Cognition targets $1B round at $47B valuation, three months after last raise

Why it matters

A $47B valuation would reprice Cognition by about 81% in just over three months, showing how aggressively investors are backing AI agents with fast, company-reported revenue growth.

Scott Wu (@ScottWu46)'s Cognition is set to raise around $1 billion at a valuation of about $47 billion, Bloomberg reported, extending the company's rapid valuation climb in AI coding. The financing remains under negotiation, and the final amount could exceed $1 billion after Cognition received nearly $10 billion of investor interest, according to people Bloomberg described as familiar with the talks. The terms could still change.

Wu co-founded Cognition with Steven Hao and Walden Yan (@walden_yan), fellow competitive programmers whose records help explain why investors have treated the founding team as part of the product thesis. Wu won three International Olympiad in Informatics gold medals and the 2011 MATHCOUNTS national championship. He later worked on performance engineering at Addepar, attended Harvard for about two years and co-founded the professional-networking startup Lunchclub before starting Cognition.

The proposed financing would arrive just over three months after Cognition raised more than $1 billion at a $26 billion post-money valuation on May 27. Lux Capital, General Catalyst and 8VC led that round, joined by Founders Fund, Ribbit Capital, Atreides, Layer Global and other investors. The new round's lead and participants have not been identified.

If the latest financing closes on the reported terms, Cognition's valuation will have risen about 81% since May.

Wu built for delegated work

Cognition introduced Devin on March 12, 2024, as an autonomous software engineering agent that operates inside its own computing environment. Devin can inspect repositories, plan tasks, edit and test code, debug failures and prepare pull requests for a human to review.

Wu's founding bet centered on agents that could keep working after an engineer handed off a task. In a May interview with Colossus, Wu described a market moving toward background agents rather than software confined to answering prompts or suggesting the next line of code. That approach placed Cognition in a harder technical category: Devin had to sustain a plan, use development tools and recover from errors across longer jobs.

Cognition's early demonstrations drew criticism from independent reviewers, who found Devin slower and less reliable than the launch videos suggested, according to Forbes. Cognition subsequently narrowed its recommended use cases around bounded engineering work, including migrations, tests, security fixes and backlog tasks. That progression matters to the financing case. Cognition has moved from selling the idea of an artificial software engineer to selling measurable engineering capacity inside large organizations.

Wu has also assembled an unusually competition-heavy organization around that goal. Cognition says its founding group collectively holds 10 IOI gold medals and includes former builders from Scale AI, Cursor, Google DeepMind, Waymo and Nuro. The company also says 21 of its first 35 employees had previously founded a business.

The revenue case carries an acquisition

Cognition said in May that enterprise usage had grown more than tenfold since the start of 2026 and that annualized revenue run rate had reached $492 million. By July, Wu and Cognition executive Jeff Wang said the figure had passed $500 million and the team had grown from 44 to 350 people over the preceding year.

Those are Cognition's figures, and run-rate revenue annualizes recent activity rather than reporting revenue already recognized over a full year. Cognition has not published a breakdown of subscription revenue, usage charges or services, leaving investors to judge the durability and margins behind the headline number.

The growth figures also include Cognition's acquisition of Windsurf, the AI coding editor business it bought in July 2025. At the time, Cognition said Windsurf brought $82 million in annual recurring revenue, more than 350 enterprise customers and hundreds of thousands of daily users. Cognition later folded the editor into Devin Desktop, giving Wu a product that reaches developers while they write code as well as an agent designed to complete jobs asynchronously.

That deal widened Cognition's distribution and added an established revenue stream. It also complicates comparisons between Devin's organic growth and Cognition's combined run rate. Cognition's July account said revenue run rate rose from $73 million to more than $500 million after the businesses came together, without separating how much of the increase came from Devin, the former Windsurf product line or cross-selling between them.

Cognition described its May financing as more than $1 billion at a $26 billion post-money valuation. TechCrunch reported the round as more than $1 billion at a $25 billion pre-money valuation, which produces a roughly $26 billion post-money figure. The distinction matters because the proposed $47 billion figure has not yet been described as pre-money or post-money. Until the round closes, the clean comparison is between reported headline valuations rather than final ownership terms.

Investors are pricing the next phase early

The proposed round gives Wu capital to keep competing in a category crowded by well-funded model providers and coding-tool specialists. Cognition must pay for cloud infrastructure, model training, enterprise deployment and the technical staff needed to support agents running inside sensitive corporate codebases. Its purchase of Windsurf also showed that Wu is willing to use Cognition's balance sheet to acquire distribution rather than build every interface internally.

The valuation assumes Cognition can preserve an independent position between foundation-model providers and software-development platforms. Cognition describes itself as model-independent, selecting different models for different engineering jobs while training systems of its own. That strategy can help Cognition shop for price and performance, though it also leaves Cognition reliant on suppliers that sell competing coding products.

For Wu, the financing would turn an early product lead into a heavily capitalized attempt to own the layer where AI models perform software work. Investors appear ready to fund that attempt before Cognition has reported a full year of results at its current scale. The $47 billion price will become real only if the round closes on those terms; the nearly $10 billion of interest shows that plenty of investors already want a place in the queue.