A $2 billion round arrives while the company burns cash faster than it earns it.
Four months ago, Cognition was worth $26 billion. It is now worth $48 billion. The $2 billion round it just closed values the maker of Devin at more than 50 times its annualized run-rate revenue of $900 million. Andreessen Horowitz led, alongside Accel, Founders Fund, General Catalyst, and Avenir.
Cognition’s revenue has nearly doubled since May, when it reported $492 million annualized. That is fast growth by any measure. The valuation moved faster. Cursor, the nearest comparison, was in talks at a $50 billion valuation in April with annualized revenue already past $2 billion. Investors are now paying more per dollar of revenue than they did in the spring for the larger company.
- Revenue now: Cognition at $900 million annualized, up from $492 million in May.
- Cursor’s mark: Past $2 billion annualized during its April funding talks.
- Year-end forecast: Cognition $4–5 billion; Cursor tracking above $6 billion.
- Exit price: SpaceX bought Cursor for $60 billion in April.
The costs are not small. It rents an Nvidia cluster with an annual bill in the hundreds of millions. Total cash burn could reach $800 million this year, according to The Information. Compute is also what ended the last comparison: investors familiar with Cursor’s financials say it sold to SpaceX largely because it could not get enough of it. Cognition’s answer is to train its own model on open source foundations, which would cut its dependence on OpenAI and Anthropic. If that works, the cost curve bends toward breakeven. If it does not, the burn keeps climbing.
Andreessen Horowitz backed Cursor and profited from the SpaceX sale. It is now leading a round in a direct rival. That detail carries weight. A firm that saw Cursor’s economics up close still believes a second company can build a large business here. Cognition is expected to reach $4 billion to $5 billion in annualized revenue by the end of 2026, while Cursor was tracking past $6 billion. Scott Wu founded the company in 2024. Mercedes-Benz, NASA, Goldman Sachs, and Citi are already customers.
The next number worth watching is not the valuation but the compute bill.
If you are picking a coding assistant for your team, this round argues against waiting for one winner to emerge. Two vendors are now funded well enough to keep shipping for years. The real diligence question has shifted: ask how your vendor secures compute and whether it controls its own model. That is what decided Cursor’s fate, not product quality.
