SANTA CLARA, UNITED STATES — Nvidia is reportedly paying $6 billion to license Poolside technology and recruiting 109 employees as it builds out its open-weight AI model strategy.
Nvidia has agreed to pay about $6 billion for a non-exclusive license to artificial-intelligence technology developed by Poolside and has offered jobs to 109 of the startup's employees, according to multiple reports citing people familiar with the transaction and an investor communication.
The arrangement would give the world's dominant supplier of advanced AI processors access to technology Poolside has used to develop and train its own models without requiring Nvidia to acquire the company outright. Nvidia is also reportedly investing another $1 billion in Poolside at a $12 billion pre-money valuation.
The distinction is significant. The reported $6 billion payment is a technology license rather than a purchase price for Poolside, and the license is described as non-exclusive. Poolside's founders are expected to remain with the startup while much of its technical workforce is offered positions at Nvidia.
Neither Nvidia's newsroom nor Poolside's public news pages reviewed Sunday contained a detailed announcement setting out the transaction terms. The figures should therefore continue to be attributed to the reporting until either company makes its own contractual disclosure.
Poolside technology could accelerate Nvidia's model ambitions
Poolside was founded in 2023 and has focused heavily on artificial intelligence for software development.
Its current work includes the Laguna family of models and infrastructure for training and operating AI agents. Poolside's own website says its broader objective is to pursue increasingly capable AI systems through software, while its recent technical work has included models designed for coding and long-running agentic tasks.
The reported Nvidia agreement centers not just on individual model weights but on Poolside's broader model-development technology. The Next Web, citing the investor letter, described the licensed asset as software Poolside used to build AI models and reported that 109 employees received Nvidia job offers.
That could give Nvidia both technology and personnel for an area in which it has been expanding aggressively: developing its own families of open and open-weight models.
The company already produces Nemotron, a family designed for reasoning, agents, coding and other enterprise applications. Nvidia's official materials describe Nemotron 3 as an open model family built to support specialized agentic AI, with models, training data and supporting software forming part of a broader developer ecosystem.
More recent Nvidia materials say Nemotron has expanded further into multimodal, voice, coding and complex workflow applications.
Nvidia moves further up the AI stack
The transaction illustrates how far Nvidia's strategy has moved beyond selling GPUs.
The company remains best known for the processors used to train and run large AI models, giving it a central position in the infrastructure boom created by OpenAI, Anthropic, Meta, Google and other developers.
But Nvidia increasingly supplies much more than silicon: networking equipment, software libraries, cloud services, development frameworks, AI models and financing arrangements surrounding large-scale computing infrastructure.
Earlier this month, Nvidia announced partnerships with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR designed to mobilize more than $500 billion of third-party capital over time for AI computing infrastructure.
Reuters also reported this month that Nvidia committed to guarantees of up to $105 billion connected with an OpenAI data-center project in Ohio, alongside a $1.5 billion investment in project developer SB Energy.
The Poolside arrangement would extend that expansion in a different direction: deeper into the intellectual property and talent required to create the models that ultimately consume Nvidia's computing infrastructure.
Open-weight models are becoming more strategically important
Nvidia's interest in Poolside comes as competition over open-weight models intensifies.
Unlike fully closed systems whose weights remain controlled by their developers, open-weight models make trained parameters available under terms that can allow outside developers to modify, fine-tune or operate them independently.
Nvidia's Nemotron license explicitly permits commercial use and creation and distribution of derivative works subject to its terms.
The company has repeatedly characterized openness as strategically important to AI development. When it introduced Nemotron 3, Nvidia said the family was intended to give developers the ability to build efficient and specialized AI agents while retaining greater transparency and control over their models.
The Wall Street Journal reported that Nvidia wants Poolside employees to help develop its largest Nemotron models as part of an effort to create a stronger U.S.-based open-weight alternative to systems developed by Chinese groups including DeepSeek and Moonshot AI.
That objective also brings Nvidia closer to competition with some companies that buy enormous quantities of its hardware.
OpenAI, Google, Meta and other major Nvidia customers develop their own model technology, while several are simultaneously pursuing proprietary AI accelerators designed to reduce dependence on Nvidia chips.
A stronger Nvidia model portfolio therefore creates a complicated strategic relationship: the company can sell infrastructure to virtually every major AI developer while increasingly offering software and models that overlap with parts of their businesses.
Poolside remains a company after the reported deal
Another unusual aspect is what Nvidia apparently is not buying.
Newcomer and other outlets report that Poolside will remain independent. Its founders — Jason Warner, Eiso Kant and other senior leadership — are not expected to become Nvidia employees under the current arrangement.
The reported $1 billion Nvidia investment would instead capitalize the company that remains after the license and employee moves, at a $12 billion pre-money valuation.
That structure follows a difficult period for Poolside.
The Financial Times reported in April that a proposed $2 billion financing had stalled after problems surrounding a large data-center project and the collapse of an arrangement with CoreWeave. At the time, Nvidia had been discussing an investment of as much as $1 billion in the fundraising.
Poolside had previously raised $500 million in 2024 at a $3 billion valuation, according to the FT.
The new structure, if completed on the reported terms, therefore gives Poolside investors substantial value from the technology license while leaving a separately financed company in operation.
The next disclosure will matter
The headline $6 billion figure is large even by the standards of the current AI investment cycle, but several details remain unresolved publicly.
It is not yet clear how the license payment will be recognized financially, precisely which technology and intellectual-property rights Nvidia receives, whether all 109 employees will accept their job offers or when the operational transfer will be completed.
The non-exclusive nature of the reported license also means Poolside retains rights to the technology rather than surrendering it entirely to Nvidia.
Those terms make the transaction materially different from a conventional acquisition and are central to understanding its economic significance.
For Nvidia, the clearest strategic consequence is that the company would gain a substantial group of model-building engineers and access to an established AI-development system as it tries to make Nemotron more competitive.
For Poolside, the arrangement potentially converts years of model-development work into billions of dollars without formally selling the company.
The next important step is an official Nvidia or Poolside disclosure establishing the final transaction terms and confirming which reported elements have been completed.
More from Business
View all in this section- Loading related stories…