Cognition Raises Over $1 Billion at $26 Billion Valuation as Devin's Revenue Hits $492M Run‑Rate in Under Two Years

Cognition, the San Francisco‑based company behind Devin, the autonomous AI software engineer, has raised over $1 billion in a Series D funding round at a $26 billion post‑money valuation. The round was co‑led by Lux Capital, General Catalyst, and 8VC, with participation from Ribbit Capital, Atreides Management, Founders Fund, and a number of existing investors. Total capital raised since founding in 2023 now stands at more than $2.5 billion.

The valuation more than doubles Cognition's last mark of $10.2 billion, set just eight months ago in September 2025 when it raised $400 million from a round led by Founders Fund. The pace of appreciation, 2.5x in under a year, puts Cognition among the fastest‑rising valuations in the history of the enterprise software market, a fact that reflects both genuine commercial momentum and the degree to which investors are pricing the agentic AI coding category as winner‑take‑most.

The Revenue Story That Justifies the Valuation

The business metrics behind this round are not speculative. Cognition's annualized revenue run‑rate has reached $492 million, up from $37 million just twelve months ago in May 2025, and from approximately $1 million when Devin launched in March 2024. Enterprise usage has grown more than 10x since January 2026 alone.

The current customer roster includes organizations that evaluate software infrastructure at the highest standards of performance and security requirements. Goldman Sachs, Citi, Santander, and Deutsche Bank are among the financial services clients. Mercedes‑Benz and Dell are on the commercial enterprise side. The US Army and US Navy are active users, making Cognition one of the few AI coding platform companies that has passed the procurement and security review process for federal defense deployment.

Fast‑growing startups have also adopted Devin across their internal engineering operations, including Exa, Modal, Eight Sleep, and OpenRouter, the AI model routing platform that itself announced a $113 million raise earlier this week. The diversity of the customer base, from global banks to defense agencies to AI‑native startups, reflects a product that has moved beyond the developer‑enthusiast audience that greeted the March 2024 launch.

Internally, Cognition's own engineering team now has more than 90 percent of its code written by Devin. CEO Scott Wu cited this openly as evidence of the product's reliability: a company whose business depends on the credibility of its AI coding agent would not trust that agent with its own production codebase unless the quality bar was genuinely high.

From Contested Demo to $492M ARR

The trajectory of Devin's commercial adoption is worth tracing because it is more complicated than the current funding numbers suggest.

When Cognition launched Devin in March 2024, the announcement drew significant attention and significant skepticism simultaneously. Several engineers published detailed analyses arguing that the launch demo had been edited to compress timelines and that the underlying task completion rate was lower than presented. Those critiques were fair at the time. The demo showed Devin completing a software development task end to end, from reading a GitHub issue through writing, testing, and deploying the solution. Independent testers found that the live product required more human intervention than the demo implied.

Two years later, that early‑stage gap between demo and reality has closed to the point where Goldman Sachs and the US Army are paying customers. Devin has been rebuilt iteratively across multiple versions, with each release improving task completion rates on the SWE‑bench and related benchmarks that measure AI coding agent capability against real software engineering workloads.

The competitive landscape the company now operates in has also intensified considerably. Anthropic's Claude Code, OpenAI's Codex, and GitHub Copilot Workspace are all building toward the same agentic coding capability. Cursor, the AI code editor that recently announced a deal with SpaceX, is valued at $9 billion and competes for developer workflow budget. Cognition's response to that competitive pressure is embodied in its acquisition of Windsurf, the AI coding environment formerly known as Codeium, which Cognition completed earlier this year. That acquisition brought a large developer user base and a complementary product surface into the Cognition platform.

The Multi‑Model Strategy

One of the more interesting disclosures accompanying this raise is how Cognition actually runs Devin in production. Rather than relying on a single underlying AI model, Devin uses a mix of Cognition's own purpose‑built models and outputs from OpenAI and Anthropic, routed to whichever combination performs best for the specific task type.

Wu's framing on this choice was direct: as frontier model competition intensifies and the gap between leading models narrows, the systems that can intelligently combine multiple models will outperform those that are locked into a single provider. That architecture makes Cognition model‑agnostic at the infrastructure level while maintaining a proprietary layer in the software engineering logic that routes, orchestrates, and validates outputs across the underlying model mix.

This positioning also explains why Cognition is not threatened by the continued improvement of OpenAI and Anthropic's base models. Better models mean better raw material for Devin's orchestration layer, not a substitute for the product itself.

Use of Capital

The $1 billion Series D will fund four priority areas:

• AI model research and improvement, expanding Cognition's proprietary model capabilities for software engineering tasks
• Infrastructure scaling to support the 10x enterprise usage growth that has already occurred this year and the acceleration expected through 2026
• Customer experience and product development, expanding Devin's task coverage and integrations across enterprise software development environments
• Additional acquisitions, with Cognition explicitly signaling it is looking for further consolidation opportunities in the AI coding category following the Windsurf deal

The company's target is to cross $1 billion in annualized revenue before the end of 2026, starting from the current $492 million run‑rate. That target requires approximately doubling the current revenue pace across the remaining seven months of the year, an aggressive timeline that the 10x usage growth figure since January suggests is within reach if enterprise deal velocity holds.