Cognition Hits a $48B Valuation as Revenue Nears $900M

In Cognition’s September 8 Series E announcement, the company said it raised more than $2 billion at a $48 billion valuation, named Andreessen Horowitz and Accel as the round’s new lead investors, and put run-rate revenue at almost $900 million, up from $492 million in May.

A September 8 Reuters report independently corroborates the $2 billion round, the $48 billion valuation and the comparison with Cognition’s $26 billion valuation in May. The rapid repricing matters because it places one of the largest private-enterprise software bets on the company behind Devin, an agent designed to carry software-engineering tasks through multiple stages rather than merely suggest code.

Valuation and reported revenue rose at nearly the same rate

The May financing supplies the clearest baseline: TechCrunch’s May financing account put the round above $1 billion, the pre-money valuation at $25 billion, the post-money valuation at $26 billion and Cognition’s annualized revenue run rate at $492 million.

From $26 billion to $48 billion, the valuation increased by $22 billion, or about 84.6%, in roughly three and a half months. Because the September revenue figure is “almost” $900 million rather than an exact amount, run-rate revenue grew by slightly less than $408 million and slightly less than 82.9%. On the available figures, the two measures advanced at strikingly similar rates.

The implied valuation-to-run-rate-revenue multiple consequently changed very little. The May snapshot produces a multiple of about 52.8 times, while $48 billion divided by $900 million equals 53.3 times; the actual September multiple would be somewhat higher because revenue remained below that threshold. The repricing therefore appears to reflect a much larger revenue baseline more than a dramatic expansion in the multiple investors assigned to it.

That comparison is necessarily limited. Run-rate revenue annualizes a recent period and is not revenue recognized over a completed financial year. Publicly available financing information does not provide audited accounts, margins, contract duration, retention, customer concentration or the allocation of revenue between Devin and Windsurf, leaving the quality and durability of the sales base untested.

A crowded syndicate is concentrating capital around Cognition

Andreessen Horowitz and Accel joined as lead investors, while Founders Fund, General Catalyst and Avenir returned. The syndicate also includes prominent venture and institutional names such as Benchmark, Bessemer Venture Partners, Kleiner Perkins, Greylock, Lightspeed, T. Rowe Price, Lux Capital, 8VC, Bain Capital Ventures and Nvidia.

The list is broad, but its significance is concentration of conviction: a large group of major investors has accepted the same sharply higher price for one autonomous-coding company. The short interval since May allowed little time for conventional annual financial evidence to accumulate, making the reported commercial acceleration and expectations for continued enterprise adoption central to the investment case.

The financing also indicates that investors see potential value in an agent layer independent of the companies supplying foundation models. That thesis depends on Cognition controlling the workflow in which agents receive assignments, navigate repositories, use suitable models and return work for review. If that layer remains differentiated as underlying models improve, it could retain customer relationships and pricing power; the financing itself does not prove that outcome.

Devin is expanding into event-driven engineering work

SiliconANGLE’s account of the product expansion details Auto-Triage for initial production-incident investigations, Security Swarm for finding and triaging vulnerabilities, Automations triggered by events in Slack, GitHub and Linear, and Cognition’s position that Devin can combine outside and internally developed models rather than depend on one provider.

These functions increase the number of places where Devin can enter an engineering process. A chat-based assistant waits for someone to initiate each exchange; an event-driven agent can begin work when an incident, repository change or project-management update occurs. Successful deployment would let Cognition sell usage across maintenance, security and operations instead of relying only on prompted code generation.

The publicly identified deployments include engineering work at Nvidia, GE Aerospace, Citi, Mercedes-Benz and Modal. These are company-reported adoption claims, not independent measurements of cost savings, code quality or autonomous completion. Deployment size, customer spending, renewal rates and the amount of human review required for these workflows remain undisclosed.

Independent performance evidence is more qualified

A 2026 conference analysis of coding-agent pull requests examined 7,156 submissions from five agents and measured Devin’s acceptance rate improving by 0.77 percentage points per week across a 32-week observation window, while documentation submissions across the dataset had an 82.1% acceptance rate compared with 66.1% for new-feature work.

In statistically significant task-specific comparisons, Devin underperformed Codex, Copilot and Cursor on fixes, and no agent led every category. That result does not directly test Devin’s newly introduced incident, security or automation functions, but it shows why adoption and financing figures cannot stand in for a universal product-performance claim.

The evidence is observational rather than causal. Pull-request acceptance is not equivalent to code quality, and differences in repositories, users, model versions, task distribution and observation periods can affect the outcome. It nevertheless provides a useful independent counterweight to commercial claims by showing improvement in one measure alongside substantial variation by task.

As of September 9, Cognition has verified financing on unusually large terms and has presented a revenue trajectory that largely matches the pace of its valuation increase. What remains unknown is whether that run rate converts into durable recognized revenue and whether Devin’s broader workflows deliver consistent results across customers. Audited financial data, retention measures and independent evaluations of the newer functions will determine whether the $48 billion price reflects lasting software economics or an exceptionally aggressive bet on the autonomous-coding market.

Also read:

• Part 3. The Illusion of Value: Why 90% of Crypto Projects Have Billion-Dollar Market Caps But Almost Zero Traffic and Revenue
• Agentic Marketing Revolution: Okara’s AI CMO Agent Hits 10 Million Views and Takes Down Its Own Infrastructure
• Uzum’s $2.3 Billion Valuation Holds—Its Unicorn Milestone Came Earlier