Devin Hits $492M ARR as Cognition Raises $1 Billion at $26 Billion Valuation

When Cognition launched Devin in March 2024 as the world’s first “AI software engineer,” the announcement was met with a familiar mix of Silicon Valley hype and skepticism. Early demos showed the AI autonomously navigating codebases, writing tests, and deploying fixes — impressive, but benchmarks were contested and real-world adoption was thin. Barely two years later, the skepticism has been obliterated by revenue numbers that defy normal growth curves.

Cognition has raised $1 billion at a $26 billion post-money valuation, the company confirmed on May 27, 2026. The round was led by Lux Capital, General Catalyst, and 8VC, with participation from Founders Fund, Elad Gil, Soma Capital, and a roster of existing backers. The valuation represents a roughly 2.5x increase from Cognition’s last round — a $400 million raise at a $10.2 billion post-money valuation closed just eight months ago in September 2025.

The Numbers That Explain Everything

The fundraising mechanics are secondary to what’s driving them. Devin’s annualized revenue run rate has climbed from $37 million in May 2025 to $492 million in May 2026 — a 1,230 percent increase in twelve months. Enterprise usage has grown more than 10x this year alone, with month-over-month growth holding at approximately 50 percent for six consecutive months.

To put this in context: Salesforce needed roughly eight years to reach $400 million in annual revenue after its founding. Devin got there in about 24 months from launch, and it did so in a market segment — enterprise software development — that was considered among the hardest for AI to crack because of the complexity, security requirements, and institutional caution involved.

Cognition also disclosed a striking internal metric: between 89 and 90 percent of the company’s own codebase is now written by Devin. The company is not just selling an autonomous coding product; it is running its entire engineering operation on one.

Enterprise Customers at Scale

Cognition counts Mercedes-Benz, NASA, Goldman Sachs, and Santander among its enterprise customers — a list that cuts across industries and geographies in a way that signals Devin has crossed the chasm from developer curiosity to institutional software infrastructure.

The diversity is significant. NASA’s adoption indicates Devin can operate in high-stakes, heavily regulated environments where code correctness is not negotiable. Goldman Sachs’s use signals that financial institutions, which have historically been the most conservative adopters of AI in production systems, are now treating autonomous coding agents as operational tools rather than experiments.

Mercedes-Benz’s involvement hints at use cases beyond traditional software companies — vehicle software, embedded systems, and manufacturing automation represent enormous addressable markets that Cognition has barely begun to capture.

What Devin Actually Does (and Doesn’t Do)

The product is most accurately described as a long-horizon coding agent rather than a simple coding assistant. Where tools like GitHub Copilot suggest the next line or function, Devin takes multi-step tasks — “implement this feature,” “fix this class of bugs across the codebase,” “write and run tests until coverage reaches 90 percent” — and executes them over extended sessions that can last hours.

Devin operates in a sandboxed environment with access to a browser, terminal, and code editor. It can write code, run it, observe the output, debug failures, and iterate — all without human intervention on each step. Users define the objective and review-approval gates; Devin handles the implementation loop.

The autonomous nature of this workflow is both the product’s core value proposition and its primary source of organizational friction. Cognition’s own research, shared with investors, documents what engineers internally call the “delegation gap” — developers trust AI for assistance on roughly 60 percent of their work but fully delegate only 0 to 20 percent. The gap isn’t primarily technical; it’s about verification costs. Engineers need to be confident that AI-produced code meets quality, security, and correctness standards before it enters production, and building those verification workflows takes time.

Cognition’s enterprise contracts now include dedicated onboarding to help engineering organizations design the review scaffolding around Devin — a professional services layer that adds stickiness and recurring revenue beyond the core subscription.

The Competitive Landscape

The $1 billion raise comes as the autonomous coding agent market has become one of the most contested spaces in enterprise software. GitHub Copilot remains the most widely deployed AI coding product by install base, but it operates primarily as an in-editor assistant rather than an autonomous agent. Cursor, which raised its own large round earlier this year, has built a passionate following among individual developers but has been slower to penetrate large enterprise accounts.

OpenAI’s Codex and Anthropic’s Claude Code both operate in overlapping territory, though they function primarily as API-accessed models that third-party developers integrate into their own tooling, rather than as packaged enterprise products with the account management and compliance infrastructure that large organizations require.

Cognition’s differentiation increasingly rests on three pillars: the depth of enterprise integrations and security certifications it has accumulated, the reliability record that Devin has built across its installed base, and the compound advantage of training data generated by real enterprise usage at scale. Every production deployment generates proprietary signal that Cognition uses to improve future model iterations.

What $1 Billion Buys

The proceeds are earmarked for three areas: model research to extend Devin’s task horizon and accuracy on complex multi-file changes; enterprise infrastructure to support compliance requirements in regulated industries; and international expansion, with European and Asia-Pacific markets identified as the next major growth opportunities.

The funding also provides capital to hire the model researchers and safety engineers needed to maintain Devin’s lead as the competitive field intensifies. Cognition currently employs roughly 200 people, a headcount that founders Scott Wu and Steven Hao have said they plan to scale carefully to avoid the organizational friction that slows model iteration cycles.

A Benchmark for What’s Possible

The Cognition story matters beyond the company itself because it establishes a reference point for the autonomous AI agent market. Revenue at this scale, growing at this rate, in enterprise accounts of this caliber, demonstrates that the transition from AI coding assistants to AI coding agents is not a future event being priced into a speculative valuation — it is happening now, at production scale, with auditable financial outcomes.

The $26 billion post-money valuation implies investors believe Cognition can capture a substantial slice of the global software development market, which Gartner estimates at over $800 billion in annual spending on developer compensation alone. Even a modest productivity-driven displacement of that spending — say, 10 percent over five years — represents an addressable market an order of magnitude larger than Cognition’s current revenue.

Whether Devin ultimately captures that opportunity, or whether a better-capitalized competitor overtakes it, the pace of adoption it has demonstrated in the past twelve months suggests the market transition is real, irreversible, and moving faster than most enterprise software buyers anticipated.