Nvidia Pays $6B to License Poolside's AI Model Factory

Nvidia has agreed to pay $6 billion to license AI model development software from startup Poolside while separately investing $1 billion in the company at a $12 billion pre-money valuation, according to a shareholder letter reviewed by The Wall Street Journal.

The arrangement centers on Poolside's Model Factory, the internal system the startup used to develop its AI models. The license is non-exclusive, allowing Poolside to offer the same technology to other companies. More than 100 Poolside engineers will transfer to Nvidia to work on the company's Nemotron open-weight model project, though the three founders will remain independent to pursue unspecified research.

Why it matters

This transaction reveals how Nvidia is positioning itself beyond chip sales to compete directly in AI model development against frontier labs like OpenAI and Anthropic, as well as Chinese competitors including DeepSeek. The deal structure—licensing plus investment rather than acquisition—allows Nvidia to secure critical talent and technology while sidestepping the regulatory scrutiny that traditional buyouts would trigger in today's antitrust environment.

A pattern of strategic licensing

The Poolside arrangement follows Nvidia's established playbook. The company previously paid Groq $20 billion for access to its inference technology and key engineers, while Groq continued operating independently and raising additional capital. A similar deal with chip-interconnect startup Enfabrica cost approximately $900 million. These three transactions alone represent roughly $27 billion in commitments structured to avoid acquisition-related regulatory review.

Born from failed fundraising

The deal emerged after Poolside struggled to secure the computing resources needed to remain competitive. The shareholder letter disclosed that the startup had a six-week window at the end of last year to raise $2 billion for a 40,000 GB300 cluster scheduled to come online in January. When that fundraising effort fell short, Poolside lost access to the cluster.

The $6 billion licensing payment will be distributed to Poolside's investors before the end of next year. The shareholder letter emphasized that the arrangement is "not an acquisition and it is not an acquihire."

Building competitive open-weight models

Nvidia plans to use the licensed technology and incoming engineers to develop one of the most capable open-weight AI models available. This effort puts the chip maker in direct competition with Chinese models such as DeepSeek and Kimi K3, as well as leading U.S. labs. Poolside's own open-weight model, Laguna S, had been positioned as a Western alternative to those Chinese offerings.

The details were first reported by The Wall Street Journal and The Next Web.