Cognition, the startup behind the autonomous coding agent Devin, is in early talks with investors about a new funding round that could value the company at $40 billion or more, according to Bloomberg, which cited people familiar with the discussions. The talks were reported August 12, just under three months after Cognition closed a $1 billion round at a $26 billion valuation in May — a jump of more than 50% in roughly 90 days if the new round closes near that mark.

Revenue growth driving the talks

The reported trigger for the new round is revenue: Cognition's annualized revenue run rate is now approaching $1 billion, roughly double what the company reported at the time of its May raise. Founded in 2023, Cognition built its business around Devin, an agent designed to work through real engineering tasks rather than just suggest code — it can read through a codebase, draft a plan, write the code, run tests, and fix its own mistakes with limited human supervision. Bloomberg's sourcing indicates any new round would likely bring in more than $1 billion, though the outlet cautioned the talks are early and Cognition could ultimately decide not to raise, or to seek different terms.

Part of a broader AI-coding funding wave

Cognition's reported valuation jump lands amid a broader run-up in valuations for AI coding-agent startups, as enterprises increasingly hand agents tasks that used to require a human engineer end-to-end — writing, testing, and shipping code rather than just autocompleting it. Cognition has also been expanding beyond Devin itself: it acquired the AI browser-automation startup Poke earlier this year, folding its technology into Devin's toolset. A $40 billion valuation would put Cognition in the same tier as some of the best-funded AI labs, on the back of a product category — autonomous coding agents — that barely existed three years ago.

Why it matters

If the round closes near the reported figures, it would mark one of the fastest valuation increases of any AI coding startup this cycle, and a signal that investors are still willing to underwrite steep valuation jumps for agent companies that can show real usage growth rather than just model benchmarks. It also raises the bar for rivals in the AI-coding-agent space — including GitHub Copilot, Cursor, and Anthropic's own Claude Code — all of which are competing for the same enterprise engineering budgets that Devin's growth suggests are shifting toward autonomous, rather than assistive, tools.