AI Coding Startup Cognition in Talks to Raise at a $40 Billion-Plus Valuation

# AI Coding Startup Cognition in Talks to Raise at a $40 Billion-Plus Valuation

Three months ago, Cognition was worth $26 billion. If its bankers get their way, it will soon be worth more than $40 billion — a jump of more than 50% engineered in roughly the span of a single fiscal quarter.

The maker of Devin, the autonomous "AI software engineer" that writes and ships code with limited human supervision, is in early talks with investors about a new financing round that could value the startup at $40 billion or more, according to reporting from Bloomberg and TechCrunch published on August 12. The discussions are preliminary, and terms could change. But the mere fact that they are happening — barely twelve weeks after Cognition closed a $1 billion round at a $25 billion pre-money, $26 billion post-money valuation in late May — is a vivid marker of how fast money is moving through the AI coding sector.

The Numbers Behind the Leap

The case for the higher price tag rests on a single, fast-moving figure: revenue. When Cognition announced its May raise, founder and chief executive Scott Wu said the company had reached a $492 million annualized revenue run-rate, with enterprise usage of Devin growing 50% month-over-month for six straight months. By the time the new talks surfaced, that run-rate was approaching $1 billion — roughly double the level reported around the prior financing, and reached in a matter of months.

That growth is what lets investors square an eye-watering number with a straight face. At a $40 billion valuation against a $1 billion run-rate, Cognition would be priced at about 40 times revenue. That is an extraordinary multiple by any traditional measure — and yet it is actually lower than the roughly 52 times revenue implied by the $26 billion mark in May. In the strange arithmetic of the current AI boom, doubling your revenue can make a valuation nearly 55% higher look cheaper than the one before it.

Cognition's ascent has been vertiginous even by the standards of the moment. The company raised $400 million at a $10.2 billion post-money valuation last September. It more than doubled that to $26 billion in May. A $40 billion-plus round would nearly quadruple the September figure inside of a year.

From a New York Apartment to the Enterprise

Cognition launched Devin in March 2024, pitching it not as a coding copilot but as a fully autonomous agent capable of taking on complete engineering tasks. Wu has framed the company's mission in expansive terms.

> "We envision a world of software abundance where engineers become architects, solving the most challenging problems and focusing on their creative visions while tasking an army of autonomous agents to support them on everything else."

That vision has translated into a specific commercial niche. Rather than positioning Devin as a human replacement, Cognition has aimed it at the long-tail "grunt work" many developers dislike — migrating applications between platforms, modernizing legacy code, and clearing backlogs of unglamorous tickets. That framing has resonated with large, conservative institutions. Cognition counts Goldman Sachs, Citi, Mercedes-Benz, Dell, Cisco, Palantir, Nubank, and both the U.S. Army and Navy among its customers, and has launched a dedicated government offering.

A pivotal move came in July 2025, when Cognition acquired the remnants of Windsurf — the AI coding startup whose planned $3 billion sale to OpenAI collapsed before Google poached its chief executive and top engineers in a licensing-and-acqui-hire deal. The purchase more than doubled Cognition's ARR and, more strategically, gave it both halves of the AI coding stack: an integrated development environment (Windsurf) for developers who want to stay in the driver's seat, and an autonomous agent (Devin) for tasks they would rather delegate entirely.

An Arms Race With Deep Pockets

Cognition's fundraising tells a broader story about a market that, a year ago, many assumed the large model labs would simply absorb. Anthropic's Claude Code, OpenAI's Codex, and Google's Jules have all captured meaningful share, and each is backed by a foundation-model war chest that a standalone startup cannot match. Cursor-maker Anysphere has ridden the same wave to its own soaring valuation. The prevailing bet was that coding, being so close to the underlying models, would be a feature the labs owned rather than a business independents could defend.

The willingness of top-tier venture firms — Lux Capital, General Catalyst, 8VC, Founders Fund, Ribbit Capital and others backed the May round — to keep writing ever-larger checks is a wager that the opposite is true: that enterprises want a vendor whose entire focus is shipping working software into their existing systems, not a coding surface bolted onto a chatbot. Cognition's month-over-month enterprise growth and its blue-chip customer roster are the strongest evidence yet for that thesis.

The risks are equally plain. Revenue multiples of 40 times leave no margin for a stumble, and much of Cognition's cost structure depends on model providers that are also its competitors. A pricing war, a capability leap by a lab, or a slowdown in enterprise experimentation budgets could all reprice these numbers quickly.

What to Watch

Three things will tell the tale in the coming weeks. First, whether the round actually closes at or above $40 billion, and who leads it — a new marquee investor would signal fresh conviction, while another insider-led round would suggest existing backers are protecting their positions. Second, whether that near-$1 billion run-rate holds its slope or begins to flatten as the easy enterprise pilots convert into harder-to-win production deployments. And third, how Anthropic, OpenAI, and Google respond — because in a market this crowded and this well-funded, the difference between a $40 billion validation and a $40 billion overhang may come down to who ships the better agent next.