Cognition, the startup behind the Devin coding agent, just raised $2 billion at a $48 billion valuation. That’s according to TechCrunch AI, which reports the round landed only four months after Cognition’s last raise valued it at $26 billion. Andreessen Horowitz led the deal, joined by Accel, Founders Fund, General Catalyst, and Avenir.
The headline number matters, but the signal underneath it matters more. Nearly doubling a valuation in a single quarter tells you investors don’t believe AI coding is a winner-take-all race. There’s room for several big players, and VCs are paying up to hold seats at that table.
📈 The numbers behind the raise
Cognition’s growth is the real story. TechCrunch AI reports its annualized run-rate revenue jumped from $492 million in May to $900 million now. Run-rate usually means one month’s top line multiplied by 12, and Cognition didn’t explain its exact math, so treat it as a fast-moving snapshot rather than booked annual revenue.
Here’s how the raise stacks up:
- Valuation: $48 billion, up from $26 billion four months ago
- New capital: $2 billion
- Run-rate revenue: $900 million, nearly doubled since May
- 2026 target: $4 billion to $5 billion in annualized revenue, per The Information
- Customers: Mercedes-Benz, NASA, Goldman Sachs, and Citi
What stands out is the multiple. Cognition is now valued richer per dollar of revenue than rival Cursor was in the spring, when Cursor had already passed $2 billion in annualized revenue during talks to raise at a $50 billion valuation.
🔍 Why the Cursor comparison matters
Cursor’s story is the cautionary note here. It was in talks at $50 billion in April, then sold to SpaceX for $60 billion later that month. TechCrunch AI reports the sale happened largely because Cursor was severely compute-constrained. It had the revenue and the demand, but not enough chips to keep scaling on its own.
Cognition could hit the same wall. It leases an Nvidia server cluster that costs hundreds of millions a year, and The Information estimates total cash burn could reach $800 million in 2026. Whether that becomes a Cursor-style squeeze is still unclear.
One detail worth sitting with: a16z was a major backer of Cursor and made a killing on the SpaceX sale. Now it’s back, leading a round in a direct Cursor competitor. When the same firm bets big on two rivals in the same category, it’s telling you the category is bigger than any single winner.
🛠️ The build-your-own-model play
Cognition is following the same cost playbook Cursor ran before joining SpaceX. It’s training its own model on open source foundations to cut its dependence on expensive third-party models from OpenAI and Anthropic. Over time, that’s how it trims burn and moves toward breakeven.
The logic is simple. Every query routed through someone else’s frontier model is a check written to a competitor. Owning more of the stack means better margins and more control over roadmap and pricing. It also means Cognition takes on the cost and risk of training, which is exactly where the compute constraint bites hardest.
💡 What it means for you
If you build with or buy AI coding tools, a few things follow from this raise:
- Expect more competition, not consolidation. Multiple well-funded players means faster feature cycles and more pricing pressure in your favor.
- Watch the compute question. The tools that secure chip supply will ship and scale. The ones that don’t may get acquired, like Cursor.
- Model independence is becoming table stakes. Vendors training their own models are betting on durability, and that’s a signal of who plans to be around in two years.
Cognition was founded in 2024 by math prodigy Scott Wu, and TechCrunch AI notes it’s expected to reach $4 billion to $5 billion in annualized revenue by the end of 2026. Cursor, for context, was tracking toward more than $6 billion before its sale.
The race for AI coding dominance is far from settled, and the capital pouring in says the biggest moves are still ahead. For the full breakdown, check the original report at TechCrunch AI.