Cognition's multiple and burn test for AI coding valuations

Cognition AI secured $2 billion of Series E financing, valuing the company at $48 billion. A growth investor underwriting an AI coding company can start with the multiple: at the May round and the Series E, price divided by run-rate revenue lands near 53x. That flat multiple is the first signal; burn, workforce terms, and investor-customer overlap are the cost of the category bet.

The May financing was $1 billion at a $26 billion valuation, and the subsequent round lifted the company's price to nearly twice that level. At the May financing, Cognition reported $492 million in run-rate revenue; by the new round, that figure had climbed to almost $900 million. New investors Andreessen Horowitz and Accel headed the financing, while Founders Fund, General Catalyst, and Avenir continued as backers. The investor list matters because it can change how independent the revenue line is.

A flat multiple is a category signal

When a multiple expands, investors are often paying for a winner-take-all position. When it stays flat while revenue grows, the market may be buying access to a category. Cognition's rounds show that pattern. That combination separates two bets: a single product that must dominate, and a software category that can support several players.

AI coding spans tools, agents, and workflows that companies may adopt in layers. A flat multiple can mean the market is comfortable with several winners, or it can mean it is not yet confident that any one company has a durable edge. The investor's job is to tell those readings apart before the next round prices one of them in.

Workforce terms reveal the burn cost

Revenue growth can be expensive if the company must buy it with compute, sales, and engineering. Cognition's workforce terms make that cost visible. The company cut 30 positions and offered the approximately 200 employees who stayed either nine months of pay or a schedule of six-day weeks lasting 80 hours or more. Those terms show the category is being built under pressure, and they give a concrete way to ask whether burn is buying durable usage or just speed.

Burn shows up in hiring, retention, and labor terms. A company can grow revenue while still spending heavily to keep that growth moving. The check is whether the spending is creating a durable advantage or simply outrunning competitors. If the latter is true, the multiple needs a discount.

Overlap and acquisitions change the revenue line

Nvidia Corp. is both an investor in Cognition's Series E and a customer of Cognition's Devin product. That link can help a company land enterprise deals, but it also makes the revenue line less independent. For a growth investor, the test is whether the same buyer would still pay if the investor relationship changed.

Overlap can signal that a strategic buyer sees real value in the product. It can also mean that part of the revenue is tied to a relationship the company does not fully control. The cleaner test is to separate the revenue that would survive without the investor relationship from the revenue that depends on it. If the dependent slice is large, the valuation is partly a bet on the relationship, not just the product.

Cognition bought the Windsurf AI coding editor in July 2025, following the collapse of OpenAI Group PBC's takeover attempt and a technology licensing deal in which Google LLC hired Windsurf's founders. That acquisition broadens the product surface, but it also adds integration work and talent risk. A valuation that includes a recently acquired product should be read as a portfolio, not a single line item.

Run the test before committing capital

• Divide valuation by run-rate revenue at each round; a flat or compressed multiple with revenue growth signals category breadth, while multiple expansion signals single-franchise risk.
• Check burn and workforce terms to see whether spending is buying durable usage or speed.
• Check investor-customer overlap to see how much revenue depends on a relationship the company does not control.