On 13 July 2026, TechCrunch reported that Nous Research is finalising a new funding round of at least $75 million led by Robot Ventures, with participation from Union Square Ventures and other prominent investors. The round values the twenty-six-month-old startup at $1.5 billion. It is the second major capital event in three months for a company that had previously raised $70 million in total and that has built its product in full public view. Hermes, Nous Research's open-source AI agent, now sits at the intersection of venture capital momentum, developer adoption and a quietly growing question inside enterprise procurement departments: is the future of autonomous AI open source?
What Hermes is and how it works
Hermes is not a chat interface. It is an autonomous agent framework designed to run locally on a user's machine, on a virtual private server, or via Nous Research's cloud-hosted tiers. Users can assign tasks that span web search, coding, image understanding and messaging, and the agent executes them over time, reporting back through channels including Telegram and Discord.
According to Nous Research's documentation, Hermes ships with a built-in skills catalogue that covers common automation workflows. The more notable architectural claim is that the agent is designed to learn from usage patterns and generate additional skills without manual engineering. In practice, this means a developer or operations team could deploy Hermes against an internal API, observe how it handles repetitive tasks, and find that the agent has created new capabilities that were not present in the original distribution.
The deployment flexibility matters. Organisations that do not want proprietary data leaving their network can run Hermes entirely on-premise or inside a private cloud. Those that need faster onboarding can use the hosted version, currently priced between $20 and $200 per month depending on usage tier. That dual model is unusual among the major agentic platforms: OpenAI's ChatGPT Work, Anthropic's Claude Cowork and Microsoft's Copilot agents are cloud-native services with limited self-hosting options, while most open-source frameworks require significant engineering investment before they can operate in production.
Hermes has attracted substantial developer attention. According to TechCrunch, the project's GitHub repository has roughly 214,000 stars and nearly 40,000 forks, placing it among the most followed AI-agent projects on the platform.
Why a $1.5 billion valuation matters for the market
A $1.5 billion valuation for an open-source agent framework signals a maturation of investor confidence that extends well beyond the startup itself. Three implications are worth noting.
First, open-source agents are now priced as infrastructure, not experiments. Prior funding rounds in the open-source agent space were modest seed and Series A cheques justified by GitHub stars and Discord activity. A $75 million round at unicorn scale suggests that institutional investors believe Hermes or its commercial derivatives can capture meaningful enterprise software spend. The comparison is not with hobbyist side projects; it is with platforms like Databricks, MongoDB and Confluent, which all built multi-billion-dollar businesses on top of open-source cores.
Second, the venture market is bifurcating around governance models. Closed-source AI platforms have raised at breathtaking valuations partly because they control the model, the infrastructure and the data pipeline. Open-source alternatives have historically struggled to monetise at comparable multiples because users can self-host without paying licence fees. Nous Research's cloud-hosted tiers and enterprise support model may be proving that the hybrid playbook open-source core, paid managed services works for agents in the same way it worked for databases and container orchestration. If so, every major open-source foundation model project will now be courted by venture investors looking for the next Red Hat.
Third, the timing is competitive. Nous Research's fundraising follows closely on the launches of OpenAI's ChatGPT Work, Anthropic's Claude Cowork and Microsoft's Copilot Cowork, all of which target the same desktop and cloud automation use cases. Raising capital now, rather than waiting for post-launch traction metrics, suggests the team believes it can accelerate feature development and channel partnerships while the closed-source incumbents are still defining pricing and data-handling policies.
What it means for enterprise buyers
Enterprise procurement teams evaluating agentic platforms should treat the Hermes funding as a market signal with direct operational implications.
Data residency and sovereignty are now enforceable. Because Hermes can run entirely inside an organisation's own infrastructure, the usual objections about sending proprietary data to third-party model providers do not apply in the same way. For regulated sectors such as financial services, healthcare and government, self-hosted agents offer a compliance path that cloud-only competitors cannot match without undergoing extensive security audits and contractual negotiations. The trade-off is operational complexity: running an agent framework locally requires MLOps, infrastructure monitoring and security patching that many organisations do not yet have in house.
Total cost of ownership is becoming a genuine comparison point. Cloud agent platforms typically charge per seat, per task or per token, and costs scale unpredictably with usage. An open-source agent with a fixed deployment cost and optional paid support can be materially cheaper at scale, especially for use cases involving large document volumes, continuous monitoring or multi-step workflows that consume substantial API credits. Buyers should model three-year costs for both approaches, including the hidden engineering labour of self-hosting, before committing to a vendor.
The skills ecosystem introduces supply-chain risk. Hermes's ability to generate and share skills creates a marketplace dynamic that is powerful but ungoverned. An enterprise that deploys a community-generated skill for Salesforce integration or Jira automation is effectively running third-party code inside its agent layer with the same privileges as the core platform. Security teams should demand the same scrutiny for agent skills that they apply to browser extensions, CI/CD plugins and open-source libraries: code review, sandbox testing, licence compliance and a clear update policy.
What it means for suppliers
For companies building AI agent products, services or tooling, the Nous Research round creates both opportunity and pressure.
The open-source ecosystem is a distribution channel. Vendors that can package proprietary data connectors, compliance overlays or vertical industry skills as Hermes-compatible modules have access to a large and engaged developer community. This mirrors the Salesforce AppExchange or ServiceNow Store models: the platform provides reach, and the partner provides specialised value.
Enterprise support and governance are the monetisable layer. Nous Research's cloud tiers and future enterprise offerings will likely focus on managed infrastructure, audit logging and security policy enforcement rather than core agent features. Suppliers with expertise in these areas, firewalling, identity management, observability, model governance can position themselves as complementary partners rather than competitors.
Closed-source incumbents must refine their lock-in narrative. If open-source agents achieve parity on core autonomy, reasoning and tool-use capabilities, the remaining differentiation will sit in data integrations, security certifications and enterprise workflow depth. Vendors that have relied on model exclusivity as their primary defence will need to invest more heavily in ecosystem partnerships and compliance roadmaps.
The Agentic Expo angle
Agentic Expo exists because the agentic AI market is fragmenting across models, platforms and governance philosophies faster than most enterprise buyers can track. The Nous Research valuation is evidence that open-source agents are not a fringe alternative. They are a funded, growing category with real enterprise deployment paths and serious venture backing.
By March 2027, when Agentic Expo opens at Olympia London, the distinction between closed-source and open-source agent stacks will not be academic. It will be a procurement decision that affects data sovereignty, total cost, security posture and vendor flexibility. The exhibitors and speakers on our floor will be the people who have already made that choice, implemented it, and can explain what worked and what did not. That is the conversation Hermes has just accelerated. Our job is to bring the people having it into the same room.