Anthropic just disclosed $17 billion in third-quarter revenue — up roughly 1,127% from the same quarter a year ago — as it filed an amended S-1 with the SEC that gives investors the clearest financial picture yet of the Claude-maker before its road show begins.
What the numbers actually say
For the nine months ended September 30, Anthropic generated $33.24 billion in revenue, up roughly 1,200% year over year. The headline loss for that period was $104 billion, though most of that reflects the change in fair value of convertible notes and derivative liabilities rather than operational burn. The operating reality is still severe: compute and infrastructure expenses rose $14.4 billion for the nine-month period, up 329% year over year.
The growth rate itself is where investors will focus hardest. Quarter-to-quarter revenue growth dropped to around 47% in Q3, down sharply from 144% in Q2 and 132% in Q1. That deceleration is real, and it arrives at an awkward moment — just as Microsoft, Google and Amazon's latest earnings stoked Wall Street anxiety about whether data center capacity will scale fast enough to meet Anthropic's demand.
The most striking single figure in the amended filing is the aggregate committed spend: $895.4 billion in purchase commitments and lease agreements, substantially all of it tied to compute and infrastructure. Anthropic gets much of its capacity from Amazon, Google and SpaceX, and also relies on neoclouds for some compute. Those commitments are a bet on a market that doesn't fully exist yet.
The same bet xAI and other frontier labs have been making: lock in infrastructure at scale before the revenue to justify it arrives, because the cost of being late is extinction. It rhymes with the hyperscaler capex arms race of the early cloud era, where the winners were defined less by current cash flow than by who could hold their nerve through the build phase. OpenAI filed its own prospectus quietly in June and has signaled a 2027 debut. The difference is that Anthropic is moving first, doing so with a deceleration already baked into the prospectus, and carrying more than 90 pages of disclosed risk factors covering political headwinds, complex infrastructure needs and the difficulty of balancing safety and commercial objectives.
What this scale means
A company committing $895.4 billion to compute while growing revenue at four-digit percentages is not optimising a business — it is betting on becoming a new kind of infrastructure layer. The Q3 deceleration matters less than what the commitments imply: Anthropic is positioning itself as that layer, not as an application vendor. PE partners evaluating AI exposure should read this as the moment the frontier-model market stops being a venture story and starts being an infrastructure story — with the valuation multiples, lock-in dynamics and competitive moats that infrastructure commands. The companies that treat Claude as a commodity API today may find the terms of that relationship look very different once Anthropic trades publicly and its cost of capital changes.
