Cognition AI Raises $1B at $26B as Devin Revenue Climbs From $37M to $492M ARR in One Year
Cognition AI closed a $1B round at a $26B pre-money valuation, with CEO Scott Wu disclosing annualised revenue climbed from $37M to $492M over the past year — a 13x run-rate increase driven by production deployments at Goldman Sachs and Mercedes-Benz.
The round, anchored around Devin — Cognition’s flagship autonomous coding agent — positions the company alongside Cursor and Blitzy as one of the highest-valued independent coding-agent plays, even as the broader coding tool market has seen consolidation around Claude Code and Codex.
What Devin Actually Does
Wu frames Devin not as a model but as an agent infrastructure layer. Rather than committing to a single foundation model, Cognition routes tasks across OpenAI and Anthropic depending on the work type. The agent handles planning, testing, deployment, and multi-step workflow execution inside sandboxed environments, with repo context and terminal access baked in.
“There’s about 30-35 million software engineers in the world today,” Wu said. “We want to make all of them 10 times more efficient, and then we think there is a lot more than 10 times more software to build.”
The model-agnostic positioning is deliberate. It insulates Devin from model deprecations and lets Cognition arbitrage capability differences across providers — routing creative synthesis to Claude and code execution to GPT-5.x as benchmarks shift.
Windsurf Acquisition and the Consolidation Context
Cognition has also absorbed the remains of Windsurf, the coding startup whose top talent and IP were acquired by Google in a $2.4B deal. The shell acquisition adds Windsurf’s remaining engineering context and user base without the team that made it.
The acquisition fits the broader coding-agent consolidation arc: Google absorbed Windsurf’s model capability, Cognition absorbed the distribution and brand. Cursor remains independent at $2B ARR and $50B valuation.
Revenue Reality Check
At $492M ARR, Cognition is the fastest-scaled B2B SaaS company at this stage on record, though the numbers require context. Enterprise AI ARR is notoriously front-loaded with pilot and commitment spend that can contract on renewal. Both Goldman Sachs and Mercedes-Benz deployments are early-stage production rather than org-wide rollouts.
What distinguishes Devin from the broader agent market is its positioning as infrastructure rather than a point tool — selling to engineering platforms and enterprise IT rather than individual developers. That targeting explains the Goldman/Mercedes mix over the startup-skewing CLaude Code user base.
Competitive Stack
| Agent | ARR | Valuation | Approach |
|---|---|---|---|
| Devin (Cognition) | $492M | $26B | Model-agnostic, enterprise infra |
| Cursor | $2B | $50B | IDE-native, individual developers |
| Claude Code | N/A (Anthropic rev) | $900B | Single-model, platform |
| Blitzy | N/A | $1.4B | 1,000-agent parallelism |
Cognition’s bet is that the long-term winner is not the best model or the best IDE, but the best agent harness that can swap the underlying intelligence as the frontier shifts.