Cognition Raises $2 Billion at $48 Billion Valuation, Doubling Revenue in Four Months

Cognition, the company behind the AI coding assistant Devin, has raised $2 billion at a $48 billion valuation, according to TechCrunch. The round was led by Andreessen Horowitz, Accel, Founders Fund, General Catalyst, and Avenir.

The valuation arrives just four months after Cognition's previous fundraise, a Series D that valued the company at $26 billion and raised over $1 billion. That May round was led by Lux Capital, General Catalyst, and 8VC, with support from existing investors, as announced in a blog post titled "Devins in More Places." The $48 billion figure also exceeds expectations reported by Bloomberg in August 2026, when the company was described as being in early talks for a round that could boost its valuation to about $40 billion.

Founded in 2024 by Scott Wu, Cognition develops Devin, which it describes as "the first autonomous software engineer." Devin plans, writes, tests, and ships production code on its own, operating inside a team's existing codebase and toolchain. The product is deployed at major enterprise customers including Mercedes-Benz, NASA, Goldman Sachs, and Citi.

Cognition reported that its annualized run-rate revenue grew from $492 million to $900 million between its May 2026 fundraise and this round. The company is expected to reach $4 billion to $5 billion in annualized revenue by the end of 2026. Annualized run-rate revenue is a common startup metric: it takes revenue from a recent period and projects it over a full year, giving investors a sense of current pace rather than historical totals.

Those revenue figures sit alongside substantial infrastructure costs. Cognition leases an Nvidia server cluster that costs hundreds of millions of dollars annually, which could push its total cash burn to $800 million in 2026. Cash burn is the amount a company spends beyond what it earns, and for AI companies, the biggest line item is often compute power — the processors and data center capacity needed to run large models. The company's acquisition of Windsurf, an IDE platform, announced in July 2025, broadened its product surface beyond the standalone Devin agent into the development environment itself. An IDE, or integrated development environment, is the application where programmers actually write and manage code.

The competitive landscape for AI coding tools has been reshaping rapidly. In April 2026, coding assistant Cursor agreed to sell to SpaceX for $60 billion. Cognition's ability to raise at $48 billion as an independent company indicates that investors see room for multiple large outcomes in the AI coding category, rather than a single winner-take-all dynamic.

The broader context here is worth examining closely. The gap between Cognition's $900 million annualized run-rate revenue and a potential $800 million in annual cash burn is thin. The company's revenue is growing fast, roughly doubling in four months. Its compute costs, tied to leased Nvidia infrastructure, are largely fixed in the near term. If Cognition reaches $4–5 billion in annualized revenue by year-end, the revenue-to-burn ratio improves dramatically, assuming infrastructure costs scale sub-linearly with revenue — that is, that serving more customers does not require proportionally more computing power. That is a significant assumption. Serving more enterprise customers may require additional compute capacity, and inference costs for autonomous code generation are not trivial, given the iterative, multi-step nature of the task. Inference is the process of a trained AI model actually producing output, and autonomous coding requires many back-and-forth inference steps per task.

The $48 billion valuation, set against $900 million in annualized revenue, implies a revenue multiple above 50x. For enterprise software, that is aggressive even by recent AI-sector standards. But investors are pricing a trajectory: four months ago, revenue was $492 million. The multiple compresses quickly if Cognition hits its $4–5 billion year-end target, landing closer to 10–12x, which is in line with high-growth enterprise AI companies.

Cognition's customer base is the less visible part of this story. NASA, Goldman Sachs, Citi, and Mercedes-Benz are not organizations that adopt tools lightly or quickly. Their procurement cycles are long, their security requirements are stringent, and their codebases are large, complex, and often built on older systems. If Devin is genuinely planning, writing, testing, and shipping production code inside those environments, that is a meaningful signal about the maturity of autonomous coding agents, distinct from the copilot pattern of suggesting the next line of code as a developer types.

The Windsurf acquisition adds a dimension to this. Owning the IDE gives Cognition a delivery surface that sits between the developer and the autonomous agent, potentially capturing more of the development workflow and the telemetry that comes with it — telemetry being the usage data generated as developers work. How that integration is progressing since the July 2025 announcement remains an open question.

What is clear is that Cognition has moved from a $26 billion to a $48 billion valuation in four months, backed by a syndicate of five major firms, with revenue that nearly doubled in the same period and a credible path to $4–5 billion by year-end. The cash burn is real, the compute dependency on Nvidia is real, and the valuation multiple leaves no margin for a growth stall. But the trajectory, at least on paper, is steep.