Founders Jason Warner and Eiso Kant stay as Nvidia also invests $1B at a $12B pre-money valuation, according to reports.
By Ryan Merket · Published
Primary source: The Wall Street Journal
Why it matters #
Nvidia is securing the technology and talent to build U.S. open-weight models, expanding beyond chips while preserving Poolside as an independent supplier.
Nvidia has agreed to pay $6 billion for a non-exclusive license to Poolside's AI model-development technology and offer jobs to 109 Poolside employees, giving the chipmaker a ready-made operation for building U.S. open-weight models. The Wall Street Journal reported the agreement on August 22nd, describing Nvidia's goal as a direct challenge to Chinese labs including DeepSeek and Moonshot AI, as well as closed-model providers OpenAI and Anthropic.
Poolside co-founders Jason Warner (@jasoncwarner) and Eiso Kant (@eisokant) are expected to remain at Poolside. Nvidia will also invest $1 billion in Poolside at a $12 billion pre-money valuation, according to reports based on a Poolside investor letter. Poolside plans to distribute the $6 billion licensing payment to its investors by the end of 2027.
The arrangement gives Nvidia access to the system Poolside calls its Model Factory, along with job offers for many of the researchers and engineers who built it. The license is non-exclusive, leaving Poolside able to keep using the technology while Nvidia incorporates the software and any employees who accept its offers.
Warner was GitHub's chief technology officer and oversaw the organization that incubated GitHub Copilot before spending two years as a managing director at Redpoint Ventures. When Warner founded Poolside with Kant in 2023, he described software development as Poolside's route toward artificial general intelligence. Kant previously founded developer-software businesses source{d} and Athenian, according to his Axon board biography.
Nvidia is licensing the production line
The most valuable asset in the transaction is the process behind Poolside's models. Poolside describes the Model Factory as an integrated collection of data pipelines, distributed training software, evaluation systems, inference infrastructure and reinforcement-learning tools. Poolside built the system to replace slow, manually coordinated training projects with a repeatable development process.
Poolside's technical report for Laguna M.1 and Laguna XS.2 says both models were trained from scratch within the Model Factory. The models consumed more than 30 trillion training tokens, and Poolside says it moved Laguna XS.2 from the start of training to release in five weeks.
Poolside followed those releases with Laguna S 2.1 in July, a 118-billion-parameter mixture-of-experts model with 8 billion active parameters per token. Poolside says the model supports a context window of up to 1 million tokens and competes with much larger models on agentic coding benchmarks. Those performance figures come from Poolside's evaluations, although Poolside published its final evaluation trajectories for outside inspection.
Nvidia already supports the Laguna architecture in its NeMo AutoModel documentation. The licensing agreement moves Nvidia beyond supporting Poolside's models on Nvidia hardware. Nvidia gains the machinery and personnel needed to produce additional models under its own direction.
A liquidity event with no acquisition
The $6 billion payment also provides a substantial return for Poolside's existing backers without transferring ownership of Poolside. Poolside raised a $26 million seed round led by Redpoint Ventures and a $100 million Series A led by Felicis Ventures before closing a $500 million Series B in 2024.
Bain Capital Ventures led that Series B, with DST Global, StepStone Group, Schroders Capital, Premji Invest, Dorsal Capital, BAM Elevate, Adams Street and Fin Capital participating. Nvidia, Citi Ventures, Capital One Ventures, HSBC Ventures, LG Technology Ventures and eBay Ventures also joined the round.
The transaction follows a structure Nvidia has already used for scarce AI technology and personnel. Nvidia signed a non-exclusive technology license with Groq in December 2025 while hiring Groq founder Jonathan Ross and other employees. Nvidia previously paid more than $900 million to license Enfabrica technology and recruit Enfabrica executives and engineers.
Licensing-and-hiring transactions leave the target operating independently and avoid the change of control produced by a conventional purchase. Similar structures involving Microsoft and Inflection, Amazon and Adept, and Google and Character.AI have already drawn regulatory attention. A Federal Trade Commission staff report warned that large technology partnerships with AI developers can affect access to computing resources, talent and sensitive technical information.
For Nvidia, Poolside supplies an American open-weight effort without requiring Nvidia to build the entire research operation internally. It also gives Nvidia a model strategy that does not depend on the closed labs buying its chips. Open-weight models can be downloaded, customized and run on infrastructure controlled by customers, making them particularly useful for governments and regulated industries that cannot send sensitive data to an external API. Poolside now faces a different operating test. Warner and Kant retain Poolside, its technology and a new $1 billion investment, while 109 employees decide whether to join Nvidia. Nvidia gets immediate access to the model-building process Poolside spent three years assembling. Poolside keeps the factory, but it may have to restaff the floor.