Analysis
Anthropic posted $16.3 billion in revenue in the first half of 2026 against an $80.7 billion net loss, according to a leaked draft of its S-1 reviewed by the Wall Street Journal and reported by TechCrunch.
The draft also shows Anthropic believes it could commit up to $750 billion to compute capacity in the coming years, as the Claude maker prepares for a targeted November IPO that reports have pegged near a $2 trillion valuation.
How fast is Anthropic losing money?
The H1 2026 numbers mark a sharp acceleration in Anthropic's cash burn. The company's revenue run-rate had already reached roughly $47 billion by late May, after a Series H that valued it near $1 trillion. Q3 revenue alone reportedly hit $16.8 billion -- up 45% sequentially from Q2 -- in figures Anthropic gave prospective investors, meaning a single quarter nearly matched all of H1's revenue. The H1 loss, by contrast, ran roughly five times larger than H1 revenue. Anthropic has also leaned harder on debt this fall: a large facility led by Blackstone closed in October, following billions raised from a banking syndicate in September.
A supply problem money can't fix overnight
The more unusual disclosure concerns cloud capacity, not cash. The filing says Anthropic is struggling to secure enough compute from AWS, Google and Microsoft, because those hyperscalers increasingly see Anthropic as a competitor rather than a customer. That's pushing Anthropic toward leasing its own data centers and working with third-party operators it describes as having a "limited operating history" -- a risk disclosure that doubles as a warning about Anthropic's own supply chain. Roughly a fifth of Anthropic's sales come from just two customers, a concentration risk rarely spelled out this plainly ahead of a public listing.
Anthropic isn't racing alone. OpenAI is reportedly rushing to close an unusually fast round that could value it above $1.4 trillion, explicitly timed to land before Anthropic's IPO, with SoftBank and Abu Dhabi's MGX in talks to join -- a story Pulse previously covered in detail, as PYMNTS reported. Neither company's timing is coincidental -- both are trying to lock in capital and investor attention before the other's defining capital-markets moment.
What the headline number misses
A $750 billion compute commitment sounds like a down payment on AI dominance, but it's disclosed in a risk-factors section as a ceiling, not a signed, fully-funded contract. Much of it is still in negotiation, in the same filing that flags Anthropic's dependence on cloud providers who'd rather compete with it than sell to it. Investors underwriting an IPO near $2 trillion are effectively betting Anthropic can out-negotiate AWS, Google and Microsoft for capacity those companies increasingly want for themselves, while absorbing a loss rate that, this half, ran far ahead of revenue.
Anthropic competes most directly with OpenAI and Google DeepMind for frontier-model customers, and with xAI for capital and compute. All three are burning cash at a scale that would have been unthinkable in prior software cycles, but Anthropic's disclosed loss-to-revenue ratio this half is stark even against that backdrop.
The confidential draft hasn't surfaced on EDGAR, so the November timeline and the valuation figure both remain unconfirmed outside reporting on a leaked document. Whether Anthropic's actual S-1, once public, shows a cleaner path to covering that compute bill through signed contracts -- rather than negotiating leverage -- will matter more to the eventual price than anything in this leak.