Cognition builds Devin, an autonomous AI software engineer for enterprise development teams. Its new Devin AI funding round — more than $1 billion announced on Wednesday, May 27, 2026 — comes as companies try to cut engineering backlogs without hiring endlessly. Founded in 2023 by Scott Wu, Steven Hao, and Walden Yan, the San Francisco startup is now being valued like a company that could define a product category, not just ride a hype cycle. Investors are betting there’s still room for an independent AI coding company even while model giants crowd the market.
What does Devin AI actually do?
Devin takes a software task in plain English, inspects a codebase, sketches a plan, writes code, runs commands, tests its own work, and hands the result back inside a developer workflow. In practice, it starts by searching the repo for relevant files and snippets. Then it produces an initial assessment with findings and implementation questions before moving into a more detailed plan a team can approve or edit.
The product isn’t just a chat box. Devin gives users an embedded IDE, a terminal, and a browser view. Teams can watch what it’s doing in real time, jump in when needed, or take over directly. That matters. The difference between a flashy demo and something engineers will trust usually comes down to visibility and control.
It also goes past single-ticket work. Devin can spin up managed child sessions in parallel, each in its own isolated VM, to handle chunks of a larger migration or code change. It can analyze earlier sessions and turn successful work into reusable playbooks. It also maintains an internal knowledge base and schedules recurring jobs like nightly checks or routine maintenance.
It plugs into the systems enterprise teams already use, including GitHub, GitLab, Bitbucket, Jira, Slack, and Microsoft Teams. So the pitch isn’t “replace your engineers.” It’s closer to “give your engineers another worker that can actually execute.”
Who built Cognition and Devin AI?
The founding story
Cognition started in 2023 with Scott Wu, Steven Hao, and Walden Yan. All 3 founders came out of elite competitive programming circles, which helps explain why the company went straight after autonomous software engineering instead of building yet another thin wrapper around a frontier model. They weren’t chasing a generic AI app. They were chasing one of the hardest agent problems they could find.
Why these founders had real market fit
Wu, Cognition’s CEO, had already built at startup scale before Devin. He previously co-founded Lunchclub and made Forbes’ 2020 30 Under 30 list. He’s also known in programming circles for a string of top-tier competitive results, including 3 International Olympiad in Informatics gold medals and a third-place finish in Google Code Jam. That mix — consumer startup execution plus deep technical credibility — is rare, and VCs tend to pay up for it.
Hao, the CTO, brought applied AI infrastructure experience from Scale AI, where he worked as a senior engineer before starting Cognition. Yan, Cognition’s CPO, previously co-founded DeepReason. Together, the resume starts to make sense: one founder with startup-building reps and one with hard production AI experience. The third had prior founder scar tissue.
Traction and the new round
The company’s customer list now includes Mercedes-Benz, NASA, Goldman Sachs, and Santander. Cognition also disclosed a $492 million annualized revenue run-rate, with enterprise usage of Devin growing 50% month over month for the last 6 months. For a company this young, that’s not normal. It changes a funding round from belief to aggressive extrapolation.
That helps explain the price. Cognition raised more than $1 billion at a $25 billion pre-money valuation and a $26 billion post-money valuation. The round follows a $400 million raise in September 2025 that valued the company at $10.2 billion post-money, so the markup in just 8 months is huge.
Lux Capital, General Catalyst, and 8VC led the new financing. Existing backers including Elad Gil, Soma Capital, Omri Casspi, and Founders Fund also participated, alongside new investors such as Ribbit Capital, Atreides, and Layer Global.
How Cognition stacks up against Codex, Claude Code, Cursor, and Jules
This part matters. A year ago, it looked like model companies would own AI coding end to end. OpenAI’s Codex is a cloud-based software engineering agent that can run many tasks in parallel. Anthropic’s Claude Code lives closer to the terminal and developer tooling stack, with strong workflow automation and MCP-based connectivity. Cursor has become the most visible AI-native editor. It pushes agents across desktop, web, mobile, Slack, and GitHub. Google’s Jules is another asynchronous coding agent built to read code, fix bugs, and work in a cloud environment.
