Nous Raises $90 Million to Turn Its Widely Used AI Agent Into a Business Product
Nous Research has closed a $90 million Series B at a $1.5 billion valuation. The company confirmed the financing on Oct. 7, 2026, alongside the launch of an enterprise agent product TechCrunch.
Robot Ventures led the round. Nvidia, Union Square Ventures, Menlo Ventures, Samsung and 1789 Capital also took part. Total funding for the company now stands at $158 million.
The close follows earlier reporting. On July 13, TechCrunch reported Nous was in talks for at least $75 million at a $1.5 billion valuation TechCrunch. The final $90 million came in above that floor, with the valuation unchanged.
Distribution is central to the pitch. Its open-source Hermes agent, software that can carry out multi-step tasks with tools, has been cloned, or copied, more than 24 million times. Nous estimates Hermes accounts for about 2.5% of global AI token usage, where tokens are the small chunks of text AI systems read and produce TechCrunch.
That reach now feeds a paid tier. Hermes for Businesses lets companies run customized agents for multi-step workflows while keeping data private and secure. The emphasis is on deployment the customer controls, rather than a shared hosted assistant.
Underneath is Hermes Agent, which the company describes as a self-improving agent with a built-in learning loop Nous Research. It creates skills from experience and refines them during use. It also saves knowledge across sessions and builds a more detailed model of the user over time.
Revenue is early but growing. Nous was at roughly $36 million in annualized revenue, meaning revenue measured as a yearly rate, by mid-September 2026. It expects to pass $100 million on that basis before the end of 2026 TechCrunch.
Looking at what this means for business buyers, the pattern is familiar from earlier open-source infrastructure. Wide, open experimentation builds practical knowledge about failures and integrations. Then a smaller group pays for privacy, management and customization. The 24 million clones matter less as a headline number than as a source of workflow lessons.
In my view, the technical claim to watch is persistent, self-improving behavior inside company boundaries. Building skills from experience and remembering context can cut repeated instructions and setup work. It also raises the bar for testing, version control and access limits. An agent that keeps learning is useful only if administrators can audit what it learned, roll it back and restrict it by role.
Worth flagging is the token share estimate. If 2.5% of global use already runs through Hermes-based setups, Nous has usage data that most model makers lack. That can guide work on response speed, handling long inputs and reliable tool use. It also means business customers will examine data separation closely.
The broader context here is a shift from models to longer-lasting agents. Copying a model is cheap. Running an agent that keeps memory, chains tools in sequence and improves without leaking data takes steady operational work. If Nous converts even a small share of its open user base to Hermes for Businesses and holds its revenue path, it will have turned wide distribution into a lasting business.
