A new funding push only months after a mega-round

Cognition, the company behind the AI coding agent Devin, is reportedly already speaking with investors about another large funding round that could value the startup at at least $40 billion. The talks come only a few months after Cognition raised $1 billion in May at a $26 billion valuation, according to the report.

Bloomberg’s sources said the new valuation would be tied in part to Cognition reaching a $1 billion annualized revenue run rate. A revenue run rate is a way to estimate the current pace of revenue on a yearly basis; it is useful for fast-growing companies, though it is not the same as booked full-year revenue.

The numbers behind the valuation jump

When Cognition announced its previous financing, Scott Wu confirmed to TechCrunch that the company had reached a $492 million annualized revenue run rate. He also said enterprise usage of Devin had been growing 50% month over month for the prior six months.

The reported fundraising story therefore rests on several concrete data points:

• $1 billion raised in the May round;
• $26 billion valuation in that round;
• At least $40 billion potential valuation in the new talks;
• $492 million annualized revenue run rate previously confirmed;
• 50% month-over-month growth in enterprise usage over six months.

Together, those figures show why investors may be willing to revisit the company so quickly: the market is not only rewarding AI coding demos, but also looking for signs that enterprises are paying for and expanding usage of these tools.

What Devin is being sold to do

Devin is described as an AI coding agent. In simple terms, an agent is software that can work toward a goal through multiple steps, rather than merely offering one-off code suggestions. In a software engineering context, that can include reading code, making changes, and helping push technical tasks forward.

Wu has said Devin is not being sold as a replacement for human programmers. Instead, the product is often used for long-tail engineering work that many developers find tedious, such as updating older software or moving applications from one platform to another.

That positioning matters. Large organizations often have extensive legacy systems and backlogs of maintenance work. These tasks are necessary, but they may not be the most attractive use of scarce engineering time. If an AI coding agent can reliably handle part of that workload, it becomes easier to justify enterprise spending even before more ambitious claims about fully autonomous software development are proven.

Enterprise adoption is the key signal

Cognition says its customers include Mercedes-Benz, NASA, and Goldman Sachs. Those names span manufacturing, aerospace, and finance, suggesting that the company is targeting complex enterprise software environments rather than only technology-native customers.

For such buyers, adoption depends on more than code generation. A tool has to fit into existing workflows, support maintenance and migration tasks, and be useful enough for engineering teams to keep using it. The report does not provide deployment details or customer-by-customer usage figures, so the clearest available signal remains the combination of revenue run rate and usage growth previously disclosed by the company.

What it means for the AI coding market

The reported talks show that AI coding is moving into a phase where valuation and revenue execution are advancing together. Investors appear to be betting that AI agents can reshape parts of the software engineering budget, especially repetitive maintenance and migration work that consumes time across large codebases.

The same valuation surge also raises the bar. The market will increasingly look beyond impressive demonstrations and ask whether these products can retain enterprise customers, deepen usage, and perform consistently in real-world codebases. If Cognition can turn its reported momentum into durable revenue, AI coding agents may become a standard layer of the enterprise development stack. If growth slows, investors will likely reassess how much of the current enthusiasm is already priced in.