Wednesday, Oct 7 | --:--
Back to home

Nvidia’s $6B Poolside License Is How You Buy a Model Lab Without Buying the Company

Newcomer published Poolside’s investor letter: Nvidia will pay $6 billion for a non-exclusive license to Model Factory, invest $1 billion at a $12 billion pre-money valuation, and offer jobs to 109 Laguna engineers. Founders stay; Poolside says it is not an acquisition. Bloomberg and The Information confirmed August 21. Largest specimen yet of the Groq-shaped license-plus-hire structure.

Times of AI Desk 5 min read Santa Clara, CA View as Markdown
Cover illustration for Nvidia’s $6B Poolside License Is How You Buy a Model Lab Without Buying the Company

Six billion dollars for a license plus a 109-person lift-out, while the target stays “independent,” is how Nvidia buys a model lab without buying a company. It is the largest specimen yet of the Groq-shaped structure.

Newcomer published a Poolside letter to investors. Nvidia will pay $6 billion for a non-exclusive license to Model Factory — the stack behind Poolside’s Laguna open-weight coding models — invest $1 billion at a $12 billion pre-money valuation, and extend offers to 109 employees. Eiso Kant, Jason Warner, and Margarida Garcia stay. The letter: “not an acquisition and it is not an acquihire.” The $6 billion is slated to be distributed to existing investors by end of 2027. Bloomberg (updated Aug 21) and The Information / WSJ corroborated. Neither company issued a newsroom primary; the letter is the source.

Deal sheet

Piece Letter / Newcomer Aug 20 + Bloomberg/TI Aug 21
License $6B, non-exclusive, Model Factory
Equity $1B @ $12B pre-money (~$13B post)
Talent Offers to 109 Laguna engineers → Nvidia Nemotron
Founders Remain at Poolside; unspecified research continues
Structure precedent Same shape as Nvidia’s Groq license-plus-hire (Dec 2025), which lawmakers called a merger-review workaround
Distribution License proceeds to LPs by YE 2027 (Inovia: $76.20/share cited in secondary coverage)

Poolside told investors that competing in open-weight models without more Nvidia hardware was not viable. Nvidia gets a U.S. open-weight coding factory to put against DeepSeek / Kimi K3 and its own Nemotron 3.5 Lightning (Aug 11).

Not Nemotron 3.5 (Aug 11), not Groq’s $350M Series A (Aug 17) — that Groq is the neocloud that remained after the earlier license. Not PORTS-Pike.

Limits

  • No Nvidia 8-K in the research window; terms are the shareholder letter as reported.
  • Treat as exclusive + concrete $, not a signed SEC primary.
  • Whether antitrust treats the third deal as a pattern, and whether Laguna weights actually land inside Nemotron, is open.

Sources

Prior Coverage

Earlier Times of AI reporting on this thread.

Scroll to continue reading