Cognition’s bet is narrower and bolder. Rather than being the general model maker or the everyday editor first, it’s trying to own the “AI coworker” layer for software teams — the part where work gets scoped, delegated, executed, checked, and repeated. It also picked up the remaining pieces of Windsurf in 2025 after Google’s acqui-hire move, which gave it more product and talent leverage right as this market got crowded.
Why does this Devin AI funding round matter?
This isn’t just another giant AI valuation stapled onto a nice demo.
Cognition is being funded like a company that could become a core enterprise software vendor. That changes the questions customers, rivals, and investors will ask next. The issue is no longer whether teams will try autonomous coding agents. They already are. The issue is which products become durable enough for regulated, high-stakes organizations to trust with bigger chunks of production work.
The fresh capital should give Cognition room to harden Devin where enterprise buyers care most — reliability and permissions. Observability, workflow depth, security, and support matter too. That’s also where this round feels less like consumer AI theater and more like infrastructure building. Big banks and industrial companies don’t buy tools because they’re fun. They buy them because the tooling fits how software gets shipped.
Investors are clearly backing that thesis. They’re not just financing model access. They’re financing the application layer that sits on top of those models and turns raw capability into repeatable engineering output.
How big is the AI coding agents market?
Pretty big already. The global AI code assistants market was estimated at $8.5 billion in 2025 and is projected to reach about $42.9 billion by 2033, with a 22.5% compound annual growth rate from 2026 through 2033. That kind of curve is why capital keeps flooding into code-generation and agent startups.
But adoption isn’t the same as trust. Stack Overflow’s 2025 developer survey found a widening trust gap around AI tools, with 46% of developers saying they don’t trust the accuracy of AI output. That’s good context for Cognition. It suggests the winners won’t be the loudest products. They’ll be the ones that can show their work and fit into existing engineering systems. They also need to give teams a sane way to supervise the agent.
That’s why the timing works. The market is large enough to matter, crowded enough to be brutal, and skeptical enough that enterprise execution still counts for a lot.
What should investors and customers watch next?
Cognition has already won the hardest thing to fake — attention from serious enterprise buyers and serious capital at the same time.
Now it has to prove that autonomous coding agents can hold up after the demo, after procurement, and after security review. That’s the next test for Devin AI funding as a story. If Cognition keeps converting experimental use into standard enterprise workflow, this round will look expensive only in hindsight.
Read how Stord raised $250M at a $3B valuation to build independent commerce infrastructure that helps e-commerce brands manage warehouses, fulfillment, and shipping without relying on Amazon.
FAQ
– What is the latest Cognition funding round for Devin AI?
Cognition raised more than $1 billion on May 27, 2026. The financing valued the company at $25 billion pre-money and $26 billion post-money, a massive jump from the $10.2 billion post-money valuation attached to its September 2025 round.
– How does Devin AI work for software teams?
Devin takes a task in natural language, searches the relevant codebase, proposes a plan, and then writes, runs, and tests code inside its own working environment. Teams can watch the process through an IDE, terminal, and browser view. They can approve plans before execution and let Devin split large jobs into parallel managed sessions.
– Who founded Cognition?
Cognition was founded in 2023 by Scott Wu, Steven Hao, and Walden Yan. Wu previously co-founded Lunchclub, Hao worked at Scale AI, and Yan earlier built DeepReason, which is a big reason the company has looked unusually credible in AI software engineering from day 1.
– Is Cognition in the AI coding assistant market or something broader?
It sits in the AI coding assistant market, but its product ambition is broader than autocomplete or code suggestions. Cognition is aiming at the emerging AI coding agent category, where software tasks are not just assisted but planned, executed, and managed asynchronously across enterprise workflows.
