Cognition's $1B Round: $26B Valuation, Revenue & Devin Use

A billion dollars, eight months after the last one

Cognition announced on May 27, 2026 that it had raised more than $1 billion at a $26 billion post-money valuation ($25 billion pre-money). Lux Capital, General Catalyst and 8VC led the round. Existing investors including Founders Fund, Elad Gil, Alpha Wave, Bain Capital Ventures and Soma Capital returned, alongside new participants Ribbit Capital, Atreides and Layer Global. [1][2]

The prior round valued Cognition at $10.2 billion post-money in September 2025. Eight months separate the two. [1]

Atlas interpretation: $26 billion against $10.2 billion is a 2.5x markup in eight months, which is fast even by the standard AI coding startups were setting through 2025 and 2026. It is a smaller jump than some frontier-lab rounds of the same period, but Cognition is a much younger, much smaller company by revenue, so the percentage move is the more telling number here than the absolute size of the round. [1]

Revenue and usage, by two different measures

Cognition disclosed annualized run-rate revenue of $492 million. TechCrunch reported enterprise usage growing 50 percent month over month for the preceding six months. Cognition's own announcement described enterprise usage differently, as having grown more than tenfold since the start of 2026, and named customers including Citi, Mercedes-Benz, Goldman Sachs, Elevance, Dell, Santander, and the U.S. Army and Navy. [1][2]

Atlas interpretation: A steady 50 percent month-over-month rate and a cumulative 10x-since-January figure are not the same claim, and depending on the exact months involved they are not obviously consistent with each other either. Neither source shows the underlying monthly numbers, so there is no way to reconcile the two framings from what was published. Both are Cognition-supplied figures relayed by reporters covering a round Cognition was raising, which is the usual position for growth statistics attached to a funding announcement: directionally informative, not independently audited. [1][2]

89 percent, self-reported

Cognition's announcement stated that 89 percent of code committed by its own engineers is committed by Devin, the coding agent the company sells. It also said Devin had been in general use for two years at that point and cited an SWE-1.6 model running at up to 950 tokens per second, plus specific customer outcomes such as Mercedes-Benz compressing an eight-month legacy modernization project into eight days and Itau automatically fixing 70 percent of flagged security vulnerabilities. [2]

Atlas interpretation: The 89 percent figure is a dogfooding claim about Cognition's own engineering org, not an independently measured or third-party-audited statistic, and it says nothing about what fraction of a commit's logic originated with Devin versus a human prompting and editing it. It is nonetheless the company selling the strongest possible evidence that its own product is essential at the company itself, which is a different and more concrete claim than a benchmark score or a customer testimonial. The Mercedes-Benz and Itau figures carry the same caveat: they come from the vendor's own announcement, with no cited customer statement or independent write-up to check them against. [2]