Cognition, the startup behind the autonomous coding agent Devin, has closed a Series E funding round of more than $2 billion at a $48 billion valuation, the company confirmed on September 8. The round was led by new investors Andreessen Horowitz and Accel, with existing backers Founders Fund, General Catalyst and Avenir also participating alongside a broad syndicate that includes Benchmark, Kleiner Perkins, Greylock and Lightspeed, among others.

From talks to a closed round

The close follows weeks of reported deal talks: Bloomberg and TechCrunch reported in mid-August that Cognition was in early discussions to raise at around $40 billion. The final terms landed well above that figure, marking roughly a 90% jump from the $26 billion valuation Cognition reached in its previous round just four months earlier, in May 2026. That May round itself had been a large step up — $1 billion raised at a $25 billion pre-money valuation.

Revenue nearly doubled since May

Cognition's run-rate revenue has grown from about $492 million at the time of its May raise to nearly $900 million now, according to figures cited around the announcement. Devin is designed to work through real engineering tasks with limited human supervision — reading a codebase, drafting a plan, writing code, running tests and fixing its own mistakes — rather than functioning purely as an autocomplete-style assistant. Cognition has also broadened its product line this year, including folding browser-automation startup Poke into Devin's toolset following an acquisition in July.

Why it matters

The round is one of the largest for any AI-agent startup this year and underscores how quickly investors are willing to re-price companies that can show fast-growing usage, not just capable models. A $48 billion valuation puts Cognition alongside some of the best-funded AI companies in any category, built almost entirely around a single agent product in a segment — autonomous software engineering — that has existed commercially for barely two years. It also raises the competitive bar for rivals chasing the same enterprise engineering budgets, including GitHub Copilot, Cursor and Anthropic's Claude Code, as the market increasingly rewards agents that can complete tasks end-to-end rather than merely assist with them.