Nvidia to use $6bn Poolside deal to build open-weight models

US AI leadership will be achieved through an open ecosystem of models, not frontier dominance, Jensen Huang said.

Nvidia is planning to use its $6bn deal with AI start-up Poolside to build open-weight AI models that can rival those made by China’s DeepSeek and Moonshot, the Wall Street Journal (WSJ) reported over the weekend.

The reports come after the chipmaking giant promised a combined $7bn investment into the AI start-up, including $1bn in equity at a pre-money valuation of $12bn. This marks a fourfold jump from the company’s 2025 valuation of $3bn.

Founded in 2023, Poolside is behind the Laguna series of open-weight models, which are small enough to run on a single Nvidia DGX GPU.

Nvidia’s reported plans to make its own open-weight models would also put it in direct competition with other US AI companies it backs, including OpenAI and Anthropic, as global businesses compete for dominance in the space.

Open-weight models are generally far cheaper to run and allow developers to make customisations to suit their needs.

And China is by far the leader in this space, according to data collected by Hugging Face, which found that Alibaba’s Qwen and related models were downloaded more than 3bn times over the past six months, placing it ahead of models from Google, OpenAI, Meta and other rivals by a large margin.

Nvidia founder and CEO Jensen Huang, in a recently published letter, highlighted the importance of open weights for AI leadership. “Our AI leadership will be judged not by one frontier AI model, but by whether the United States builds a strong, open ecosystem that diffuses into every sector,” he wrote.

“Open weights expand access to the AI economy. Start-ups, established businesses, universities and public institutions can build on advanced models without training one from scratch or paying frontier model prices for every task.

“Relying solely on closed models is not inherently safe; they can be breached, misused or fail in ways that outsiders cannot detect. And concentrating advanced AI capabilities behind a small number of closed models compounds that risk.”

Alongside its multibillion-dollar commitment to Poolside, Nvidia is also nabbing the start-up’s 109 employees. The non-exclusive nature of the deal means leadership will not be joining Nvidia and will continue working on their own projects.

Deals like these give Nvidia access to a company’s technical prowess, including research and staff, while minimising impact on the chipmaker’s overall efforts to develop its own technology.

In 2025, the company signed US AI company Groq in a licensing deal that saw much of its talent hired away by the chipmaking giant. The same year, it spent more than $900m to hire Enfabrica CEO Rochan Sankar and other employees at the AI start-up and license the company’s technology.

Nvidia is also reportedly in talks with South Korean chipmaker Rebellions for a similar partnership that could result in an investment or an acquisition.

Poolside employees joining Nvidia will reportedly work on the company’s Nemotron project – which aims to develop open-weight models – according to WSJ sources.

Don’t miss out on the knowledge you need to succeed. Sign up for the Daily Brief, Silicon Republic’s digest of need-to-know sci-tech news.