Cognition AI Raises $2B at $48B: Devin Reaches $900M ARR
September 8, 2026. Cognition AI closed a Series E funding round of more than $2 billion at a $48 billion post-money valuation, the company announced on its blog and confirmed by Reuters and Bloomberg. The round nearly doubles the company’s $26 billion valuation from just four months prior, when it raised $1 billion in a Series D. Andreessen Horowitz and Accel co-led as new investors, joined by existing backers Founders Fund, General Catalyst, and Avenir, plus a syndicate of more than 30 additional firms including Benchmark, Bessemer, Kleiner Perkins, Greylock, Lightspeed, Lux Capital, and Nvidia itself.
The headline is not just the money. It is what the money is anchored to: Cognition’s run-rate revenue grew from $492 million in May to nearly $900 million by the time the Series E closed, according to the company’s own announcement, corroborated by TechCrunch and SiliconANGLE. An ~83% increase in annualized run-rate in roughly four months, tied to named enterprise deployments, moves the agentic coding thesis from category optimism to revenue evidence.
Cognition’s Funding Trajectory at a Glance
| Round | Date | Amount | Valuation | Run-Rate Revenue |
|---|---|---|---|---|
| Series C | Early 2026 | Undisclosed | ~$9B | Undisclosed |
| Series D | May 27, 2026 | $1B+ | $26B | $492M |
| Series E | Sep 8, 2026 | $2B+ | $48B | ~$900M |
Sources: Cognition blog, Reuters, TechCrunch
The company was founded in 2024 by Scott Wu, a math competition champion who built Cognition on a single thesis: software engineers should function as architects, not line-by-line coders, delegating execution to swarms of autonomous agents. Two years after launching Devin as what Cognition called “the first AI software engineer,” the market is handing that thesis a billion-dollar validation.
Where Devin Is Being Deployed
The $900 million run-rate is not coming from one sector. SiliconANGLE and Cognition’s announcement detail Devin operating across six distinct enterprise environments today:
- NVIDIA: chip design workflows
- GE Aerospace: aviation software engineering
- Citigroup: financial services systems
- Goldman Sachs: financial services systems
- Mercedes-Benz: automotive engineering
- Modal Labs: AI infrastructure
The breadth matters as much as the depth. A year ago, the open question was whether agentic coding would convert beyond early-adopter startups. Semiconductor design, aerospace, investment banking, and automotive manufacturing are among the most regulated and risk-averse engineering environments in the world. Deployment in all four, by named Fortune 500 names, answers the “will enterprises actually pay for this?” question with real data.
Nvidia’s role is especially notable: the chipmaker is simultaneously a Series E investor and a named Devin customer, a circular-capital structure that mirrors patterns seen elsewhere in the AI infrastructure stack. Nvidia funds the applications that consume its own GPUs, aligning incentives across the supply chain at the application layer.
What Cognition Launched Alongside the Round
The Series E was not just a financial announcement. Cognition used the moment to ship three new capabilities that extend Devin beyond traditional coding workflows:
Auto-Triage assigns Devin to investigate and resolve software incidents autonomously. When a service degrades or goes down, Devin pulls logs, traces root causes, proposes and tests a fix, and routes to a human engineer only when it cannot resolve the issue within defined parameters. The goal is cutting mean time to resolution for software incidents without additional on-call burden.
Security Swarm coordinates multiple Devin agents to run continuous security validation against a codebase. Rather than scheduled penetration tests or static analysis scans, Security Swarm runs ongoing checks as code is merged, surfacing vulnerabilities before they reach production. Enterprise security teams get persistent coverage without proportionally scaling headcount.
Automations is a no-code workflow builder that lets non-engineers schedule and trigger recurring Devin tasks: running test suites, generating weekly change reports, updating documentation, or auditing dependency versions. The abstraction layer moves Devin from a developer tool to a general-purpose engineering automation platform accessible to ops and product teams.
Together, these capabilities expand Cognition’s addressable market well beyond software-engineering seats toward the broader enterprise operations audience. Any team that owns a codebase or depends on software infrastructure becomes a potential buyer.
The Investor Signal: Run-Rate or Nothing
Two consecutive September mega-rounds, Cognition’s $2B and Mistral’s 3B euro raise announced the same week, share a structural feature: both were priced on explicitly disclosed revenue run-rates, not category narrative. Cognition’s valuation of approximately 53x its company-stated run-rate prices in substantial compounding, which is a bet on the ARR trajectory continuing.
The practical consequence for enterprise buyers is a shift in the negotiating dynamic. When AI vendors can anchor their valuation to real, disclosed, growing revenue, they operate from a position of demonstrated demand rather than theoretical potential. Procurement teams that have been deferring agentic coding evaluations are now negotiating with vendors who have evidence that their competitors have already moved.
For Enera’s enterprise clients thinking about AI-native transformation, the Cognition raise is a useful forcing function: the question is no longer whether autonomous coding agents work at scale, but which deployment approach fits each team’s risk tolerance and workflow design.
What Enera Is Watching
Three dynamics are worth tracking as Cognition scales from $900M toward the $4 to $5 billion ARR the company is projected to reach by end of 2026, per reporting from The Information:
Retention durability. Run-rate is a snapshot. The $48B valuation prices in compounding that has not yet been earned over multi-year contract cycles. The real test is whether Devin’s enterprise contracts renew at expanding scope, not just at original scope. Regulated industries, especially finance and aerospace, tend to have long evaluation cycles but also long retention curves when a tool embeds into mission-critical workflows.
Margin structure at maturity. NVIDIA’s dual investor-customer role raises the long-term question of who captures margin as the agentic coding market consolidates. The GPU supplier funds the application that consumes its compute, aligning incentives now. But as model efficiency improves and inference costs fall, the margin equation for application-layer companies will shift. How Cognition prices across that transition will determine whether its current revenue multiple is defensible.
Agent governance requirements. As Devin moves into semiconductor design and aviation, enterprise security teams are asking who audits what the agent changed and how reversibility is guaranteed. The new Security Swarm capability addresses part of this, but enterprises in regulated sectors will want formal audit trails, role-based agent permissions, and integration with existing change-management processes. The vendors who solve agent governance alongside agent capability will hold the enterprise contracts longer.
For GTM teams evaluating where agentic AI fits into their enterprise stack, the Cognition Series E makes one thing very clear: autonomous software engineering is no longer a forward-looking capability. It is a present-tense competitive variable. The engineering teams deploying Devin today are shipping code at a different velocity than those who are not.
If you are building an agentic enterprise strategy and need help mapping where autonomous agents fit your operations and GTM workflows, talk to Enera.
Sources: Cognition blog | Reuters | TechCrunch | SiliconANGLE | FourWeekMBA analysis | Unite.AI