TL;DR
Cognition AI has raised more than $1 billion at a $26 billion valuation for Devin, its AI coding agent. Revenue grew from $37 million to $492 million in 12 months, with Goldman Sachs, Mercedes-Benz, and the US government among its customers.
Cognition AI has raised more than $1 billion in new funding at a $26 billion valuation, more than doubling its worth since a September round that valued the company at $10.2 billion. The round was co-led by Lux Capital, General Catalyst, and 8VC, with participation from Ribbit Capital, Atreides Management, and Peter Thiel’s Founders Fund. The company has now raised more than $2.5 billion in total funding.
The numbers behind the valuation tell the story. Cognition’s revenue run rate has grown from $37 million in May 2025 to $492 million today, a 13-fold increase in 12 months. The company says it aims to cross $1 billion in annualised revenue later this year. Its customers include Goldman Sachs, Mercedes-Benz, NASA, Santander, and several parts of the US government.
What Devin does
Cognition’s flagship product is Devin, an AI agent designed to automate the programming process end-to-end. Unlike code completion tools that suggest lines or blocks of code while a human developer writes, Devin operates as a full coding agent that can take a task description and produce working software autonomously. It plans, writes, debugs, and deploys code across complex multi-step workflows.
The most striking claim comes from inside Cognition itself. Scott Wu, the company’s co-founder and CEO, said that more than 90% of the company’s internal code is now written by Devin. That figure, if accurate, makes Cognition one of the most aggressive practitioners of its own product in the history of enterprise software. It is a company that has automated its own engineering function using the tool it sells to automate engineering functions at other companies.
Cognition runs Devin on a mix of its own proprietary models and models from OpenAI and Anthropic. Wu framed this as a strategic advantage rather than a dependency. As the model layer gets more competitive, he said, working with a combination of models produces better results than relying on any single provider. Cognition routes customers to the best tools for their specific needs, positioning itself as an orchestration layer rather than a model company.
The AI coding market is on fire
Cognition’s raise lands in the hottest category in venture capital. Cursor, the AI coding editor built by Anysphere, hit $2 billion in annual recurring revenue in roughly three years and was in talks to raise $2 billion at a $50 billion valuation before SpaceX struck a deal in April for the right to acquire the company for $60 billion. OpenAI and Anthropic are both investing heavily in coding capabilities within their foundation models. Salesforce expects to spend $300 million on Anthropic tokens this year, primarily for coding use cases.
The competitive dynamics are unusual. Cognition uses models from OpenAI and Anthropic, the same companies that are building competing coding products. Wu addressed this directly, arguing that Cognition’s value is in the agent layer, the orchestration, planning, and execution logic that sits on top of foundation models, rather than in the models themselves. This is a bet that the model layer will commoditise while the agent layer captures durable value. It is also a bet that OpenAI and Anthropic will not build equally capable agent products that undercut their own API customers.
The Windsurf acquisition and consolidation wave
Cognition is also growing through acquisition. In July 2025, the company acquired the remaining assets of Windsurf, a coding startup that had been the subject of a bidding war between OpenAI and Google. Google ultimately struck a $2.4 billion deal for Windsurf’s top engineering talent and licensing rights, while Cognition picked up what was left, including technology, customers, and employees who chose not to join Google.
The Windsurf deal illustrates the consolidation dynamics at play. AI-native enterprise spending surged 94% year on year in early 2026 while traditional SaaS growth cooled to 8%, and the capital flowing into AI coding tools is driving a wave of acquisitions, talent raids, and competitive positioning that resembles the early days of cloud computing compressed into months rather than years.
Cognition said it plans to use the new funding to refine its models, improve the customer experience, and potentially make more acquisitions. Wu emphasised on Bloomberg Television that the raise allows Cognition to remain independent, a pointed comment given the SpaceX-Cursor deal and the broader trend of AI startups being absorbed by larger platforms.
The independence question
Whether Cognition can stay independent at this scale and growth rate is the open question. The company is valued at $26 billion with $492 million in revenue, a multiple of roughly 53 times. That valuation holds only if growth continues at its current pace and the AI coding market does not compress around a few dominant players. Every major software company is building AI coding capabilities, and the foundation model providers are steadily improving their native coding performance with every model release.
Cognition’s defence is execution speed and the claim that the agent layer, not the model layer, is where enduring value lives. If Devin can genuinely automate 90% of a company’s coding output, the product is not a developer tool. It is a replacement for a significant portion of the software engineering workforce. That is the same logic driving layoffs at Meta and Microsoft, where companies are converting payroll budgets into AI infrastructure spending. Cognition is selling the tool that makes that conversion possible.
Founded in 2023, Cognition went from a demo that went viral on social media to a $26 billion company in less than three years. The speed is unprecedented, but so is the market it is operating in. AI coding is not a niche product category. It is a direct bet on the proposition that software can write itself, and that the companies that automate programming fastest will capture a disproportionate share of the $600 billion global software market.
