Nvidia has struck one of the most unusual deals in recent AI memory: a $6 billion non-exclusive license to Poolside AI’s Model Factory, the internal platform the startup used to build its Laguna family of coding models. On top of that, Nvidia is investing $1 billion into Poolside at a $12 billion pre-money valuation, and 109 Poolside employees have received Nvidia job offers.

The three founders are staying put. Poolside’s leadership was emphatic in its letter to investors: “This is not an acquisition and it is not an acquihire.”

What Poolside Built

Poolside spent several years building its Model Factory, a purpose-built platform for producing AI models that can write, debug, and optimize code. The infrastructure behind it runs 10,000 to 20,000 experiments per month, with streaming data feeding directly into training, reproducible experimentation pipelines, and agents that increasingly write the code used to train future models.

The output is the Laguna model family. The key variants are:

• Laguna XS.2 - An open-weight model for local agentic coding workflows, available to any developer
• Laguna S 2.1 - A 118-billion-parameter open-weight model that matches or outperforms much larger systems on coding benchmarks
• Laguna M.1 - A 225-billion-parameter Mixture of Experts model built for government, defense, and regulated enterprise environments that need data sovereignty and maximum reasoning capability

These models solve a genuine problem for enterprises: they want AI that can handle long-horizon software engineering tasks, runs on infrastructure they control, and can be deployed without routing sensitive code through third-party APIs.

What Nvidia Gets

Nvidia is licensing the platform itself, not just the trained models. The Model Factory is the manufacturing process, not the product, and the license is non-exclusive, meaning Poolside can sell access to it elsewhere. Nvidia gets the infrastructure know-how it needs to produce specialized coding AI at scale.

This follows a pattern Nvidia has been building out. Rather than acquiring AI companies outright, the company licenses technology, hires substantial portions of teams, and leaves the original entity operating. It is the third deal Nvidia has structured this way.

What Poolside Gets

For Poolside, this is a clean outcome that lets the founders keep building. The $6 billion goes to investors by the end of 2027. The $1 billion in fresh equity keeps the company funded and valued at $12 billion. The founders retain control.

The non-exclusive nature of the license matters: Poolside can still sign deals with other AI infrastructure players, cloud providers, or enterprise software companies who want the same manufacturing capability.

What This Means for Business

The enterprise software development market is undergoing a real transformation. Businesses are no longer asking whether AI can help with code. They are asking how they maintain control over the models doing the work, particularly in sectors where regulatory compliance and data sovereignty matter.

The Model Factory approach Poolside pioneered, and that Nvidia just paid $6 billion to access, signals something important: the real value in enterprise AI is shifting from individual models to the systems that build and maintain them. A model is a point-in-time artifact. A model factory is a continuous capability.

For companies evaluating AI-assisted software development today, a few things follow from this deal:

Agentic coding is moving mainstream. Nvidia does not spend $7 billion acquiring capabilities for niche markets. The enterprise demand for AI that can autonomously write, test, and iterate on software is large enough to justify this investment.

Open-weight models are becoming a serious enterprise option. Poolside’s Laguna S 2.1 being open-weight is not accidental. Government agencies and regulated businesses that cannot send code to external APIs need models they can run on their own infrastructure. The market for self-hostable frontier coding AI is real.

The who-trains-the-model question is getting more interesting. When Nvidia is licensing model factory infrastructure, not just buying models, it is thinking about manufacturing capability at scale. The same logic applies to enterprises considering which AI development platform they rely on.

For companies already using AI for development work, the Poolside/Nvidia deal is a signal that the underlying infrastructure is maturing fast. For companies that have not yet deployed AI in their development workflows, the gap between leaders and laggards is widening.

If your business runs on software, your competitive position increasingly depends on how well you can build and ship it. The tools are becoming more capable. The question is whether your organisation is positioned to use them.

[Enterprise DNA’s Omni team works with businesses to assess and deploy AI across development and operational workflows. If you want to understand what this shift means for your specific situation, book a call with Sam.]