Cognition has raised more than $1 billion at a $25 billion pre-money valuation—$26 billion post-money—the company confirmed on May 27. The round was co-led by Lux Capital, General Catalyst, and 8VC, with Founders Fund, Ribbit Capital, Atreides Management, and Layer Global among the participants. It marks roughly a 2.5x step-up from the $10.2 billion post-money valuation Cognition carried when it closed a $400 million round just eight months earlier in September.
The pace is the story. A startup that booked low single-digit-million ARR in late 2024 is now raising at a valuation that rivals public infrastructure names, on the back of an enterprise coding-agent market that has gone from speculative to load-bearing in under two years.
The numbers behind the markup
Reporting around the round puts Cognition's annualized revenue run-rate at about $492 million, with enterprise usage growing roughly 50% month-over-month over the past six months. Devin, the company's autonomous software-engineer agent, reached $73 million ARR by mid-2025; the acquisition of Windsurf's remaining assets last year more than doubled combined run-rate and pulled in a largely non-overlapping enterprise base. Named customers span Goldman Sachs, Citi, Dell, Cisco, Ramp, Palantir, Mercedes-Benz, NASA, Santander, Nubank, and Mercado Libre.
CEO Scott Wu's pitch is a two-surface product: Devin for asynchronous, fully delegated tasks, and the Windsurf IDE for the in-the-loop editing where engineers want to make each call themselves with AI assistance. Cognition's bet is that production teams want both, not one or the other.
Why an independent agent vendor survived
The conventional read heading into 2026 was that the frontier labs would swallow this category outright—Anthropic's Claude Code, OpenAI's Codex, and Google's Jules all ship coding agents bundled with the models underneath them. A $1 billion round at a $26 billion valuation is a direct counter-argument: enterprises are willing to standardize on a model-agnostic vendor that sits above the labs and orchestrates work across IDE and agent.
That positioning is also Cognition's risk. Its agents run on third-party frontier models, so margin and capability are partly hostage to the same companies it competes with on the surface layer. The capital cushion—against a reported lifetime net burn under $20 million before this raise—buys runway to push its own SWE models and lock in enterprise distribution before that pricing leverage tightens.
What changes for builders
For engineering orgs evaluating coding agents, the signal is that Cognition is now funded to behave like permanent infrastructure rather than an acquisition target. Expect deeper enterprise SLAs, more aggressive IDE-plus-agent integration, and continued consolidation pressure on smaller coding-tool startups. The valuation also resets the bar for the category: $26 billion is now the number every rival coding-agent vendor gets measured against in its next raise.
