Cognition AI just closed a Series D round north of $1 billion, pushing its post-money valuation to $26 billion. That’s up from $10.2 billion following a $400 million raise in September 2025, meaning the company more than doubled its valuation in roughly eight months.
For a company founded in November 2023, total funding now exceeds $2.5 billion.
What Cognition actually does
Cognition builds Devin, widely recognized as the first autonomous AI software engineer. Launched in March 2024, Devin doesn’t just autocomplete code snippets. It handles entire engineering tasks end to end, from understanding specifications to writing, testing, and deploying code.
Cognition’s annualized recurring revenue has ballooned from $37 million in May 2025 to $492 million. Devin now writes approximately 90% of Cognition’s own internal code.
CEO Scott Wu has described Devin as a “tireless, skilled teammate” that frees human engineers to focus on higher-level design and architecture work.
The Series D was led by Lux Capital, General Catalyst, and 8VC. Previous backers Founders Fund and Elad Gil participated again, and Ribbit Capital came in as a new investor. The client roster now spans from Goldman Sachs and Citibank to small engineering teams.
The crypto connection most people missed
Cognition didn’t start out as a general-purpose AI company. Its early efforts were rooted in the cryptocurrency space before pivoting toward broader AI applications. The founders, Scott Wu, Steven Hao, and Walden Yan, initially cut their teeth on crypto-related projects before realizing the underlying technology had far wider commercial potential.
Devin has been deployed in blockchain tooling projects, including Sui integration through Crossmint’s GOAT SDK.
Cognition also recently acquired Windsurf, an AI code editor, expanding its toolkit beyond Devin’s autonomous capabilities.
What this means for investors and the broader market
The ARR trajectory is the key metric here. Growing from $37 million to $492 million in a single year suggests enterprise adoption is accelerating. When banks like Goldman Sachs and Citibank are paying customers, it signals that even the most risk-averse institutions see autonomous AI engineering as production-ready rather than experimental.
The risk, naturally, is concentration. When one AI agent writes 90% of a company’s code, you’re placing enormous trust in the reliability and security of that system. A subtle, systematic flaw in AI-generated code could compound across thousands of deployments before anyone notices. For blockchain applications, where code is often immutable once deployed, this risk is amplified considerably.