NVIDIA Pays $6 Billion for Poolside Model Factory License, Extends Job Offers to 109 Employees
NVIDIA has agreed to pay $6 billion for a non-exclusive license to develop AI models using Poolside's software, while also extending job offers to 109 employees and investing an additional $1 billion.
A letter sent to investors by Poolside indicates that the license pertains to the company's "model factory" system used for building AI models. The transaction is explicitly not an acquisition or acquisition-style employment, and the three founders will remain in their positions, with the company continuing to operate independently.
NVIDIA is also investing $1 billion in Poolside at a pre-money valuation of $12 billion, bringing the post-money valuation to approximately $13 billion. The $6 billion licensing fee is planned to be distributed to Poolside's existing investors by the end of 2027.
Poolside focuses on coding AI agents and developing open-source models, with its Laguna series models running on NVIDIA chips. This transaction marks NVIDIA's third similar structural arrangement recently, having previously adopted a licensing plus talent plus equity model with other AI companies.
Details of the transaction show that NVIDIA is rapidly acquiring model-building capabilities and related talent while avoiding regulatory scrutiny associated with a full acquisition. Poolside, on the other hand, receives significant capital and continues to develop independently.
From a market mechanism perspective, funds are flowing from NVIDIA to AI model development startups, strengthening the chip giant's positioning in software and talent. The event is driven by technology acquisition, benefiting Poolside and its early shareholders who receive licensing fees and investments, while independent model developers face intensified competition from larger companies.
Source: Public Information
ABAB AI Insight
NVIDIA has repeatedly secured AI software and talent through non-acquisition structures in recent years, with similar transactions involving Groq and Enfabrica, indicating its accelerated path toward full-stack model building beyond its hardware dominance.
On the capital front, NVIDIA mobilizes cash reserves through a dual-track approach of licensing fees and equity investments, directly injecting resources into model factory technology and core teams, motivated by the desire to shorten its model iteration cycle and bind upstream software capabilities, with specific actions including non-exclusive licensing and targeted talent invitations.
Similar cases illustrate large tech companies' "licensing + talent" model for AI startups, as well as early chip manufacturers' acquisition paths to expand into software ecosystems; the current AI industry is at a stage where model factory capabilities are becoming scarce assets, leading larger companies to prefer flexible binding rather than full acquisitions.
Essentially, this represents a technological substitution: model-building software and talent are migrating from independent startups to chip platforms, with the mechanism being to rapidly internalize R&D results through high licensing fees while retaining the independence of the licensee to avoid regulatory and integration costs.
ABAB News · Cognitive Law
The $6 billion purchase is not for the company, but for the key to the model factory.
Non-exclusive licensing allows both large companies and startups to accelerate simultaneously.
Talent invitations are more flexible than acquisitions and harder to be halted by regulators.