Anthropic's Q3 Revenue Hit $16.8 Billion, but IPO Valuation Talk Cooled to $1.5 Trillion

Anthropic is having a strange week: the numbers keep getting bigger and the mood keeps getting cooler. The company told prospective IPO investors that its third-quarter revenue reached $16.8 billion, up 45% from the second quarter, and that it turned a profit on an adjusted basis during the quarter, according to reporting from the Financial Times and Bloomberg. That is real money at a pace most software companies never touch.

Yet the same investor conversations are producing a smaller headline number. Reuters, citing four sources, reports that some existing and prospective investors now expect Anthropic’s IPO to land around $1.5 trillion rather than the roughly $2 trillion the company had been targeting. The reason is a single letter doing a lot of work: deceleration.

The math that spooked the market

The quarter-over-quarter trajectory tells the story. Q1 revenue was about $4.7 billion. Q2 came in near $11.6 billion, a jump of roughly 147% in a single quarter. Q3’s $16.8 billion is a 45% sequential gain, which is still extraordinary by any normal standard, but it is less than a third of the previous quarter’s growth rate. Investors pricing a two-trillion-dollar IPO are effectively underwriting years of continued hypergrowth, so the shape of that curve matters more than any single quarter’s absolute number.

The market context did Anthropic no favors. AI-linked stocks have been selling off, with the SOXX chip index falling for eight straight sessions, and reporting around the IPO notes rising anxiety about debt-financed data center expansion. Investor Gavin Baker argued the bear case misses the point, saying Anthropic has positive free cash flow, accelerating October growth, and unit economics “much better than expected,” with about 75% of compute spending going to training rather than inference.

Why this IPO matters beyond Anthropic

Anthropic’s listing is widely seen as the litmus test for the entire private AI market, ahead of a possible eventual OpenAI IPO. If the most revenue-efficient frontier lab cannot command its target multiple, every downstream valuation, from model labs to the infrastructure providers selling them GPUs, gets repriced. Reporting around the deal suggests the company could stage the largest IPO in history, with timing talk centered on after the November midterms.

The funding arms race that produced these numbers has been visible all year: Anthropic locked up a $40 billion deal trading equity for a million Google TPUs, secured 2GW of AMD MI455X capacity in a $5B arrangement, and booked 250,000 next-generation GPUs in a $10 billion cloud agreement. Revenue at this scale exists to pay for commitments like those.

For the engineers and founders watching rather than trading, the takeaway is less about the valuation than the growth pattern. Frontier AI revenue is still compounding fast enough to fund the largest infrastructure buildout in corporate history, but the market has started demanding proof that the compounding lasts. The IPO prospectus, when it lands, will be the most consequential AI document of the year.