It’s here.
Anthropic‘s long-awaited IPO filing offers the public’s first glimpse into the financials of a leading US frontier lab. The S-1 shows a company with rocketing revenue that’s trying to outrun the hundreds of billions of dollars it has committed to computing contracts.
“The compute commitments are enormous,” said Harrison Rolfes, a senior research analyst covering private companies at PitchBook.
Revenue hit $11.56 billion in Q2, up from $787 million during the same time last year. While the filing leaves Q3 blank, the company claims revenue was 86 times the $197 million Anthropic netted in Q4 2024. That would put Q3 revenue around $16.9 billion, a much slower rate of growth.
On paper, Anthropic’s losses are immense. In Q2, the company lost $66.3 billion, though most of its losses can be attributed to accounting rules. The company owes money in the form of convertible notes that increase in value as the company’s valuation rises.
Much of the company’s money is going to fund the compute that its AI is trained and run on.
The company spent $5.6 billion on compute contracts, up from $1.39 billion in spending a year earlier.
The charts below break down Anthropic’s S-1. From its revenue to its compute commitments and spending, the filing provides the biggest view to date on just how expensive it is to run an AI lab.
Latest Research
Too Central to Fail: The AI Venture Ecosystem Built Around One Company’s Strategy
Learn about our editorial standards.