Robinhood’s venture arm is playing both sides of the AI coding race. The firm put $10 million into Cognition, the company behind the Devin AI coding agent, according to an exclusive report in Fortune. This follows a $75 million investment in OpenAI made back in April.
The Deal Details
Robinhood Ventures participated in Cognition’s Series E funding round in September, which raised more than $2 billion at a $48 billion valuation. Robinhood is disclosing its participation for the first time. Other backers in that round include Andreessen Horowitz, Accel, and Founders Fund.
Cognition reported an annualized revenue run rate of nearly $900 million at the time of the September round, up from $492 million in May.
Why Robinhood Is Backing Rivals
Sarah Pinto, head of Robinhood Ventures, told Fortune directly: “It’s not a winner-take-all market.” Robinhood’s own engineers already use multiple AI coding tools. On the consumer side, Robinhood lets customers choose between different AI models to run its trading agents, which informs this multi-horse strategy.
Pinto first met Cognition cofounder and CEO Scott Wu in 2024. She said his vision for where AI was heading “really never changed” as the technology and market evolved, which she described as “pretty extraordinary.” Cognition’s other cofounders are Steven Hao, Walden Yan, and Russell Kaplan.
The Fund Structure
The Cognition investment is held through Robinhood Ventures Fund I, a publicly traded closed-end fund that began trading in March. It gives retail investors exposure to private companies by buying and selling shares on a stock exchange, rather than needing a direct stake. Current holdings include Databricks, Revolut, and ElevenLabs, among others.
The fund’s net asset value jumped roughly 17% for the quarter ended September 30, its largest quarterly increase to date.
The Operator Angle
The competitive landscape for AI coding tools now includes OpenAI’s Codex, Anthropic’s Claude Code, GitHub Copilot, and Cognition’s Devin. Robinhood’s bet is that developers will use more than one. If that’s true, the companies that build the most reliable specialized tools win alongside the platforms, not instead of them.
