Nvidia pays Poolside $6B in licensing deal

Nvidia has agreed to pay AI coding startup Poolside $6 billion to non-exclusively license Model Factory, the software Poolside built to train its own models, and will separately invest $1 billion in the company at a $12 billion pre-money valuation, according to reporting from Newcomer confirmed by TheNextWeb and PYMNTS on August 20-21. Nvidia is also extending hiring offers to 109 Poolside staff who worked on Laguna, Poolside’s open-weight coding model. Poolside’s co-founders, Jason Warner and Eiso Kant, are staying, and the company says it will keep operating independently. It plans to pass the $6 billion licensing payout through to its investors sometime in 2027.

The deal is explicitly structured to avoid being called an acquisition or an acquihire. It is the third time in roughly eight months Nvidia has used this playbook: a non-exclusive IP license plus a talent-hiring offer, layered with an equity stake, to absorb the core output of an AI startup without buying it. Nvidia used similar structures with Groq (a $20 billion licensing-and-hiring deal in December, after which Groq still raised $650 million and later a $350 million Series C with Nvidia participating) and with hardware startup Enfabrica (roughly $900 million). Across the three deals, Nvidia has now committed close to $27 billion.

What it means for you

If you’re building on Poolside’s Laguna models or evaluating them as a routing option against other Chinese and Western open-weight releases, nothing changes immediately: Poolside says it’s staying independent and the founders aren’t leaving. But treat “took Nvidia money” as a weaker signal than it used to be. These licensing-plus-hiring deals let Nvidia harvest a startup’s best researchers and its training pipeline while leaving the shell company technically un-acquired, which means Poolside’s future roadmap now runs through people who no longer report to Poolside. It’s the same financing logic behind Nvidia’s move to become an AI financier more broadly: buy the compute-hungry demand, structure around merger review, keep every dollar flowing back into GPUs.