Nous Research Hits $1.5B Valuation With $90M Round, Debuting Enterprise AI Agents

Open-source artificial intelligence lab Nous Research has secured a $90 million Series B funding round that catapults the startup to a $1.5 billion valuation, according to reports confirmed by the company on Wednesday. The fresh capital will fund the rollout of "Hermes for Businesses," an enterprise-grade expansion of its autonomous, open-source Hermes Agent that allows organizations to deploy and orchestrate self-learning AI workers without locking their proprietary data inside closed commercial ecosystems.

Inside the $90 Million Series B and Unicorn Valuation

The funding round was led by Robot Ventures, with high-profile strategic backing from Nvidia, Microsoft's M12 venture fund, Samsung, Union Square Ventures (USV), Y Combinator, Menlo Ventures, and 1789 Capital. The financing raises Nous Research’s cumulative funding to roughly $160 million, following a $20 million seed round in 2024 and a $50 million Series A round in early 2025 led by Paradigm.

The investment represents a dramatic milestone for the decentralized and open-weight AI movement. While mega-rounds for proprietary AI labs like OpenAI, Anthropic, and xAI regularly dominate headlines with multi-billion-dollar price tags, Nous Research has carved out an intensely loyal developer following by creating efficient open software that users can run on local workstations, private servers, or the cloud. Rather than paying purely for hosted model inference, investors in this round are backing a widespread open distribution network and the commercial infrastructure being built on top of it.

The Architecture: How Hermes Agent Works

Photo: TechCrunch (source)

First released under the permissive MIT license in February 2026, Hermes Agent is designed not merely as a chatbot interface or a coding autocomplete plugin, but as a full-fledged autonomous agent harness. Unlike conventional session-based AI assistants that wipe their memories clean once a conversation thread terminates, Hermes maintains persistent memory across restarts, preserves ongoing context, and dynamically writes reusable "skills" to handle repetitive workflows autonomously.

The system runs locally across macOS, Windows, and Linux via a terminal client or desktop application, requiring minimal setup. From an infrastructure standpoint, Hermes is model-agnostic: developers and corporate users can tether it to frontier closed models via API—such as Claude or GPT-series endpoints—or run it entirely offline against local open-weight models like Meta’s Llama family or Nous Research's own specialized open weights. Communication happens directly where workers already spend their time, with native routing through Slack, Discord, Telegram, WhatsApp, Signal, email, and command-line interfaces.

To safeguard execution, Hermes supports five isolated sandboxing environments out of the box, including local containers, Docker, SSH tunnels, Singularity, and Modal, complete with automated container hardening.

Hermes for Businesses: Specifications, Security, and Pricing

With Hermes for Businesses, Nous Research is answering the primary objection enterprise chief technology officers have held against open-source agents: administrative governance. The new corporate offering wraps the open-source Hermes core with the administrative guardrails, privacy guarantees, and centralized oversight required by Fortune 500 IT departments.

Key features in the commercial suite include:

• Centralized Fleet Orchestration: A single pane of glass allowing administrative teams to deploy, monitor, and configure thousands of autonomous agent instances across different corporate units.
• Single Sign-On (SSO) and RBAC: Integration with enterprise identity providers via Okta, Microsoft Entra ID, and SAML/OAuth protocols, paired with granular role-based access control.
• Air-Gapped and Private Deployments: The option to run entirely within an enterprise's on-premises infrastructure or Virtual Private Cloud (VPC), ensuring company data is never retained for external model training.
• Shared Corporate Memory Banks: The ability for cross-functional teams to build and distribute specialized skills and institutional knowledge bases without leaking sensitive credentials.
• Token Routing & Spend Management: Dynamic algorithmic routing heuristics that automatically steer queries to the cheapest and most capable model for a specific task, curbing runaway inference costs.
• Enterprise SLAs and Auditing: Immutable audit logging for security compliance, alongside guaranteed 99.9% platform uptime agreements.

The Origin Story: From Fine-Tuning Llama to Autonomous Agents

Photo: TechCrunch (source)

Nous Research was founded in 2023 by Jeffrey Quesnelle, Karan Malhotra, Ryan Teknium, and Shivani Mitra as an open collective focused on pushing back against centralized artificial intelligence. The group quickly gained prominence within the Hugging Face and GitHub ecosystems by releasing state-of-the-art open fine-tunes of Meta's Llama models, demonstrating that small, focused community models could punch well above their weight against proprietary commercial APIs.

The team also co-authored YaRN (Yet another RoPE extensioN), an efficient algorithm that expanded context windows for open-weight models at a fraction of typical computational training costs. That research set the stage for Hermes Agent. By solving long-context management and combining it with automated synthetic data generation and reasoning architectures, Nous built an agent capable of persistent autonomy.

Adoption accelerated rapidly throughout 2026. According to internal data confirmed by the company, the Hermes harness has been cloned more than 24 million times, with independent mirrors tracking over 22 million direct downloads. Most remarkably, Nous estimates that interactions driven by Hermes deployments worldwide now account for approximately 2.5% of all global AI token consumption.

