Cognition raised $2B at a $48B valuation, nearly doubling its price in four months
Run-rate revenue went from $492 million in May to $900 million in September. So did the burn: the company is leasing Nvidia capacity at a cost that could take 2026 cash consumption to $800 million.
Cognition has raised $2 billion at a $48 billion valuation, up from $26 billion in May. Andreessen Horowitz, Accel, Founders Fund, General Catalyst and Avenir led the round, according to TechCrunch, which reported the terms on 8 September.
The revenue growth behind the mark is real and fast. Annualised run-rate revenue rose from $492 million in May to $900 million by early September, and the company is projecting $4 billion to $5 billion annualised by the end of the year. That last figure is a company projection, not a booked number, and it implies roughly a fivefold increase in under four months.
Cognition was founded in 2024 by Scott Wu and sells Devin, an autonomous coding agent. Its enterprise list includes Mercedes-Benz, NASA, Goldman Sachs and Citi — buyers who tend to run long procurement cycles, which is a more durable signal than developer sign-ups.
What the price is actually saying
The interesting comparison is Cursor. Anysphere's editor was doing about $2 billion annualised — more than twice Cognition's revenue — when SpaceX bought it for $60 billion in April. On the multiple, Cognition is now priced above where Cursor was in the spring.
Read one way, that is exuberance. Read the way its investors describe it, it is a bet that AI coding is not a winner-take-all market: that an autonomous agent that takes a ticket and returns a pull request is a different product from an editor a developer sits inside, and that both can compound at once. The clearing price for that thesis is now $48 billion.
There is a third reading worth naming. Cursor's exit removed the category's largest independent competitor from the venture market entirely, and the capital that would have gone into its next round had to go somewhere.
The cost line
Agents are expensive in a way editors are not. A coding agent that works for hours on a task consumes orders of magnitude more inference than a developer accepting completions, and Cognition does not own the hardware. The company leases Nvidia servers at a cost running to hundreds of millions of dollars a year, and its 2026 cash burn could reach $800 million, according to TechCrunch's reporting.
Against $900 million of run-rate revenue, that is a business converting nearly all of its revenue into compute and headcount. It is defensible while revenue is compounding at this rate and while the $2 billion covers the gap. It is a different conversation if the growth curve flattens before inference costs fall, and inference cost per unit of work is the variable neither Cognition nor its investors control.
What is not known
Cognition has not disclosed gross margin, net revenue retention, or what share of the $900 million comes from a small number of large enterprise contracts. It has not said what portion of the projected $4 billion to $5 billion is contracted versus pipeline. Nor has it explained how the Nvidia lease is structured — whether it carries minimum commitments that would survive a slowdown.
Those are the numbers that decide whether $48 billion is early or late, and none of them are public. What is public is that five of the largest firms in venture capital looked at them and decided to nearly double the price they paid in May.
Runs the newsroom. Rename this profile in the studio to your own byline.
Related
Every weekday, the AI stories that moved money or shipped code.
No cross-posting, unsubscribe anytime. See all newsletters