From $26B to $48B in 100 Days: Cognition's $2B Series E Bets Devin Outgrows the Model Wars

Twelve months ago, Cognition was worth $10.2 billion. On Tuesday, the company behind Devin, the autonomous AI software engineer, announced it has raised more than $2 billion at a $48 billion valuation — a Series E that lands just 100 days after its $1B Series D and nearly doubles the company’s price tag in a single quarter.

The round was led by new investors Andreessen Horowitz and Accel, alongside existing backers Founders Fund, General Catalyst, and Avenir. The syndicate stretches to more than three dozen firms: Benchmark, Bessemer, Kleiner Perkins, Greylock, Lightspeed, Altimeter, Bond Capital, Meritech, T. Rowe Price, Lux, 8VC, D1, DST, Bain Capital Ventures, Battery, Ribbit, Stripes, and A* Capital, among others. NVIDIA also joined — as both an investor and, notably, a customer.

The number that makes $48B defensible

Valuations this aggressive live or die on the revenue line, and Cognition delivered one of the steepest curves in applied AI. Run-rate revenue grew from $492 million in May to almost $900 million today — roughly an 80% jump in four months, on top of a year in which ARR climbed from around $73 million to nearly a billion. That trajectory traces the company’s entire valuation history: roughly $350M in early 2024, $2B by April 2024, nearly $4B in March 2025, $10.2B in September 2025, $26B in May 2026, and now $48B.

The multiples remain eye-watering. At the May round, Cognition traded at roughly 53x revenue — well above premium public software peers, and expensive even against Cursor, whose ~$29B valuation on $1B+ ARR implied roughly 29x. The new $48B tag on ~$900M ARR keeps the multiple in a similar zone. Investors are not underwriting a plateau; they are underwriting continued hypergrowth, with all the risk that entails.

What Devin actually does now

Devin launched in early 2024 as a chat-driven agent that could take a ticket and ship code. Where completion tools suggest lines while a human types, Devin operates as a full coding agent: it takes a task description and produces working software autonomously — planning, writing, debugging, and deploying across multi-step workflows.

The Series E announcement makes clear the product has moved beyond that reactive loop into proactive, event-driven work. Three capabilities define the new surface area:

• Devin Automations watches Slack, GitHub, Linear, and similar systems for triggering events and kicks off work without a human opening a chat.
• Devin Auto-Triage takes a first pass at incident investigation, reading alerts and proposing remediation steps.
• Devin Security Swarm scans and maps large codebases to find and triage multi-step exploits.

The pitch is that engineers act as architects setting goals and priorities, while swarms of agents grind through execution. Cognition also says compute budgets will begin to self-allocate toward the highest-impact tasks — an idea that only holds up if agents can reliably judge their own return on investment.

Enterprise logos and the Windsurf foundation

The customer list is what makes this look like infrastructure rather than a demo. Cognition names chip design at NVIDIA, aviation at GE Aerospace, financial services at Citi, automotive at Mercedes-Benz, and AI infrastructure at Modal. Earlier reporting added Goldman Sachs and the U.S. Army and Navy.

Getting there required the Windsurf acquisition. Cognition bought the AI coding startup in July 2025, just days after Google poached its CEO, co-founder, and research leads. The deal broadened the product considerably: Devin handles the asynchronous agent layer while Windsurf adds a synchronous IDE layer (since rebranded Devin Desktop), turning Cognition from a single-agent company into a wider coding platform.

Why stay independent

The strategic rationale in the announcement is blunt: Cognition wants to remain an independent agent lab so it can mix and match the best underlying models — including its own — rather than tie customers to a single provider.

The biggest competitive threat is bundling. Microsoft/GitHub, OpenAI, Anthropic, and Google need only make autonomous coding a default feature of the stacks developers already pay for, rather than outperform Devin feature by feature. Raising $2 billion buys time and optionality to run model training, product development, and go-to-market in parallel.

It also funds an aggressive geographic push. In the past year Cognition opened offices in Washington D.C., Tokyo, Singapore, London, São Paulo, and Madrid, on top of its San Francisco, New York, and Austin hubs — a field-sales footprint, not a self-serve one, consistent with the enterprise contracts driving the ARR curve.

Who feels the pressure

The clearest beneficiary beyond Cognition is the thesis that independent agent companies can survive alongside frontier labs. A year ago, most signals pointed to model makers absorbing this market themselves; this round is a large counter-bet, and NVIDIA’s dual role as investor and customer reinforces it.

The pressure lands on any coding-agent startup that cannot show a comparable revenue slope, and on point-solution vendors in incident response and application security whose workflows Devin’s Auto-Triage and Security Swarm are absorbing.

For teams evaluating tooling, the practical shift is that autonomous agents are moving out of chat windows and into Slack alerts, CI pipelines, and security review loops. Cognition is betting that production-grade autonomy is a now problem — not a someday one — and it has just raised the war chest to prove it.