Cognition has closed a Series D of more than $1 billion at a $26 billion valuation, led by Lux Capital, General Catalyst, and 8VC, according to a post published on the Cognition blog on May 28. The round more than doubles the company’s $10.2 billion post-money valuation from roughly eight months ago.
Run-rate revenue stood at $492 million at the time of the announcement, up from earlier figures, with enterprise usage growing more than 10x since the start of 2026. Those numbers set up a valuation multiple of roughly 53x current ARR, aggressive even by the standards of a sector that has grown comfortable pricing two or three years of growth into today’s multiples.
The enterprise roster includes Citi, Goldman Sachs, Mercedes-Benz, Dell, Santander, Elevance, and the U.S. Army and Navy. Two of the named deployments are specific enough to benchmark: Mercedes-Benz reduced an eight-month legacy modernization project to eight days using Devin, and Itaú, Latin America’s largest bank, now resolves 70% of its security vulnerabilities automatically through the product.
Those numbers are company-supplied and have not been independently verified, but they are the kind of outcome metric that procurement teams at large enterprises recognize. They also explain why systems integrators Infosys and Cognizant have embedded Devin directly into their delivery workflows.
For context against the broader AI coding agent category: our May 22 piece put Cursor at a $3 billion ARR run-rate, with SpaceX holding a call option to buy the company at $60 billion once SpaceX’s own shares begin trading. Cognition is being priced at a different kind of claim: that Devin’s autonomous, multi-step execution model will capture a larger share of software engineering budget than a copilot sitting alongside a human developer. The market is making that bet at 53x current ARR. It may be right, but the arithmetic requires Devin to hold and grow share while four well-resourced competitors close in.
That competitive pressure is not theoretical. Claude Code shipped as a standalone agentic product earlier this year. Cursor has moved up-market into longer-horizon tasks. Grok Build, which we covered on May 27, is targeting the same enterprise software-engineering workflows. Each of those products has a distribution advantage: Cursor has the installed developer base, Claude Code has Anthropic’s model edge, and Grok Build has xAI’s pricing aggression.
Cognition’s stated structural answer is independence. The company describes itself as an agent lab that uses all foundation models and routes tasks to the best-fit model across more than 100 categories of software engineering work. Its own SWE-1.6 model, recently launched and reportedly the most-used model inside Windsurf, adds a proprietary layer to that story. The framing is portfolio optimization rather than single-model dependency, which is a real differentiation if Cognition can maintain evaluation depth as the model landscape keeps changing.
The valuation logic also connects to a point Simon Willison made this week: aggressive seat pricing combined with demonstrable productivity outcomes is the most reliable PMF signal the market has right now. Cognition’s $492M ARR, if accurate, suggests that signal is present. The $26B multiple is a bet that the signal sustains through the next two to three years of competitive compression.
For procurement teams currently negotiating Devin enterprise contracts, this round changes the pricing conversation in one specific way: Cognition now has enough capital to hold headline prices flat while competitors cut, and it has every incentive to lock in multi-year terms before the market reprices. Any enterprise agreement signed in the next 90 days should include explicit pricing-protection and audit-rights clauses for the revenue-per-seat calculation, because the next comparable data point, whether a competitor’s price cut or a public offering filing, will arrive well before the contract term ends.
Published on the Cognition blog on 2026-05-28.
