Cognition Raises $1B at $26B Valuation: Devin Now Writes 90% of Its Own Code

Cognition raised over $1 billion on May 27 at a $26 billion post-money valuation, up from $10.2 billion just eight months earlier. Lux Capital, General Catalyst, and 8VC co-led the round, with Ribbit Capital, Atreides, and Peter Thiel’s Founders Fund joining in. Total funding now stands at $2.5 billion.

The growth numbers are the real story. Annualized revenue run-rate reached $492 million, up from $37 million twelve months ago. Monthly enterprise usage has grown 50% for six consecutive months. The company is targeting $1 billion in ARR before the end of 2026.

Goldman Sachs, NASA, and Mercedes-Benz Are Paying Customers

Cognition’s flagship product, Devin, is an autonomous AI software engineer. It reads codebases, plans tasks, writes code, runs tests, and debugs, all without requiring a human engineer to supervise each step. The claim is a complete development loop, not just tab-complete.

The customer list spans industries well beyond tech: Goldman Sachs, Citi, Palantir, Cisco, Dell, NASA, Mercedes-Benz, and Santander. NASA and a German automaker showing up on this list matters. It suggests Devin is operational in non-software engineering contexts, validating the thesis that agentic coding tools can work where the codebase is large, domain-specific, and high-stakes.

90% of Cognition’s Own Code Is Written by Devin

CEO Scott Wu stated in the funding announcement that more than 90% of the code committed at Cognition is written by Devin itself.

This serves two purposes at once. Externally, it is a demonstration of product confidence. Internally, it is a live signal that a meaningful portion of standard software engineering tasks can be systematically handed off to an AI agent at production quality.

Earlier this year, Cognition acquired the remaining assets of Windsurf, integrating AI code editor capabilities into its platform. The current product sits in two layers: a developer tooling layer, and an agent layer capable of executing end-to-end task flows. Few direct competitors hold both.

The Competitive Stakes

The AI coding market has attracted serious capital. Anysphere’s Cursor closed a $2.3 billion round this year at a $29.3 billion valuation. Anthropic treats agentic coding as a core use case for Claude. GitHub is shifting Copilot to usage-based AI Credits billing on June 1, a structural signal that complex agent usage is becoming the primary billing driver, not just inline suggestions.

Cognition’s bet is that the agent layer, where an AI completes full engineering tasks rather than assisting with fragments, has a deeper moat than tooling alone. The counterargument is that agent task completion rates have a ceiling, and enterprise customers evaluating on proof-of-concept trials will hit that ceiling before they commit to wide deployment.

The $1 billion ARR target by year-end is a concrete test of which side of that argument holds.

---

Further Reading:

• Meta Cuts 8,000 Jobs: The Cost of AI Restructuring
• OpenAI Files Confidential S-1 for IPO

Sources:

• AI coding startup Cognition raises $1B at $25B pre-money valuation (TechCrunch, 2026-05-27)
• AI Coding Startup Cognition Raises $1 Billion at $26 Billion Value (Bloomberg, 2026-05-27)

If this was useful, subscribe to the newsletter for weekly AI PM insights and GenAI case studies.