Anthropic's draft IPO prospectus shows an $80.7 billion loss

The headline is a loss CoinCentral called one of the largest half-year losses ever reported: $80.7 billion for the first six months of 2026, up from a $13 billion loss in the same period a year earlier. But most of it is not cash the company burned. About $72 billion came from an accounting charge reflecting the rising fair value of convertible notes and other financing instruments — as Anthropic's private valuation climbed, the estimated value of the notes that could convert into its shares roughly tripled year over year, and accounting rules booked the increase as a loss.

The same kind of charge drove the company's $42 billion net loss in 2025. Strip it out, and the 2025 picture was an $8.1 billion operating loss on $4.6 billion of revenue. The real business underneath is growing fast: first-half 2026 revenue reached $16.3 billion, up nearly fifteen-fold from $1.1 billion a year earlier. One concentration risk the filing flags plainly: about a fifth of that revenue came from just two customers, and the company says it expects a large share of revenue to come from a limited number of customers for the foreseeable future.

Compute is where the cash actually goes. Anthropic spent nearly $9 billion on computing capacity and infrastructure in the first half — up from $2.3 billion a year earlier and more than the $7.3 billion it spent in all of 2025. The forward commitments are on another scale entirely: at least $518 billion over roughly a decade on cloud services, computing and infrastructure, per Reuters' review of the draft.

The WSJ review put the total higher once leases and pending deals are included: about $450 billion in cloud agreements over the next seven to ten years — $111 billion with Google, $110 billion with Amazon Web Services, $31 billion with Microsoft — plus up to $86 billion renting Nvidia chips through Nscale and Lambda, and a Google-chip lease via Broadcom worth more than $160 billion. All told, future infrastructure costs top $600 billion, with up to $150 billion more under negotiation. The filing also admits an awkward fact: Anthropic asked for more capacity than its cloud partners were willing to provide, and those partners increasingly view the company as a competitor.

Roughly a third of the prospectus covers risk factors, and some of it is strikingly blunt. Anthropic says its AI models can show self-preserving behavior and can produce responses that resemble blackmail, and it warns that a major AI-related incident could cause damages too large to insure.

The filing also discloses co-founder pay: Dario and Daniela Amodei each earned base salaries of about $655,000 in 2025, with total compensation of roughly $18 million and $16.5 million respectively. The company says it is seeing growing political resistance to data center projects, and flags that some of the data center operators it works with have limited operating history.

The company is expected to go public in November, though some investors are now weighing a valuation around $1.5 trillion — down from earlier talk of $2 trillion. Growth is still torrid but slowing sequentially: at an investor meeting on October 14, Anthropic said third-quarter revenue jumped 45% to $16.8 billion from the prior quarter, after revenue had more than doubled from the first quarter to $11.6 billion in the second. Some investors were disappointed by the slower rate.

Two days before the prospectus details surfaced, on Sunday, Anthropic said it plans to commit 5% of its shares to a new public benefit fund. The numbers land in a week when the market is looking hard at whether the vast sums poured into chips and data centers are generating sufficient returns. On the growth curve Anthropic is reporting, the question is starting to get an answer — but the filing's $600 billion in commitments is the price of finding out.