From 24 Million Clones to the Boardroom: Nous Research Confirms $90M Series B, Launches Hermes for Businesses

Open-source AI has a new unicorn — and this one comes with an enterprise sales motion. Nous Research, the startup behind the widely cloned Hermes Agent, has confirmed a $90 million Series B at a $1.5 billion valuation, while simultaneously launching “Hermes for Businesses,” a product aimed squarely at companies that want agentic AI without sending their data to someone else’s cloud.

The confirmation, reported by TechCrunch’s Marina Temkin on October 7, closes out a summer of speculation. Back in July, the round was rumored at $75 million; the final number landed higher.

The round, in numbers

• $90 million Series B — up from the ~$75 million figure reported during negotiations in July 2026
• $1.5 billion valuation, making Nous Research one of the few open-source-native AI companies to clear the unicorn threshold
• $158 million total funding raised since the company’s founding in 2023
• Robot Ventures led the round, with participation from Nvidia, Union Square Ventures, Menlo Ventures, Samsung, and 1789 Capital — the latter notable because Donald Trump Jr. is a partner there

The investor list reads like a deliberate bet on multiple theses at once: Nvidia gets a consumer of GPUs and a distribution channel for its AI stack; USV and Menlo are betting on open-source distribution economics; Samsung brings a device and enterprise ecosystem; and 1789 Capital adds a politically angled flourish to what is otherwise a classic infrastructure story.

Why Hermes matters: 24 million clones and 2.5% of global token usage

The heart of the story is adoption. The open-source Hermes Agent has been cloned more than 24 million times, and Nous estimates it now drives roughly 2.5% of global AI token usage.

Those are extraordinary numbers for a three-year-old company competing against OpenAI, Anthropic, and Google. They reflect a structural shift in how AI software gets adopted: rather than winning a procurement battle, Hermes won the developers. The agent harness — the scaffolding that connects a language model to tools, memory, and scheduling — became the de facto standard for people building their own AI assistants, and the clones followed.

That installed base is exactly what makes the enterprise pitch credible. When your product is already running in millions of developer setups, hobbyist rigs, and shadow-IT deployments, the step from “loved by tinkerers” to “approved by procurement” is shorter than it looks.

Hermes for Businesses: private agents for multi-step work

The new product, “Hermes for Businesses,” lets companies deploy customized AI agents that can handle multi-step workflows while keeping their data private and secure.

The emphasis on privacy and self-hosting is not incidental. Enterprise buyers in 2026 have two overlapping anxieties: giving a frontier lab access to sensitive internal data, and building critical workflows on APIs whose pricing and behavior can change overnight. An open-source agent harness that companies can run on their own infrastructure addresses both — the model can be swapped, the data never leaves, and the workflow logic is inspectable.

This positions Nous against a crowded field. OpenAI and Anthropic are pushing their own enterprise agent platforms, Meta’s Muse is chasing consumer and business assistant use cases, and every cloud vendor has an agents story. Nous’s differentiation is the same one that carried Linux, Kubernetes, and PostgreSQL into the enterprise: you don’t have to trust the vendor, you only have to trust the code.

The revenue trajectory: $36M now, $100M targeted for year-end

According to The Wall Street Journal, Nous was at roughly $36 million in annualized revenue by mid-September 2026 and expects to pass $100 million before the end of 2026.

Nearly tripling annualized revenue in a single quarter is aggressive, but it’s the kind of aggression that a 24-million-clone installed base makes plausible. The open-source funnel is already full; the question is conversion. If even a small fraction of the developers and teams already running Hermes convert to paid business plans with support, hosting, and enterprise controls, the math works.

It also helps that the company is not burning capital on training frontier models from scratch. Nous’s history is in fine-tuning, open training techniques, and tooling — a capital-efficient posture compared to labs spending billions on pretraining. The $90 million goes a long way when your core product is software distribution rather than a compute arms race.

What it means

Three things make this round more than another AI funding headline.

First, it validates open-source agents as a business, not just a movement. For two years the consensus was that open models would commoditize and value would accrue to closed labs. Nous’s unicorn status — built on an agent harness rather than a foundation model — suggests the durable layer may be the orchestration and deployment tooling around models, whoever trains them.

Second, the Nvidia participation is telling. The GPU maker has backed most of the major AI labs, but investing in an open-source agent company signals that Nvidia sees the agent layer — not just the model layer — as a driver of future inference demand. Every Hermes deployment is ultimately a token consumer.

Third, the enterprise turn is now unavoidable for every consumer-AI company. Nous follows OpenAI, Anthropic, and Meta in concluding that individual subscriptions, however popular, don’t produce the contract sizes that justify AI-era valuations. The open question is whether open-source credibility translates into enterprise sales execution — a very different muscle from community building.

For now, the numbers argue for optimism: 24 million clones, 2.5% of global token usage, and a revenue line that’s supposed to triple in three months. The open-source agent era has its unicorn. The next test is whether the boardroom buys what the developers already did.