Andreessen Horowitz just led a $2 billion round into Cognition at a $48 billion valuation, only four months after the same company raised at $26 billion. That kind of acceleration would be extraordinary in any market. In AI coding, it’s becoming almost routine.
As reported by TechCrunch, the round was co-led by Accel, Founders Fund, General Catalyst, and Avenir. Cognition, the startup behind the AI software engineer Devin, says its annualized run-rate revenue has grown from $492 million to $900 million since May. That’s not a small jump. And it’s arriving fast enough that investors clearly believe there’s still a window to get in.
The natural comparison is Cursor. Anysphere’s coding assistant was in talks to raise at a $50 billion valuation before agreeing to sell to SpaceX for $60 billion in April. At that point, Cursor’s annualized revenue had crossed $2 billion. Cognition is at roughly half that figure, yet its valuation is now close to what Cursor was fetching. That means Cognition is actually trading at a higher revenue multiple than Cursor was in the spring, which tells you something about how aggressively investors are pricing future growth here.
Cursor’s sale to SpaceX wasn’t purely a valuation play. According to investors familiar with its financials, Cursor was severely compute-constrained, and joining SpaceX gave it a path out of that bottleneck. Whether Cognition faces the same problem is an open question. The Information has reported that Cognition leases an Nvidia server cluster costing hundreds of millions of dollars annually, which could push its total cash burn to around $800 million this year. That’s a significant operational weight, and it’s the kind of number that makes the $2 billion raise look less like a victory lap and more like a necessity.
Still, Cognition is taking the same long-term approach Cursor was pursuing before the SpaceX deal. It’s training its own model on top of open source alternatives, with the goal of reducing dependence on expensive APIs from OpenAI and Anthropic. Getting off third-party model costs is probably the single clearest path to a healthier margin profile in this space, and it’s where most serious AI coding companies are headed.
The revenue trajectory looks promising. The Information projects Cognition will hit $4 billion to $5 billion in annualized revenue by end of 2026. Cursor was reportedly on track to exceed $6 billion by the same date, so there’s still a gap. But the fact that a16z, which made serious returns on Cursor when it sold to SpaceX, is now leading a round in a direct Cursor competitor is worth paying attention to. It suggests the firm doesn’t see this as a settled market with one dominant product.
Cognition was founded in 2024 by Scott Wu, a former competitive math prodigy. Its enterprise customer list includes Mercedes-Benz, NASA, Goldman Sachs, and Citi, which points to a deliberate push into large organizations rather than just developer tooling for startups.
The broader signal here is that AI coding is not shaking out the way winner-take-all software markets usually do. GitHub Copilot has Microsoft’s distribution. Cursor had the indie developer mindshare. Devin is pitching autonomous software engineering at the enterprise level. These are genuinely different bets, and investors appear to believe that multiple of them can be worth tens of billions of dollars at the same time. Whether that view holds once the market matures is another question entirely.
