Analysis
Anthropic told investors its preliminary second-quarter revenue topped $11.5 billion, according to CNBC. The quarter-over-quarter climb:
- Q2 2025 -- $787M
- Q1 2026 -- $4.73B
- Q2 2026 -- >$11.5B (14x+ YoY, ~2.4x Q1)
The company also posted positive adjusted operating income for the period -- its first reported profitable quarter on that metric, a milestone Fortune's reporting frames as central to the investor pitch CFO Krishna Rao is now making ahead of a possible IPO.
“Both frontier labs are converging on the same conclusion: enterprise contracts, not consumer subscriptions, are what's actually moving the revenue needle in 2026.”
The scale of the jump is worth sitting with:
- Full-year 2025 revenue -- ~$10B (all of last year, for comparison)
- Run-rate disclosed in May 2026 -- $47B (forward-looking, annualized)
- Actual Q2 2026 print -- >$11.5B
Anthropic has now more than matched its entire 2025 haul in a single quarter. The Q2 print still trails the $47 billion run-rate on an annualized basis -- a gap that reflects revenue accelerating within the quarter itself rather than the number being overstated.
## Where the revenue is coming from Anthropic doesn't break out revenue by product line in these preliminary disclosures, but the company's own public statements this year have pointed to enterprise API usage -- particularly coding tools built on Claude -- as the primary growth engine, alongside Claude Code adoption inside large engineering organizations. That mirrors the story at OpenAI, whose CFO Sarah Friar told investors this week that enterprise revenue has now overtaken consumer revenue for the first time, arriving two quarters ahead of her own earlier forecast. Both frontier labs are converging on the same conclusion: enterprise contracts, not consumer subscriptions, are what's actually moving the revenue needle in 2026.
## The competitive backdrop is not simple tailwind The growth number lands the same week Alibaba's Qwen family of open models passed 3 billion downloads, overtaking Meta and Google combined on Hugging Face's own tracking, and the same week Anthropic and OpenAI both cut prices to hold onto cost-sensitive customers drifting toward cheaper Chinese alternatives like DeepSeek, Zhipu's GLM-5.2 and Moonshot's Kimi K3, some of which undercut US frontier pricing by 60-90%. Anthropic introduced Claude Opus 5 at $5 per million input tokens and $25 per million output tokens as part of that repricing. Revenue is growing fast and the per-token price of the product driving that revenue is falling at the same time -- which means usage volume, not price, is doing essentially all of the work behind the $11.5 billion figure.
## What the number doesn't tell you Preliminary figures are exactly that: Anthropic's own disclosure notes the numbers could still change once finalized, and "positive adjusted operating income" is a non-GAAP measure that typically excludes stock-based compensation and other costs a public-market income statement would have to include in full. A company can post positive adjusted operating income while still burning significant cash on compute commitments -- Anthropic has multi-year cloud and chip agreements with Amazon, Google and others that don't show up cleanly in a single quarter's operating snapshot. None of that makes the growth rate less real; it does mean the profitability framing investors are hearing in these early IPO meetings is more selective than a finalized 10-Q would be.
## What it means for the IPO calendar For founders and GPs tracking the AI-infrastructure trade, this is the clearest data point yet that Anthropic's private-market valuation conversations -- reportedly as high as $2 trillion -- are being underwritten by actual revenue growth, not just compute-capacity narrative the way some infrastructure rounds have been this year. That distinguishes Anthropic's IPO setup from names like Cerebras, which fell roughly 14% after its second earnings report even after raising guidance, on a GAAP revenue miss versus consensus. Anthropic avoiding that kind of gap between narrative and reported number is exactly what its bankers will want to point to once the roadshow starts.
The risk sitting underneath the headline growth rate is concentration: Anthropic's enterprise revenue depends heavily on a relatively small number of very large contracts and a handful of hyperscaler compute partnerships, and any slowdown in enterprise AI capex -- a risk Arm co-founder Hermann Hauser flagged this week as the AI trade's real fragility -- would hit a 14x growth rate far harder than it would hit a more diversified consumer subscription business. Rao's investor meetings will need to answer that concentration question directly, not just the growth-rate one, before a fall listing gets priced.
