Nous Research Raises $75 Million as Agent Market Heats Up
Nous Research Raises $75 Million as Agent Market Heats Up
Nous Research, a startup building AI agent tools, is in talks to raise at least $75 million at a $1.5 billion valuation, according to three sources cited by TechCrunch. Robot Ventures is leading the round, with support from USV and other investors. The company and its backers have not publicly confirmed the details.
This funding round comes just under three months after Nous closed a $50 million Series A at a $1 billion valuation, led by the venture firm Paradigm. That earlier round was structured around the company's cryptocurrency-based AI platform on Solana. Two major funding events within a single quarter, with the valuation jumping 50 percent, suggests how quickly investors are moving on AI agent infrastructure right now.
Nous Research was founded in 2023 by Jeffrey Quesnelle, Karan Malhotra, Ryan Teknium, and Shivani Mitra. Before this new round, the company had raised $70 million from Paradigm, Robot Ventures, North Island Ventures, OSS Capital, and Balaji Srinivasan, according to Crunchbase data cited by TechCrunch. The new money is meant to expand Hermes products and the company's business model, though specifics remain undisclosed.
What Nous Research Actually Does
Hermes is an open-source AI agent — a piece of software that can take actions on its own, like searching the web or writing code — at the heart of Nous's pitch. Think of an agent as a more autonomous version of the chatbots you may have used; it can perform tasks without you having to click through each step. Hermes has gathered roughly 214,000 stars and nearly 40,000 forks on GitHub, a measure of how many developers are using and contributing to the project. Few open-source agent frameworks have reached that scale.
The software comes with built-in capabilities: web search, coding, and image understanding. The original design goal was for Hermes to learn from real usage and develop new skills over time, rather than shipping with a fixed set of abilities.
Nous makes money in two ways. First, it offers a cloud-hosted version of Hermes through a subscription service, ranging from $20 to $200 a month. This is a standard software-as-a-service model. Second, the company operates a decentralized compute network — a system where users contribute their own computers to help with AI training and inference work (running predictions on the model). This network was the foundation of the earlier funding round structured around cryptocurrency tokens.
That combination is unusual. Nous is simultaneously an open-source project with genuine grassroots adoption, a cryptocurrency-native infrastructure play, and a venture-backed software company pursuing traditional venture investors. Each identity pulls in a different crowd of backers, each with different expectations for how the company should make money and be governed.
The Pace of Capital
The speed at which this funding is happening is the more telling detail. In April, Nous closed at a $1 billion valuation. Three months later, it is reported to be raising at $1.5 billion, with no major product milestone or revenue announcement disclosed in between. This pace reflects how intensely venture capital is chasing agent infrastructure companies right now, especially those with visible developer adoption.
Whether strong GitHub numbers and a small hosted customer base actually justify a 50 percent valuation jump in twelve weeks is a fair question. Any investor betting on this round is essentially wagering that the answer is yes.
One important note: the $1.5 billion figure comes from sources with knowledge of the deal, not from Nous Research or its investors, all of whom have declined to comment or not responded to press inquiries. Deal terms in early-stage venture rounds can shift during negotiations, so this figure reflects the state of talks as reported, not a confirmed closing. Readers should treat it as provisional.
What Comes Next
Nous has published technical work that hints at what "expanding Hermes products" might involve. In previous blog posts, the company detailed integrations between its Tinker API and an Atropos reinforcement learning framework — components that would allow Hermes agents to improve by learning from feedback gathered during real use. This aligns with the stated goal of agents that can build new skills on their own.
For an industry that has seen agent startups emerge largely on the strength of demo videos and synthetic benchmarks, Nous Research's GitHub numbers offer harder evidence of developer adoption. The real question now is whether that adoption translates into the kind of steady revenue needed to support a $1.5 billion valuation, and whether the decentralized compute network can scale alongside the commercial product.
The agent tooling market has grown crowded with startups, all chasing the same opportunity. In a crowded field, what increasingly matters is not just how good the underlying model is, but whether developers actually choose to build with your tools and whether they trust your company long-term. That is the bet this round places.
