Nvidia is making a major investment in open-weight AI through a new deal with startup Poolside. The company will pay $6 billion to license Poolside’s AI technology and invest another $1 billion in the startup at a $12 billion pre-money valuation.
The deal also brings more than 100 Poolside engineers to Nvidia to work on Nemotron, its family of open-weight models. Nvidia plans to use Poolside’s technology and engineering experience to improve Nemotron as it competes with Chinese models including DeepSeek and Kimi K3 and U.S. offerings from OpenAI and Anthropic.
Poolside will continue operating independently, with its leadership staying at the company. For Nvidia, the deal adds technology and engineering talent to an open-weight AI effort already underway as Chinese models gain ground.
Why It Matters: A stronger Nemotron could give companies another option outside the proprietary models that make up much of the U.S. market. Nvidia already supplies much of the computing power behind enterprise AI and is now moving further into the models themselves. If Nemotron gains traction, Nvidia could play a larger role in enterprise AI environments.
- The Deal Covers Poolside’s Model-Building Technology: At the center of the $6 billion license is Model Factory, the internal system Poolside used to develop its AI models. The license is non-exclusive, so Poolside can continue offering the technology elsewhere. The licensing payment is expected to go to Poolside investors before the end of next year.
- More Than 100 Poolside Engineers Are Joining Nvidia: The engineers will work on Nemotron, including some of Nvidia’s largest models under development. Poolside founders Eiso Kant and Jason Warner and operations executive Margarida Garcia will stay with the startup and continue separate research. Poolside will remain independent even after much of its engineering team moves to Nvidia.
- Poolside Lost Access To A 40,000-GPU Cluster: Late last year, Poolside had six weeks to raise $2 billion for a 40,000-GPU GB300 cluster scheduled to come online in January. It did not secure the financing in time and lost the cluster. Poolside had already developed Laguna S, an open-weight model described as a Western alternative to Chinese models.
- Nvidia Is Building Up Its U.S. Open-Weight Effort: DeepSeek, Kimi K3 and other Chinese models have gained ground while many leading U.S. offerings remain proprietary. Nvidia CEO Jensen Huang has argued that U.S. leadership in AI will depend in part on building a strong open ecosystem. Nemotron gives Nvidia its own open-weight model family for that effort, while also putting it in competition with U.S. model developers including OpenAI and Anthropic.
- Nemotron Is Part Of A Larger Open-Model Effort: Nvidia launched the Nemotron Coalition in March with developers including Mistral, Thinking Machines Lab and Perplexity to share data, expertise and computing resources. The company has also backed Reflection AI, another developer focused on open models. Poolside adds another source of technology and engineering talent to work Nvidia already has underway.
Go Deeper -> Nvidia is paying $6 billion to license AI model software from startup Poolside – Quartz
Nvidia Is Spending $6 Billion to Build a Powerful U.S. Alternative to Chinese AI – WSJ
