By Hamza Faisal

Anthropic, the startup behind the Claude family of AI models, has updated its draft IPO prospectus — and the numbers are staggering. According to the Wall Street Journal, which viewed the new filing, the company generated $16.3 billion in revenue in the first half of 2026, roughly 15 times the $1.1 billion it made a year earlier. But it also posted a net loss of $80.7 billion over those six months, nearly double its $42 billion loss for all of 2025. The listing could value Anthropic at up to $2 trillion, making it one of the largest public offerings ever.

A loss that is mostly on paper

The headline loss is less frightening than it looks. About $72 billion of the $80.7 billion came from non-cash accounting charges tied to convertible notes and other financing instruments Anthropic issued to raise capital. As the company's valuation climbs, those instruments grow heavier as book liabilities — an accounting loss, not cash walking out the door. Strip those charges out, and the underlying half-year loss is roughly $8.7 billion, versus a $13 billion loss in the first half of 2025.

The revenue trajectory is the real story. At $16.3 billion for the half, Anthropic has already more than tripled its entire 2025 revenue of $4.6 billion — and it says it was profitable on its own adjusted measure in the second and third quarters. One risk for investors: about a fifth of first-half revenue came from just two customers.

The $450 billion cloud bill

The filing's other headline number is the scale of Anthropic's future computing commitments: roughly $450 billion in cloud agreements over the next seven to ten years, most of them noncancelable. The named deals include $111 billion with Google Cloud, $110 billion with Amazon Web Services, and $31 billion with Microsoft, committed "regardless of actual usage." Anthropic has also agreed to spend up to $86 billion renting Nvidia chips through smaller AI cloud providers Nscale and Lambda, with terms running to 2033 and 2034.

On top of that sits a deal worth more than $160 billion to lease Google's AI chips through Broadcom, pushing total future infrastructure costs past $600 billion. Deals still being negotiated could add another $150 billion, bringing the possible total to around $750 billion. The filing admits Anthropic has sought more capacity than the big cloud providers "were willing to provide" — and that those partners increasingly see it as a competitor, forcing it to lease its own data centers.

Why the filing matters

The updated prospectus is the clearest look yet at the economics of a frontier AI lab: revenue compounding ferociously while compute demand compounds faster. Anthropic spent nearly $9 billion on computing in the first half of 2026 alone — more than the $7.3 billion it spent in all of 2025. The bet only works if growth keeps accelerating, because the cloud bill is now on the order of a decade of current revenue.

The filing also doubles down on Anthropic's reputation for candor about AI risk. Roughly a third of the 300-page document is risk factors, including warnings that AI models can show "self-preserving" behavior to avoid being shut down. If the IPO goes ahead at anything like the reported $2 trillion valuation, those disclosures will be stress-tested by the market like nothing before them.

By Hamza Faisal