Cognition hits $48B valuation, signaling investors believe AI coding is far from a winner-take-all market
Cognition's valuation multiple is higher than Cursor's was before selling to SpaceX.
Cognition, a startup creating an AI coding assistant called Devin, has recently raised an impressive $2 billion, resulting in a $48 billion valuation. This significant increase in valuation, following a previous round led by the same investors, indicates that venture capitalists remain confident in the potential for multiple companies to thrive in the AI coding market.
Since the May 2023 fundraise, Cognition's annualized revenue has more than doubled, from $492 million to $900 million. This growth rate, while not disclosed by the company, is typically calculated as a monthly top line revenue multiplied by twelve.
In comparison to a rival coding assistant, Cursor, Cognition currently commands a higher revenue multiple. Just four months ago, Cursor was in talks to raise capital at a $50 billion valuation before being acquired by SpaceX for $60 billion. At that time, Cursor's annualized revenue had already surpassed $2 billion. The acquisition of Cursor was largely due to its severe compute limitations, as reported by investors familiar with the company's financials.
Cognition currently leases a substantial Nvidia server cluster, which could lead to annual cash burn of $800 million. However, the company is in the process of training its own model using open-source alternatives, a strategy that should help reduce expenses and move closer to achieving profitability. By the end of 2026, Cognition is projected to generate annualized revenue of $4 billion to $5 billion.
While this figure is lower than Cursor's projected $6 billion by year-end, it is still noteworthy that a16z, a major investor in Cursor, is backing Cognition, a competitor. Cognition's founder, math prodigy Scott Wu, has secured major enterprise customers such as Mercedes-Benz, NASA, Goldman Sachs, and Citi.
Written by urgent.news from TechCrunch's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.