Anthropic's S-1 shows $16.3 billion in six months and an $80.7 billion net loss

A revenue line that adds a zero a year, and a loss line that does not mean what it looks like

Anthropic filed its S-1 on October 26. Revenue went from $41 million in 2023 to $386 million in 2024 and $4.6 billion in 2025. For the six months to June 30, 2026 it was $16.3 billion against $1.2 billion a year earlier. The filing itself warns that it does not expect growth "at comparable rates". Customers paying $100,000 or more on a trailing-twelve-month basis rose from over 950 to over 11,900 in a year.

The headline loss is a different animal. The filing reports net losses of $1.5 billion, $8.3 billion and $42.0 billion for 2023 to 2025, and $80.7 billion for the first half of 2026. Of the 2025 figure, $34.3 billion was the change in fair value of convertible notes, derivative liabilities and SAFEs, so the loss grows as the company's own valuation does. 2025 also carries a $1.5 billion settlement in the Bartz case. The company is a Delaware public benefit corporation, and has already put 5% of its fully diluted shares into a public benefit fund.

The bet underneath: committing to compute before the demand shows up

The filing says substantially all of its large purchase commitments are for compute, much of it take-or-pay, owed whether or not the capacity is used. That is the part the thread kept returning to.

"Everything comes back to whether the demand is really there. Anthropic thinks it is." — a Tern.sh developer

He went on to total the commitments he saw to Google, AWS, Broadcom and SpaceX (figures in his comment; the S-1 text available to us leaves the aggregate blank) and argued that Anthropic has no choice but to commit hard, because a lab without capacity "get[s] eviscerated on either availability or last-minute datacenter leases".

One product, a WeWork comparison, and a leaderboard

The risk-factor language says open-weight models may take "a portion of the market for certain use cases". A longtime commenter set that sentence next to a public ranking:

"meanwhile the top 5 models by token consumption on the day this is posted are all Chinese/open weight models." — a longtime HN commenter

The bear case came in two parts. On concentration:

"I would imagine most of the huge revenue increase over the last year is due almost entirely to AI coding spend." — a commenter

On structure:

"There are a lot of parallels to the WeWork S1. They've made absolutely massive forward cash commitments, on a business that's increasingly commoditized" — a commenter

Valuation math drew its own objection. One reader worked backwards from Amazon's roughly 30% a year over three decades and asked how much further a company worth $1.5 to $2 trillion could grow, concluding that matching even half of Amazon's run would need a valuation in the quadrillions.

What to read first in the filing

The S-1 makes the revenue story easy and the cost story hard. A buyer should look past the net loss, which tracks the company's own valuation, to compute spend against revenue: $8.9 billion on compute for $16.3 billion of sales in six months, plus obligations that run into the 2040s.