Cognition's AI Coding Startup Hits $48 Billion Valuation, Nearly Doubling in Four Months

Cognition, the startup behind the autonomous coding agent Devin, announced on September 8, 2026 that it raised more than $2 billion in a Series E round at a $48 billion valuation, according to the company's own announcement and confirmed by Reuters.

That's nearly double the $26 billion valuation Cognition carried just four months earlier, when it raised over $1 billion in its Series D, according to Reuters and Cognition's May 27, 2026 funding announcement.

Andreessen Horowitz and Accel led the new round as first-time investors, joined by existing backers Founders Fund, General Catalyst, and Avenir, according to Reuters and TechCrunch. The syndicate also included Benchmark, Bessemer, Kleiner Perkins, Greylock, Lightspeed, T. Rowe Price, Lux Capital, and Nvidia itself, according to a company profile compiled by Dealroom and a detailed list published by Unite.ai.

The Revenue Case

Cognition says its annualized run-rate revenue climbed from $492 million in May to almost $900 million now, according to the company's own blog post cited by Reuters, TechCrunch, and Dealroom. Cognition has not disclosed exactly how it calculates that figure, though the standard method takes one month's revenue and multiplies by 12, as TechCrunch noted.

The company credits the growth to expanding use of Devin among enterprise clients including Nvidia, which uses it for chip design, GE Aerospace, Citi, Mercedes-Benz, and the cloud infrastructure firm Modal, according to Unite.ai's account of Cognition's announcement. Goldman Sachs and NASA were also named as customers by TechCrunch.

The Information reports Cognition is on pace to hit $4 billion to $5 billion in annualized revenue by the end of 2026, a figure cited by TechCrunch and a KuCoin market analysis piece.

The Cost Side Nobody's Skipping

Cognition leases an Nvidia server cluster costing hundreds of millions of dollars a year, and The Information reports that could push the company's total cash burn to $800 million in 2026, a figure repeated by both TechCrunch and KuCoin.

That's real money going out the door before the company is anywhere near profitable. Cognition says it's training its own model based on open-source alternatives, a move TechCrunch reports is meant to cut reliance on paid third-party models from OpenAI and Anthropic and bring the company closer to breakeven. Whether that strategy actually closes the gap between $900 million in revenue and $800 million in burn is an open question none of the available reporting answers.

The Cursor Comparison

The most useful benchmark here is Cursor, another AI coding assistant. Cursor was in talks in April 2026 to raise at a $50 billion valuation before agreeing instead to sell to SpaceX for $60 billion, according to TechCrunch. At the time, Cursor's annualized revenue had already topped $2 billion and was on track to exceed $6 billion by the end of the year, per TechCrunch's earlier spring reporting.

That means Cognition, at $900 million in run-rate revenue and a $48 billion valuation, is commanding a higher revenue multiple than Cursor was when it sold. TechCrunch also flags that Andreessen Horowitz, which profited handsomely backing Cursor into its SpaceX sale, turned around and led the round for a direct Cursor competitor. This is a straightforward bet that venture money doesn't think AI coding is winner-take-all.

Cursor's own path to SpaceX was driven largely by being severely compute-constrained, according to investors familiar with its finances cited by TechCrunch. Cognition faces the same expensive Nvidia dependency, and none of the current sources establish whether it will hit the same wall.

What Skeptics Would Say

The skeptical read here is straightforward: a company burning close to $800 million a year while sitting on $900 million in run-rate revenue is not a profitable business, it's a bet that scale eventually arrives before the cash runs out. A four-month doubling in valuation, from $26 billion to $48 billion, on a startup founded in 2024 by Scott Wu, is the kind of number that invites comparisons to dot-com-era exuberance. That concern is fair on its face, and none of the sources here show Cognition disclosing a profitability timeline or unit economics beyond the top-line revenue figure.

What cuts against the bubble read is that Cognition's customer list, Goldman Sachs, Citi, Mercedes-Benz, NASA, Nvidia, GE Aerospace, is not speculative logos. These are large enterprises paying for a working product, and the revenue growth from $492 million to $900 million in four months is a measured jump, not a projection.

Cognition has expanded into Washington, D.C., Tokyo, Singapore, London, São Paulo, and Madrid over the past year, adding to hubs in San Francisco, New York, and Austin, according to Unite.ai's reporting on the company's announcement. Whether Cognition's push to train its own model actually gets its burn rate under control before its next funding round is needed, or whether, like Cursor, the endgame is an acquisition rather than an IPO, remains to be seen.