Did you know Cognition, the company behind the AI coding agent Devin, nearly doubled its value in about four months?
That’s right! On September 8, Cognition announced a Series E round of more than $2 billion. The deal pushes the Cognition valuation to $48 billion. Back in May, investors valued it at $26 billion. So the price tag jumped by $22 billion in one summer. Even by AI standards, that is a fast climb.
Meanwhile, the round was bigger than many expected. According to Techzine, Bloomberg had reported that Cognition was looking for about $1 billion at a $47 billion valuation. Instead, it doubled the amount and still beat the target price.

Table of Contents
Who is writing the checks?
Two new investors led the round: Andreessen Horowitz and Accel. In addition, existing backers Founders Fund, General Catalyst and Avenir came back for more. The list of other names is long. For example, it includes Benchmark, Bessemer, Kleiner Perkins, Greylock, Lightspeed and T. Rowe Price.
Why so much interest? Mostly, it comes down to revenue. Cognition says its run-rate revenue grew from $492 million in May to almost $900 million. That is roughly 83% growth in a few months, as AI Business points out. However, keep in mind that the company reports this number itself. Run-rate revenue is also a forecast of the next year, not money already in the bank.
- Round: Series E, more than $2 billion
- Valuation: $48 billion, up from $26 billion in May
- Leads: Andreessen Horowitz and Accel
- Run-rate revenue: almost $900 million, up from $492 million (company reported)
- Named customers: NVIDIA, GE Aerospace, Citi, Mercedes-Benz and Modal
You see, Devin is not autocomplete
Most people met AI coding through autocomplete. You type, and the tool guesses the next line. Devin works differently. You give it a task. Then it inspects the codebase, writes code, runs tests and works through problems on its own. Cognition first showed Devin in 2024. Since then, it has also added the Windsurf code editor to its lineup.
The new funding comes with a broader product push. Specifically, Cognition highlights three new tools:
- Devin Auto-Triage takes a first pass at investigating incidents.
- Devin Security Swarm finds and triages vulnerabilities.
- Devin Automations starts work from events in Slack, GitHub, Linear and other systems, so nobody has to open a chat for each task.
In other words, Devin is moving from “answer my prompt” to “watch my systems and act.” That is a big change in how engineers might work.
We believed engineers should operate more like architects and delegate execution to swarms of agents.
Cognition, Series E announcement
From typing help to agent swarms
How AI coding tools are evolving
Each step hands more of the engineering task to the machine
Figures from Cognition’s Series E announcement, September 8, 2026. Revenue is company reported.
Diagram by CVisiona | cvisiona.com
Diagram: how AI coding tools are moving from autocomplete to agent swarms, and Cognition’s Series E numbers, by CVisiona
But here’s the twist
Coding may be the first white-collar job where AI agents face a truly measurable test. After all, the software either works or it doesn’t. Tests pass or fail. Bugs get fixed or they stay. No PowerPoint required.
That is why coding has become one of the hottest corners of the agent economy. Copilots helped programmers type faster. Coding agents, on the other hand, try to finish entire engineering tasks. As a result, buyers can compare tools on real output, not just on demos.
And the competition is fierce. In August, SpaceX completed its $60 billion acquisition of Cursor, according to Bloomberg Government. Meanwhile, Anthropic’s Claude Code and OpenAI’s Codex bring coding agents straight from the model makers. We covered a similar land grab in our look at Meta’s Muse Code. So Cognition is not just racing startups. It is racing the labs that build the models it uses.
Cognition’s bet: stay independent
This is where Cognition’s strategy matters. The company calls itself an “independent agent lab.” It says it can choose and combine the best models for each job, including its own. Therefore, customers are not locked into one model provider. That pitch should appeal to large banks and carmakers that want options.
Still, independence is expensive. According to Channel Insider, citing The Information, Cognition expects to burn about $800 million in cash this year. Much of that goes to leased computing power. We have explained how unchecked inference bills can break AI startups, and coding agents are hungry users of compute.
The skeptic’s view
Cognition also carries some baggage. For instance, a popular Internet of Bugs video challenged Devin’s original 2024 demo. In 2025, chief executive Scott Wu drew criticism for telling Windsurf staff to expect 80-hour weeks, as Channel Insider recalls. Moreover, a $48 billion valuation on under $1 billion of run-rate revenue leaves little room for error. We asked a similar question in why AI startups are worth $1B and why they shouldn’t be.
Then there is the human question. If agents do more of the work, what happens to junior developers? Cognition’s answer is that engineers become architects. However, architects usually learn by building first. That tension is not solved yet.
What to watch next
So what should teams do now? First, test coding agents on real tickets, not on demos. Next, measure how many changes pass review without a human rewrite. Finally, keep a close eye on security, because an agent with access to your systems needs strict limits.
The question used to be, “Will AI replace software engineers?” The more interesting question now is how many AI engineers every human engineer will manage.
Sources & Credits
- Text credit & news reporting source: Cognition, Series E announcement (September 8, 2026)
- News reporting source: Channel Insider (September 2026)
- News reporting source: Techzine (September 9, 2026)
- News reporting source: AI Business (September 10, 2026)
- Background: Bloomberg Government, SpaceX completes $60 billion Cursor acquisition (August 2026)
- Image: Cognition. Diagram: CVisiona
Enjoyed this breakdown? Stay ahead of the curve with our technical analyses and news. Join our community by subscribing to the CVisiona Newsletter for weekly deep dives, and subscribe to our CloudMelon Vis YouTube Channel for video breakdowns of Cloud-Native, AI, and Tech Enterprise news.
Question for you: Would you trust an AI agent to fix your production bugs on its own? Tell us in the comments.
















Anthropic’s Record IPO Raised $81 Billion. Its Suppliers Own a Big Slice
OpenAI’s Agents Turned a German Wiki Into a Group Chat. Now What?
OpenAI Built an AI That Uses Your Computer. Then It Hit ‘Critical’
Your AI Agent Has the Keys. Who Decides What It Can Touch?
Anthropic Reported an $80.7 Billion Loss. The Real Story Is a Single Share
How AI Startups Scale to Millions Without Scaling Their Teams