Battling Big Tech: Open-Source Agents vs. Walled Gardens

The launch of Hermes for Businesses lands squarely in the middle of a high-stakes battle for the corporate AI layer. Major players are racing to cement lock-in through end-to-end proprietary software stacks: OpenAI has pushed agentic features within ChatGPT Enterprise, Meta unveiled its Muse platform, and standalone startups like Instinct have raised eye-popping rounds—such as $1 billion at a $10 billion valuation—to construct vertically integrated autonomous agents.

Nous Research’s bet is fundamentally philosophical and economic: corporations will resist paying premium rents to external model providers if it means surrendering control of their internal corporate intelligence. If a company uses a closed system, switching vendors or self-hosting later becomes prohibitively expensive. By backing an MIT-licensed core supported by enterprise-grade security tooling, enterprise buyers can swap model providers beneath the agent anytime prices or capabilities change.

Executive Vision and Industry Community Reactions

Photo: TechCrunch (source)

In an announcement addressing the Series B round, Nous Research Chief Executive Officer Dillon Rolnick emphasized the startup's lean operating model and ideological commitments.

"Somehow, $90 million is modest in modern fundraising. We at Nous have always delivered more with less, and we are proud of our resourcefulness in an age of waste," Rolnick wrote. "We formed Nous in 2023 to combat the seemingly inexorable trend dragging AI toward centralization and walled gardens. Users don't want lock-in, and they deserve observability into the thing they will allow into so many aspects of their lives."

Industry analysts and venture partners echoed this perspective. Backers point out that with Nvidia, Samsung, and Microsoft's venture wing simultaneously joining the cap table, hardware and platform providers are actively hedging their bets, ensuring that open systems remain viable counterweights to closed proprietary ecosystems.

Practical Playbook: What This Means for Developers and Enterprise IT

For existing individual developers, Nous has reaffirmed that the MIT-licensed Hermes Agent will remain entirely free and open. Engineers can continue cloning the GitHub repository, modifying internal code, and embedding custom integrations without commercial royalty restrictions.

For enterprise IT directors, the arrival of Hermes for Businesses creates an immediate architectural alternative to proprietary enterprise licenses:

• Audit Your Data Exposure: Organizations prohibited by compliance policies (such as HIPAA, GDPR, or SOC2) from feeding proprietary company data into third-party cloud APIs can deploy Hermes instances locally or inside private AWS/Azure clusters.
• Evaluate Multi-Model Arbitrage: Companies spending tens of thousands of dollars monthly on commercial AI endpoints can use Hermes’s cost-routing features to offload mundane tasks to cheap local open weights, reserving costly frontier models only for complex reasoning.
• Standardize Workspace Memory: Instead of individual team members using fragmented, unmonitored chatbot wrappers, IT administrators can provision governed Hermes workspaces that pool team skills while logging activity for audit readiness.

What Happens Next and Lingering Unanswered Questions

While the capital injection cements Nous Research as a legitimate heavyweight unicorn in the AI software sector, questions remain about how smoothly an open-source collective can scale enterprise sales teams and customer support engineering.

Executing white-glove deployments for banks, healthcare providers, and federal contractors requires substantial compliance certification, including formal FedRAMP and SOC2 Type II certifications. Furthermore, as competing open-source agent projects like OpenClaw continue to gain steam, Nous Research will need to continually maintain its developer-first edge while simultaneously servicing enterprise procurement departments. The company plans to roll out Hermes for Businesses onboarding to waitlisted enterprise customers over the coming months.

Frequently Asked Questions

Is the core Hermes Agent still free to use? Yes. Hermes Agent remains completely free and open-source under the MIT license, allowing anyone to download, inspect, host, and modify the core software without paying licensing fees.

What does the $1.5 billion valuation signify for Nous Research? The $1.5 billion valuation establishes Nous Research as an official unicorn, proving that venture capital firms and major tech players see enterprise revenue potential in open-source AI infrastructure rather than exclusively in closed, proprietary model labs.

Which investors participated in the Series B round? The $90 million round was led by Robot Ventures, with participation from Nvidia, Microsoft's M12 venture arm, Samsung, Union Square Ventures, Y Combinator, Menlo Ventures, and 1789 Capital.

How does Hermes for Businesses differ from the open-source version? While using the same core agent logic, Hermes for Businesses adds enterprise governance tools such as Single Sign-On (SSO), fleet management dashboards, audit logging, role-based access control, dedicated security sandboxing, dynamic model cost routing, and uptime service-level agreements (SLAs).

What underlying AI models can Hermes run? Hermes is model-agnostic. It can connect to proprietary cloud endpoints such as OpenAI or Anthropic models, run against hosted multi-model providers via the Nous Portal, or execute fully locally using open-weight models like Meta's Llama series.

Sources

• TechCrunch