Cognition raises more than $2 billion at a $48 billion valuation as Devin revenue approaches $900 million

Cognition, the developer of the Devin software agent, has raised more than $2 billion at a $48 billion valuation. The Series E was led by new investors Andreessen Horowitz and Accel, joined by existing investors Founders Fund, General Catalyst and Avenir. Since the May round, annualized run-rate revenue has grown from $492 million to nearly $900 million, roughly 1.8 times in four months. The valuation works out to about 53 times annual revenue.

The list of participants is unusually long: Benchmark, Bessemer, Kleiner Perkins, Greylock, Lightspeed, Altimeter, Bond Capital, T. Rowe Price, DST, Bain Capital Ventures and several dozen more funds. Early-stage venture firms sit alongside investors who typically come in closer to a public listing. For a company founded in 2024, this is an unusual mix.

From chat to proactive agent

The main product shift Cognition points to is how work gets started. Previously, a developer opened a chat and assigned a task. Now Devin starts on its own, triggered by events in work systems.

• Devin Auto-Triage performs the first-pass analysis of incidents.
• Devin Security Swarm finds vulnerabilities and sorts them by priority.
• Devin Automations launches tasks from events in Slack, GitHub, Linear and other systems, without a separate conversation for each one.

The company names its customers specifically: chip design at NVIDIA, aviation at GE Aerospace, finance at Citi, automotive at Mercedes-Benz, AI infrastructure at Modal. The set is diverse, ranging from regulated banks to companies that build accelerators and clouds for AI themselves.

Key figures of the round

• Raised: more than $2 billion, at a $48 billion valuation.
• Run-rate revenue: $492 million in May, nearly $900 million now.
• Lead investors: Andreessen Horowitz and Accel (both new).
• New offices over the past year: Washington, Tokyo, Singapore, London, São Paulo, Madrid. Existing hubs: San Francisco, New York, Austin.

An independent agent lab

Cognition positions itself as an independent agent lab: the company selects and combines models for each task, including its own, and does not lock customers into a single provider. This is a deliberate contrast with players whose coding agent is built into their own model lineup. For a large customer, the argument is clear: model quality changes from quarter to quarter, and a contract for an agent should not depend on the fate of one supplier.

The company goes on to describe an "era of self-driving software": agents are proactive by default, software improves itself, and compute budgets flow on their own to the most valuable tasks. In this scheme, the engineer sets goals and priorities, while execution goes to agents. This is the company's own vision, not a measured result. Run-rate, moreover, is calculated from current contracts and is not the same as recognized revenue. The data was provided by Cognition itself, and the article contains no independent audit.

What teams should check before adopting

If you are considering Devin or a similar agent, it makes sense to start with narrow tasks where the result is easy to verify: first-pass incident analysis and the vulnerability queue. Here the agent saves on-call engineers hours, and a mistake is cheap as long as the final decision stays with a human.

Next, write down which events and systems the agent receives work from, and restrict its access rights as strictly as you would for a new employee. Proactive launching from Slack or GitHub is convenient, but it turns any configuration error into a mass one.

Finally, ask the vendor which models run under the hood and whether they can be swapped. The claimed provider independence is valuable only if it is written into the contract terms.