{"slug": "nvidia-expands-open-weight-ai-push-with-poolside-deal", "title": "NVIDIA Expands Open-Weight AI Push with Poolside Deal", "summary": "NVIDIA announced a $6 billion deal to license AI technology from startup Poolside and a separate $1 billion investment, valuing Poolside at $12 billion pre-money. The deal brings over 100 Poolside engineers to work on NVIDIA's Nemotron open-weight AI models, as NVIDIA expands its push to compete with Chinese open-weight models like DeepSeek and proprietary models from Anthropic and OpenAI. Poolside remains independent, and the license is non-exclusive, allowing Poolside to sell to NVIDIA's competitors.", "body_md": "NVIDIA has announced a $6 billion deal to license AI technology from startup Poolside, while separately investing $1 billion in the company, as it expands its development of open-weight AI models.\n\nThe deal gives NVIDIA access to Poolside’s AI development technology and brings more than 100 of the startup’s engineers into NVIDIA. The engineers will work on Nemotron, Nvidia’s family of open-weight AI models. Poolside will remain an independent company, with the $1 billion investment valuing the startup at $12 billion before the new funding.\n\nThe agreement is part of a major expansion of NVIDIA’s AI strategy. The dominant provider of GPUs for AI, NVIDIA is putting far more resources into the models that run on its hardware. Its goal is to build an open-weight model capable of competing with Chinese offerings such as DeepSeek and Moonshot AI’s Kimi, while also presenting an alternative to proprietary models from Anthropic and OpenAI.\n\n“The license is non-exclusive, so Poolside can sell the same system to NVIDIA’s competitors,” said Mitch Ashley, VP and Practice Lead at The Futurum Group. “Standardizing on a coding model commits you to whoever runs its pipeline, because benchmarks age faster than a two-year platform bet.”\n\nHe advised enterprise buyers: “Ask who owns that pipeline before you sign.”\n\n**China and Open-Weight**\n\nOpen-weight models give developers access to model parameters, allowing them to modify and deploy the technology for their own requirements. This approach can lower operating costs and give businesses more control than closed platforms.\n\nChina has emerged as a major force in this segment. DeepSeek attracted global attention in early 2025 with a low-cost open model, raising concerns that Chinese developers could gain an advantage as AI adoption spreads internationally. Other Chinese developers, including Moonshot and Z.AI, have also released competitive open models.\n\nNVIDIA CEO Jensen Huang has argued that US leadership in AI will depend partly on creating an open ecosystem that can spread across industries. The company formed the [Nemotron Coalition](https://nvidianews.nvidia.com/news/nvidia-launches-nemotron-coalition-of-leading-global-ai-labs-to-advance-open-frontier-models) in March with participants including Mistral, Thinking Machines Lab and Perplexity, bringing together computing resources and technical expertise.\n\nThe Poolside transaction gives NVIDIA another large infusion of engineering talent and technology for this effort. Poolside’s internal Model Factory, used to develop its AI models, is included in the license.\n\nPoolside was founded in 2023 by software developer Eiso Kant and former GitHub CTO Jason Warner. Its work has focused heavily on AI for software development, including government and defense applications. The company recently released Laguna S, an open-weight model that gained traction among Western developers.\n\nKant, Warner and operations executive Margarida Garcia will remain outside NVIDIA and continue research work.\n\nThe deal follows a challenging period for Poolside. The startup sought $2 billion in financing late last year to secure a 40,000-unit GPU GB300 cluster scheduled to become available in January. It failed to raise the money within the required six-week period and lost access to the cluster. Without additional computing capacity and capital, Poolside concluded that its ability to compete in frontier AI would soon be constrained.\n\nNVIDIA, of course, faces few such resource limitations. The company can combine Poolside’s technology and engineers with its enormous AI computing infrastructure as it develops larger Nemotron models. NVIDIA recently released Nemotron 3.5 Lightning and is working on a much larger model that could contain more than one trillion training parameters.\n\nThe strategy also creates a particularly complex competitive situation. OpenAI and Anthropic are major NVIDIA customers, so a stronger Nemotron could compete directly with their models. Several large AI developers are also creating their own processors, making them potential competitors to NVIDIA in silicon.", "url": "https://wpnews.pro/news/nvidia-expands-open-weight-ai-push-with-poolside-deal", "canonical_source": "https://techstrong.ai/articles/nvidia-expands-open-weight-ai-push-with-poolside-deal/", "published_at": "2026-08-24 19:02:21+00:00", "updated_at": "2026-08-24 19:13:03.592976+00:00", "lang": "en", "topics": ["artificial-intelligence", "ai-policy", "ai-startups", "ai-products", "ai-infrastructure"], "entities": ["NVIDIA", "Poolside", "Nemotron", "DeepSeek", "Moonshot AI", "Anthropic", "OpenAI", "Jensen Huang"], "alternates": {"html": "https://wpnews.pro/news/nvidia-expands-open-weight-ai-push-with-poolside-deal", "markdown": "https://wpnews.pro/news/nvidia-expands-open-weight-ai-push-with-poolside-deal.md", "text": "https://wpnews.pro/news/nvidia-expands-open-weight-ai-push-with-poolside-deal.txt", "jsonld": "https://wpnews.pro/news/nvidia-expands-open-weight-ai-push-with-poolside-deal.jsonld"}}