$10B to $26B in Eight Months. Cognition Didn’t Build a Coding Tool. They Built a Moat.
The Devin round isn’t a bet on AI. It’s a bet on infrastructure that’s already working.
A startup founded in late 2023 just closed a $1 billion Series D at a $26 billion valuation.
Eight months ago, that same company was worth $10.2 billion. The 2.5x jump didn’t come from a pivot or a press cycle. It came from one thing: the numbers got undeniable.
Cognition’s flagship product is Devin — an autonomous AI software engineer that doesn’t assist developers, it replaces tasks they used to do themselves. Enterprise usage has grown more than 10x since the start of 2026. Run-rate revenue is sitting at $492 million. Mercedes-Benz needed eight months to modernize a legacy system. Devin did it in eight days.
Lux Capital, General Catalyst, and 8VC co-led the round, with Ribbit Capital, Atreides, and Founders Fund also participating. Goldman Sachs, Citi, the U.S. Army, and the U.S. Navy are customers. This isn’t a demo anymore.
Mental Model: Second-Order Thinking
Most people read the Cognition headline and think: AI coding is hot, investors are piling in.
That’s first-order. Stop there and you miss what’s actually happening.
Second-order thinking asks: what does a $26 billion autonomous coding company mean for everything built on top of software engineers?
Here’s where it gets interesting. 89% of code committed by Cognition’s own engineers is now written by Devin. They’re not just selling the product — they’re their own clearest proof of concept. That feedback loop is compounding. Every improvement to Devin makes Cognition faster, which funds better models, which makes Devin better.
Second-order: the moat isn’t the model. It’s the flywheel built around the model. Cursor’s great. GitHub Copilot’s everywhere. Both assist human engineers in writing code. Devin takes ownership of the task itself and executes it autonomously. That’s a different product category with a different switching cost.
Third-order, if you want to go there: what happens to software company headcount in 36 months if this trajectory holds? That’s the question nobody in enterprise HR wants to answer out loud.
The Contrarian Take
Everyone’s calling this a sign that AI coding is overvalued. I think that’s the wrong frame.
Cognition’s revenue run rate grew from $37 million last May to $492 million today. That’s not a valuation story. That’s a revenue story that investors priced forward. At $26 billion on $492 million run-rate, you’re paying roughly 53x revenue for a company that 13x’d its top line in 12 months. Expensive? Yes. Irrational? Depends on whether the growth rate holds.
The real risk isn’t valuation. It’s concentration. Goldman, Citi, Mercedes, the U.S. military — those are enormous logos but a short list. Enterprise sales cycles for autonomous code agents are going to get longer and more complex, not shorter, as the deals get bigger. If Devin hits a few high-profile failures in production environments, the narrative flips fast.
That’s the bet Lux, General Catalyst, and 8VC are making — that Devin’s track record compounds faster than the scrutiny does.
At /mkt, we think about the same dynamic in regulated fintech. The moat we’re building isn’t a UI. It’s 14 months of FINRA correspondence, filed forms, and a broker-dealer subsidiary. Nobody clones that quickly. The question is always whether the infrastructure compounds faster than the competitive pressure. Cognition’s answer, so far, is yes.