Cognition, the company behind the autonomous coding agent Devin, is closing a new funding round of around $1bn at a valuation reported at $47bn on 2 September 2026 — up from $26bn in May and $10.2bn a year earlier. Annualised revenue has grown from $492m in late May to over $900m now, and investor demand for the round reportedly approached $10bn before it closed, well beyond the amount Cognition is actually raising.
This is the third valuation step-change for Cognition inside twelve months, and it closes out a trajectory this site flagged as “in talks” on 21 August, when the number being discussed was $40bn. The round landing higher than the reported talks, alongside revenue that’s nearly doubled in the same window, is the more interesting data point than the headline figure itself — it suggests investor appetite for autonomous coding agents isn’t cooling as the market matures, it’s accelerating faster than the companies themselves are announcing.
Why the revenue number matters more than the valuation
A $47bn valuation on its own is a bet on the future. Revenue growing from $492m to $900m+ in roughly three months is evidence that businesses are actually paying for what Devin does today, at a scale that’s hard to dismiss as pilot-project spending. That’s a meaningfully different signal to the funding rounds AI coding startups were raising even eighteen months ago, when valuations ran well ahead of any comparable revenue. It doesn’t mean the technology has closed the gap between “drafts working code” and “ships production software unsupervised” — it means enough organisations believe it’s close enough to be worth paying real money for right now.
So what
For anyone weighing how much of a software build to hand to an autonomous agent versus an experienced engineering team, the money flowing into this category is a reason to take the question seriously, not a reason to assume the answer is “hand over more.” Every serious production deployment of Devin, Claude Code or Cursor’s agent mode still sits inside human review, testing discipline and architectural oversight — the capability the market is pricing in is real, but it’s additive to good engineering practice, not a replacement for it. We build that balance into how we scope AI-assisted projects rather than treating agent capability as a reason to skip senior review. See our AI-assisted development approach, or get in touch to talk through where that line sits for your project.