Nous Research has confirmed a $90 million Series B round that values the open-source AI developer at $1.5 billion, TechCrunch reported Wednesday. The three-year-old startup is using the money to sell a business version of Hermes Agent, its open-source AI agent, to companies that want to automate multi-step work.

What Happened

Robot Ventures led the round. Nvidia, Samsung, Union Square Ventures, Menlo Ventures and 1789 Capital, where Donald Trump Jr. is a partner, also took part, according to TechCrunch. SiliconANGLE listed Y Combinator among the participants as well. The new money brings Nous' total funding to $158 million.

Alongside the round, Nous introduced Hermes for Businesses, which lets companies deploy customized agents for multi-step workflows while keeping their data private. The company now sells three paid editions: Hermes Cloud, aimed at individuals who want a quick hosted setup, and Hermes Business and Hermes Enterprise, which add team token limits and collaboration features for organizations.

RelatedProofpoint to Buy Varonis for $6.3 Billion, Citing Risks of AI Agents' Data Access

Nous gave several figures to show traction. It said Hermes Agent, released in February, has been cloned or downloaded more than 24 million times and has drawn thousands of open-source contributors. The company also claims Hermes accounts for roughly 2.5% of all AI tokens consumed worldwide. Annualized revenue reached about $36 million in mid-September, and Nous projects it will top $100 million by the end of 2026, figures SiliconANGLE attributed to The Wall Street Journal.

How Hermes Works

Hermes runs as a desktop application that can act on a user's files and programs. SiliconANGLE described it organizing documents, using browser functions and running code inside Docker containers. The agent is not tied to one model: it can call third-party large language models, including Anthropic's Claude, as well as Nous' own Hermes 4 models, and a router picks a model for each prompt based on inference cost, speed and output quality.

The Hermes 4 models were trained on a custom dataset of 5 million synthetic records, about two-thirds of them meant to strengthen reasoning, according to SiliconANGLE. Nous generated that data with an in-house tool called DataForge.

Why It Matters

Nous built its name releasing openly available Hermes language models, and Hermes Agent gave developers a self-hosted alternative to closed agent products from the large labs. The Series B is a bet that the same community following can be converted into paying enterprise customers who want support, controls and hosted options on top of free code.

The investor list also ties Nous to two hardware suppliers. Nvidia makes the accelerators and Samsung the memory that run the models behind agent workloads like those Hermes generates.

The revenue projection is aggressive. Moving from a $36 million run rate in mid-September to more than $100 million by Dec. 31 would require the business tiers to win large customers quickly, and the company has not said how many paying organizations it has today.

What to Watch

The first test is whether Hermes Business and Hermes Enterprise land named corporate customers, and on what terms Nous handles data that never leaves a customer's own systems. Another is whether the 2.5% token-share claim, which Nous has not broken down publicly, holds up as larger rivals push their own agent products. Finally, investors will look for confirmation that year-end revenue clears the $100 million mark the company has set for itself.