Nvidia just paid $6 billion for a startup that couldn't raise $2 billion. The models are free now.

Poolside spent months trying to close a $2 billion funding round.

It failed. Lost access to a 40,000-chip cluster.

Then Nvidia paid $6 billion to license the same technology, hired 109 engineers, and invested another $1 billion. The founders keep the company. Nvidia gets everything that matters.

This is the third time Nvidia has run this playbook.

Groq in December for $20 billion. Enfabrica for $900 million. Total commitment: $27 billion.

Each deal structured as a license plus hiring round, not an acquisition. No antitrust review. The target company keeps operating. Nvidia takes the talent and the IP.

The destination is Nemotron, Nvidia's open-weight model family. Free to download. Free to run. A trillion-parameter version is in training.

The economics are the same as CUDA: give away the software, sell the hardware that runs it. Every company charging for access to mid-tier models now competes against free.

If your AI vendor's business model depends on model licensing revenue, that revenue stream just got a structural threat.

Nvidia is betting that cheaper intelligence creates more demand for compute. Your AI strategy just became a hardware procurement decision.

SOURCE: https://www.forbes.com/sites/jonmarkman/2026/08/24/nvidia-pays-poolside-6b-to-license-its-model-factory-and-109-workers/
VERIFIED: Bloomberg (Aug 21), The Next Web (Aug 21), Forbes (Aug 24)
SIGNAL: Nvidia has now spent $27 billion acquiring AI talent and IP through three license-plus-hire deals, bypassing antitrust scrutiny while building free models that threaten every vendor's licensing revenue. Enterprise AI just became a hardware game.