Nous Research Hits $1.5B on Promises and Hermes Vibes

Nous Research, the Hermes agent maker nobody asked for, has apparently convinced Robot Ventures, Union Square Ventures, and a supporting cast of "prominent investors" to participate in at least $75 million in new funding at a $1.5 billion valuation. That's right: $1.5 billion for a company whose primary distribution mechanism is architectural diagrams shared on Twitter and a GitHub repository with enthusiastic stars but zero paying customers. The round is being led by Robot Ventures—a name so on-brand for 2024 VC that it reads like satire we couldn't improve.

For those unfamiliar with Nous Research's commercial innovation, the company makes Hermes agents: AI systems that integrate with Claude and GPT to perform tasks. The emphasis here is on "integrates with"—meaning Nous builds on top of Anthropic and OpenAI's models without formal partnerships or revenue-sharing arrangements that anyone has publicly disclosed. The company's actual revenue, customer list, and paying user base remain classified information, which in VC speak means "we prefer to discuss TAM instead." Hermes itself appears to exist primarily as a reference architecture that developers can implement, fork, or ignore—a business model more accurately described as "open-source marketing" than "software company."

Robot Ventures, the lead investor, specializes in AI infrastructure and robotics, a sector currently experiencing the kind of irrational exuberance last seen in cryptocurrency circa 2021. USV, a firm with a reputation for thoughtful long-term investing, has apparently decided that $1.5 billion for an agent-building layer is the right entry point in a market where every AI company claims to be building agents, and where actual enterprise adoption remains glacial. One wonders what internal model justified this valuation, or whether the spreadsheet simply read: "Agents are hot. Hermes is an agent thing. Therefore: $1.5B."

The pitch, one imagines, goes something like this: "Nous Research enables enterprises to deploy autonomous agent systems at scale, leveraging best-in-class LLMs through an extensible architecture." Translation: "We glued some Python to Claude and it worked on our demo." The "significant participation from other prominent investors" is classic VC hedge language—it means secondary players saw Robot Ventures and USV on the cap table and decided not to be left out of the next unicorn meme. FOMO remains the only asset class that consistently appreciates.

The risks are architectural and commercial: first, any moat Nous imagines it has evaporates the moment Anthropic or OpenAI adds agent capabilities directly to their APIs, which they will, because it's trivial and owned by companies with actual revenue. Second, enterprises don't buy agent frameworks—they license agent services from vendors they already know. Third, the open-source nature of Hermes means that value accrual flows upstream to Claude and GPT, not horizontally to Nous. The company has built a beautiful platform that makes other people's models more useful, which is approximately the inverse of a defensible business model.

What this deal reveals is the persistent belief among top-tier VCs that raw momentum and impressive positioning can substitute for unit economics and customer demand. Nous Research may genuinely be building something interesting, but $1.5 billion at this stage—before meaningful revenue, before proof of enterprise adoption, before API-level integration from the model companies—is a valuation that flatters the investor more than the company. In five years, either Nous will be a forgotten acqui-hire by Anthropic, or we'll be reading about this round in "Top VC Disasters of the 2020s." Both outcomes are already baked in.

The AI bubble doesn't need unicorns anymore—it just needs a good story about what happens when you let the agents run the funding meetings.