Cognition, the startup behind the autonomous coding agent Devin, said on Friday, September 25, that it has crossed $1 billion in annualized revenue run rate. It reached that mark less than two years after Devin became generally available and less than three years after the company was founded. The figure is roughly double the $492 million run rate Cognition reported in May. It comes 17 days after the company raised $2 billion at a $48 billion valuation, which put it among the most richly priced private software companies in the world.
In a short post on its blog and on X, Cognition dedicated the milestone to its customers. It said Devin now works alongside engineering teams at GE Aerospace, Rivian, Rohlik and Exa, "among many others." The company restated the thesis it was founded on in January 2024: "the world needs far more software than it can build." It argued that every company is now a software company, from banks to automakers to governments. Cognition said its goal is "a world where software is good by default, delightful, and everywhere."
A Doubling in Four Months
The growth has been fast even by the standards of the AI coding market. Cognition said in May, when it raised $1 billion at a $26 billion valuation, that its run rate stood at $492 million. When it announced a Series E of more than $2 billion on September 8, it said that figure had grown to almost $900 million. Andreessen Horowitz and Accel led that round, joined by existing backers Founders Fund, General Catalyst and Avenir. Bloomberg reported on Friday that Cognition is on track for $1 billion based on its September performance, citing a person familiar with the matter. Cognition declined to comment to Bloomberg on the latest financials.
The customer list keeps getting bigger. Beyond the names in the milestone post, Cognition has said Devin supports chip design at Nvidia, aviation at GE Aerospace, financial services at Citi, automotive work at Mercedes-Benz and AI infrastructure at Modal. Benzinga reported that the company also counts Goldman Sachs, Elevance Health, Dell Technologies, Santander, the U.S. Army and the U.S. Navy as Devin customers. Startup users include Eight Sleep and OpenRouter.
The product has also moved well past what shipped at general availability on December 10, 2024, when pricing started at $500 per month for engineering teams. Cognition now sells Devin Auto-Triage for first-pass incident investigation, Devin Security Swarm for finding and triaging vulnerabilities, and Devin Automations, which start work from events in Slack, GitHub, Linear and other systems. The company calls itself an independent "agent lab" that mixes third-party and in-house models so customers are not tied to one provider.
Chief executive Scott Wu has said Devin is meant to take over work rather than suggest the next line of code. He has also rejected the idea that it is built to replace engineers. "We've never thought about it as replacing humans," Wu told TechCrunch in May. "When we started building Devin, it's kind of a funny thing, but we really just thought of it as: this is your buddy who helps you build more." In the same interview, Wu said Devin works at "somewhere between a junior and a mid-level engineer" depending on the task.
Why It Matters
Cognition's figure is one of the clearest signs yet that enterprises will pay heavily for agents that take on whole engineering tasks, not just autocomplete tools. It also suggests that AI coding is not a winner-take-all market. Cursor, the most visible rival, sold to SpaceX for $60 billion in a deal that closed in August. Before that sale, Cursor had been in talks to raise at a $50 billion valuation on more than $2 billion in annualized revenue. TechCrunch noted that Cognition's $48 billion price gives it a higher revenue multiple than Cursor had in the spring. Andreessen Horowitz, a major Cursor backer, came back to co-lead Cognition's round, which suggests investors see room for more than one large winner.
The headline number needs some context. A run rate usually multiplies one recent month of revenue by 12. It is a snapshot of momentum, not revenue booked over a full year. Cognition has not disclosed actual revenue for the period, how the $1 billion breaks down, or how much of it comes from recurring contracts rather than usage-based spending that can shift from month to month. Costs matter too. The Information has reported that Cognition leases an Nvidia server cluster costing hundreds of millions of dollars a year and that its cash burn could reach $800 million in 2026. That is one reason the company is training its own models, building on open-source alternatives, to reduce its reliance on OpenAI and Anthropic.
What to Watch
The next test is whether Cognition can hold this pace. The Information has reported that the company is expected to reach $4 billion to $5 billion in annualized revenue by the end of 2026, which would mean roughly quadrupling in a single quarter. Three things will show whether the growth lasts: how often pilots at companies like GE Aerospace and Rivian turn into company-wide deployments, whether Cognition's in-house models bring down compute costs, and whether a rival well-funded by SpaceX responds on price. Wu has described the road ahead in simple terms. "I think we are in for a wild ride," he told TechCrunch.
“We've never thought about it as replacing humans.”— Scott Wu, Co-founder and CEO, Cognition