Analysis
Devin Writes 89% of Its Own Code. Investors Just Valued That at $26 Billion.
Cognition's Series D closes at a $26 billion post-money valuation after Devin's annualised revenue grew from $37 million to $492 million in twelve months. The round is the clearest signal yet that independent AI coding agents can survive in a market where every frontier lab is building one.
By Michael C ·
Eight months ago, Cognition AI was valued at $10.2 billion. On 27 May 2026, the company announced it had closed a Series D at a $26 billion post-money valuation, raising more than $1 billion in a round led by Lux Capital, General Catalyst, and 8VC. The round also drew in Founders Fund, Elad Gil, Ribbit Capital, Atreides, and Layer Global. The headline number is striking, but the metric underneath it is more so: Devin, Cognition's autonomous AI software engineer, has grown from $37 million in annualised revenue to $492 million in twelve months — a 13x increase, with enterprise usage expanding more than 10x since the start of 2026.
Revenue went from $37 million to $492 million ARR in twelve months. Enterprise usage grew more than 10x. Devin now writes more than 89% of Cognition's own codebase.
The most striking detail in Cognition's announcement is not the valuation or the revenue figure — it is the fact that Devin now writes more than 89% of Cognition's own codebase. The company that built an AI software engineer is itself being built by that AI software engineer. This is not a marketing claim about productivity improvements; it is a structural fact about how Cognition operates. Scott Wu, Cognition's CEO, has been careful to frame this as a collaboration rather than a replacement: 'AI coding agents shouldn't replace humans,' he told TechCrunch. But the operational reality at Cognition is that the ratio of AI-generated to human-generated code has crossed a threshold that most software companies have not yet approached.
The Competitive Landscape That Should Have Killed Cognition
When Cognition raised its $400 million round in September 2025, the conventional wisdom in the AI industry was that independent coding agent startups were in trouble. Anthropic had launched Claude Code. OpenAI had released Codex. Google had acquired Windsurf — the AI coding tool that had been Cognition's most direct competitor — in a deal that also brought much of Windsurf's team to Google. The model makers were moving down the stack, and the assumption was that they would commoditise the coding agent layer the same way they had commoditised earlier AI application categories.
That assumption has not played out, at least not yet. Cognition's revenue growth through this period suggests that enterprise customers are not simply defaulting to the coding agent bundled with their frontier model subscription. They are making deliberate purchasing decisions in favour of a dedicated coding agent platform. The reasons are not entirely clear from public disclosures, but the customer list — Mercedes-Benz, NASA, Goldman Sachs, Santander — suggests that large enterprises with complex, heterogeneous codebases are finding value in a purpose-built agent that is not optimised for general-purpose tasks.
The Windsurf Acquisition and What It Means for the Market
Cognition's announcement noted that it had acquired the remaining assets of Windsurf after Google's acqui-hire took most of the team. This is a notable detail: Cognition is not just surviving the consolidation of the AI coding market — it is picking up the pieces of a competitor that was absorbed by one of the largest technology companies in the world. Whether the Windsurf assets add meaningful capability to Devin or serve primarily as a defensive acquisition to prevent another competitor from using them is not clear from the announcement.
The broader market context matters here. Cursor, the AI coding tool built by Anysphere, has reached $2 billion in annualised revenue and is reportedly raising at a $30 billion valuation. GitHub Copilot continues to grow within Microsoft's enterprise base. The AI coding market is not winner-take-all — multiple products are scaling simultaneously, which suggests that the market itself is expanding faster than any single product can capture. Cognition's $492 million ARR at a 53x revenue multiple compares to Cursor's $2 billion ARR at approximately 30x, a premium that reflects either higher growth expectations for Devin or a different risk profile.
What the $1 Billion Will Be Used For
Cognition's announcement described the use of proceeds in broad terms: expanding the Devin platform, growing the enterprise sales team, and investing in research to extend Devin's capabilities. The company has been relatively quiet about its technical roadmap, but the emphasis on enterprise usage growth suggests that the near-term priority is deepening relationships with large customers rather than launching new consumer-facing products.
The 50% month-over-month revenue growth that Cognition has sustained for the past six months is not a rate that can continue indefinitely at this revenue scale. At $492 million ARR, sustaining 50% monthly growth would require reaching approximately $7 billion in annualised revenue by the end of 2026 — a figure that would make Cognition one of the fastest-growing enterprise software companies in history. The more likely trajectory is a deceleration as the company moves from early adopters to mainstream enterprise customers, who typically require longer sales cycles and more extensive integration work.
The Deeper Question: What Does $26 Billion Buy?
The valuation multiple on Cognition's round is high by any historical standard for enterprise software. At 53x annualised revenue, investors are pricing in not just continued growth but a structural position in the AI coding market that is difficult to displace. The thesis, implicitly, is that Devin is not just a product — it is a platform that will become increasingly embedded in enterprise software development workflows, creating switching costs and data advantages that make it defensible against both frontier lab competition and other independent coding agents.
Whether that thesis is correct will become clearer over the next twelve to eighteen months. The AI coding market is moving fast enough that a product that is dominant today could be marginalised by a capability jump from a frontier model. But Cognition's revenue growth, customer list, and the fact that Devin is now writing most of its own code suggest that the company has built something that enterprise customers find genuinely useful — and that is a more durable foundation than a valuation multiple alone.