53x Revenue and 50% a Month: The Cognition Bet Nobody's Doing the Math On
An AI that writes 90% of its own code just doubled its price tag in eight months. The interesting question isn't whether that's crazy. It's which clock runs out first.
Eight months ago, Cognition was worth $10.2 billion. Yesterday it closed a round at $26 billion post-money. That’s a 2.5x jump in eight months. The makers of Devin, the autonomous AI software engineer, raised more than $1 billion at a $25 billion pre-money valuation, with Lux Capital, General Catalyst, and 8VC co-leading and Founders Fund and Ribbit Capital piling in.
Now the part that makes people choke on their coffee. Cognition is running at a $492 million annualized revenue run-rate, which puts the valuation at roughly 53x ARR. For reference, public software companies usually trade in the single digits to low teens. So the easy take writes itself: bubble, peak froth, the end is near.
I think the easy take is doing first-order math on a second-order problem.
The model: Second-Order Thinking.
First-order thinking stops at the immediate, obvious effect. The price is 53x revenue, therefore it’s overpriced. Done. Second-order thinking asks the harder question: and then what? What does the world look like two or three steps after the obvious one?
Run it forward. Devin’s enterprise usage has grown 50% month-over-month for six straight months, and revenue went from about $37 million a year ago to $492 million now. If that growth holds even partway, the denominator in “53x ARR” balloons fast, and the multiple compresses on its own. The number that looks insane today can look cheap in four quarters. That’s the bull case, and it’s a real one.
But second-order thinking cuts both ways, which is the whole point of using it honestly. The other clock is the moat. Cognition’s own CEO says the future isn’t one model but systems that combine many models and tools. Anthropic, OpenAI, and Google have all shipped competing coding agents. So the real race isn’t “is 53x too high.” It’s whether revenue compounds faster than the moat erodes. Growth rate versus commoditization. Two timers, one room.
My take.
Stop arguing about the multiple. The multiple is a snapshot, and snapshots lie about anything growing this fast. The signal worth watching is the gap between usage growth and competitive pressure. When 50% month-over-month starts slowing before the moat is proven, that’s your tell. Not before.
I build in regulated markets at /mkt, where the second-order question is permanent. Every product decision sits inside rules that move slower than the tech. You learn fast that the first-order answer is rarely the one that pays off. The second-order one almost always is.
None of this is investment advice. It’s a thinking exercise. Do your own work, and do it two steps deep.
If this was useful, share it with someone who builds things. And if you want the full toolkit of 50 mental models, my book is coming soon.
Spencer Gareiss is CPO at /mkt. Former Senior TPM at Robinhood (derivatives & prediction markets). Army Reserve Captain. Author of "Mental Models: How to Think, Act, and Win." Nothing here is investment advice or a solicitation of any security.
