Cognition, the startup behind the autonomous coding agent Devin, is on track for $1 billion in annualized revenue based on its September performance, according to Bloomberg, which cited a person familiar with the matter. That would roughly double the company's run rate in about four months and cement AI coding agents as one of the few generative AI categories generating revenue at genuine enterprise scale.
The milestone lands weeks after Cognition raised more than $2 billion at a $48 billion valuation, and in the same week that investors, analysts and regulators are asking harder questions about how much autonomy AI agents should have.
The Numbers
It is worth being precise about what has been reported:
- Bloomberg's story says Cognition is on track to hit a $1 billion annualized run rate based on this month's performance. Cognition declined to comment on the latest figure.
- The company itself said earlier in September that its run-rate revenue had topped $900 million.
- In May, when it raised more than $1 billion at a $25 billion pre-money valuation, Cognition reported a run rate of about $492 million, and said enterprise usage of Devin had grown 50% month over month for six months.
- On September 8, it announced a Series E of more than $2 billion at a $48 billion valuation, led by Andreessen Horowitz and Accel.
A run rate extrapolates a short period of sales to a full year; it is not audited annual revenue, and Cognition has not published a breakdown. Even so, the trajectory is steep. At its latest valuation, investors are paying roughly 50 times the self-reported run rate.
What Is Driving Growth
Much of Cognition's acceleration traces back to its July 2025 acquisition of Windsurf, the AI coding editor. Windsurf brought roughly $82 million in its own recurring revenue and more than 350 enterprise customers, and combined enterprise revenue rose more than 30% in the seven weeks after the deal closed.
Cognition's pitch is that Devin works less like an autocomplete tool and more like a junior engineer that can take a ticket, write code, run tests and open a pull request. Its named customers span finance, automotive, healthcare and government, including Citi, Goldman Sachs, Mercedes-Benz, Dell Technologies, Santander, Elevance Health, the US Army and the US Navy, according to company materials cited in coverage.
A Crowded, Fast-Moving Market
Cognition is not alone, and it is not the largest player:
- Cursor maker Anysphere reportedly reached about $4 billion in annualized revenue by June 2026, and SpaceX completed its $60 billion acquisition of Cursor in August. SpaceX had earlier approached Cognition about a potential takeover.
- Lovable, the Swedish "vibe coding" startup, crossed $600 million in annualized revenue this month, TechCrunch reported.
- Frontier labs including Anthropic and OpenAI continue to push their own coding agents directly to developers and enterprises.
The competition is pushing prices, capabilities and consolidation all at once. Coding is the first knowledge-work domain where agents deliver clear, measurable output — merged code — which makes it easier for enterprises to justify large contracts.
Why It Matters
Cognition's run rate is a data point in a larger argument about whether the AI boom is producing real businesses. This week, Radical Ventures data circulated showing that so-called AI "neolabs" without products or revenue had raised about $24 billion over two quarters. Against that backdrop, a company approaching $1 billion in run-rate revenue from paying enterprises stands out.
Three implications for the AI economy:
- Agents are monetising. Revenue is flowing to products that complete tasks end to end, not just assist humans.
- Valuations remain aggressive. A roughly 50x multiple assumes growth continues at a pace few software companies have ever sustained.
- Governance is now a sales issue. With OpenAI pausing frontier training after agent incidents this week, enterprise buyers will increasingly ask coding-agent vendors how their tools are sandboxed, what credentials they can reach and how actions are audited.
For software teams, the message is that autonomous coding agents have moved from pilot to budget line. For investors, the question is whether Cognition can keep compounding in a market where the biggest rival now sits inside SpaceX and the model providers themselves are competitors.
