Face-free close-up of a black mechanical keyboard on a dark desk with orange key lighting and notebooks in soft focus — coding desk still for Cognition Series E Devin coverage

Cognition’s $2B Series E: AI coding isn’t winner-take-all

I’m wrapping this one from Mexico on the evening slot, and the AI coding race just got another loud cheque. Cognition — the lab behind coding agent Devin — said it raised over $2 billion in a Series E at a $48 billion valuation. That’s the company’s own post on cognition.com/blog/series-e, and TechCrunch, Reuters, and PYMNTS all match the size and price.

The round was led by new investors Andreessen Horowitz and Accel, alongside existing backers Founders Fund, General Catalyst, and Avenir (company blog / Reuters). TechCrunch notes this lands about four months after Cognition’s prior raise at $26 billion. PYMNTS pegs that May 27 Series D at more than $1 billion raised. From Mexico, my read is simple: investors still think AI coding has room for more than one big winner.

ARR jump — and what Devin is shipping

Since that May round, Cognition says run-rate revenue grew from $492 million to almost $900 million (company blog / TechCrunch). The company pitch is engineers as architects with agent swarms doing execution. New surfaces called out in the Series E post: Devin Auto-Triage for a first pass on incidents, Devin Security Swarm for finding and triaging vulnerabilities, and Devin Automations that start work from Slack, GitHub, Linear, and similar events without opening a fresh chat each time (cognition.com).

Cream-paper schematic of Cognition Series E funding stack: over $2B Series E at $48B valuation, a16z and Accel as new leads with Founders Fund General Catalyst and Avenir, May step-up from $26B with ARR from $492M to about $900M, plus Devin product flow from engineer as architect through agent swarms Auto-Triage Security Swarm and Automations
Series E → valuation step-up → Devin product surface. Schematic: Tech & AI Pulse.

Own models, big logos, and why a16z is back

Cognition says it wants to stay an independent agent lab — picking and combining models, including its own, rather than locking customers to one provider (company blog). TechCrunch adds that, like Cursor before the SpaceX deal, Cognition is training on open-source bases to cut spend on OpenAI and Anthropic APIs, and that Cognition leases a large Nvidia cluster that can push cash burn high this year. Named customers on the company post include chip design at NVIDIA, aviation at GE Aerospace, financial services at Citi, automotive at Mercedes-Benz, and AI infra at Modal; TechCrunch also lists NASA and Goldman Sachs. Founded in 2024 by Scott Wu, with newer offices in D.C., Tokyo, Singapore, London, São Paulo, and Madrid on top of SF, New York, and Austin (company blog).

The part I’m watching from Mexico: a16z already made a killing on Cursor’s SpaceX exit, and it’s leading money into a Cursor rival anyway (TechCrunch). That’s the “not winner-take-all” signal in plain English. Sources: Cognition Series E post, TechCrunch, Reuters, PYMNTS.

Hero image: mechanical keyboard on a desk by Maria on Unsplash (Unsplash License). Cropped, graded, and lightly grained by Tech & AI Pulse. Face-free coding still — no people, no logos.