across four rounds from March 2025 to May 2026, selling the same core product into the same market. What changed was not the business, it was what investors were willing to pay for it.
Valuation study
Anthropic: Pricing the Pre-IPO AI Leader
How do you price a company that has no stock ticker? A valuation study of Anthropic written after it filed confidentially for an IPO.
Overview
Anthropic is a strange company to research, because you cannot buy stock in it. It is still private. In June 2026 it filed confidential paperwork with the SEC for an IPO, which got me curious about how a company gets priced before it has a ticker, and whether the number attached to it right now actually makes sense.
My argument: Anthropic's revenue growth and cost efficiency are real and well supported, but its valuation has grown faster than either one. The price investors are paying is currently ahead of what the business has actually proven.
Key points
1,000+
Enterprise demand looks real
Customers spending over $1M a year roughly doubled from about 500 in two months, and eight of the ten largest companies by revenue run significant workloads on Claude. That is deep integration, not a trial.
~60%
Costs are better than expected
Compute runs near 60 percent of revenue against an estimated 75 percent plus at OpenAI, cash burn is guided down from $5.6B to about $3B, and the company projected an unusually early profitable quarter.
Ahead of proof
But the multiple outran the fundamentals
A 15.7x jump is larger than revenue growth alone justifies, so part of that number is a future that has not happened yet. The IPO filing is the real test, when public markets can check these figures against full disclosures instead of investor projections.
The paper
Disclaimer: This is a student research paper, not professional financial analysis or investment advice. Anthropic is a private company and does not publish audited financial statements, so every figure here comes from company announcements, media reporting, and third-party financial sites rather than official filings. These numbers are largely self-reported projections and estimates, so they may be incomplete, revised, or inaccurate, and all amounts should be read as approximate and illustrative.