Robot Ventures leads Nous Research's new funding round

Nous Research is closing in on a new funding round of at least 75 million dollars that would value the company at roughly 1.5 billion dollars, and the striking part is that its flagship product is free. The round is being led by Robot Ventures with significant participation from Union Square Ventures and other prominent names, according to reporting by TechCrunch citing sources with knowledge of the deal, which drew a high level of investor interest. Nous declined to comment, so nothing is signed yet, but the figure would roughly double a valuation the company only reached about fifteen months ago following a 50 million dollar Series A led by the crypto-focused fund Paradigm. The product driving that leap is Hermes, an open-source AI agent that anyone can download and run on their own machine. It arrived shortly after locally run agents caught on, and it differentiates itself with built-in skills spanning web search, coding and image analysis, plus the ability to pick up new capabilities automatically from what a user does. It runs continuously on a computer or a virtual private server, plugs into Telegram and Discord, and has racked up roughly 214,000 stars and nearly 40,000 forks on GitHub. The business model sits alongside rather than on top of the free version, with a cloud-hosted option sold on subscriptions ranging from 20 to 200 dollars a month, and the new capital is meant to expand exactly that. What makes Nous unusual is how comfortably it straddles two worlds that rarely mix. Founded in 2023 by Jeffrey Quesnelle, Karan Malhotra, Ryan Teknium and Shivani Mitra, the lab has published research cited by Meta and DeepSeek while also building Psyche, a decentralized training network that coordinates GPU contributions from around the world through the Solana blockchain. That runs on an in-house optimizer called DisTrO, designed to cut how much data has to move between machines during training, and the team demonstrated it by pretraining a 15 billion parameter model in a distributed setup. Few labs sit this credibly in both open-source AI and crypto infrastructure. The Gulf angle is more relevant than it may look. Sovereignty has become the organizing principle of regional technology policy, and open-weight models that can be downloaded, inspected and run on infrastructure a government controls speak directly to that instinct, which is precisely why Saudi Arabia's HUMAIN and Abu Dhabi's G42 have leaned on open approaches through Allam and Falcon respectively. Decentralized training networks like Psyche offer another hedge for a region navigating export controls on advanced chips. As Gulf institutions weigh how much of their AI stack to own outright, a lab giving away its agent while commanding a unicorn valuation is a useful case study in where the value actually accrues.