Top AI Chipmaker Looks To Make Its Own AI Models

Nvidia reportedly plans to spend billions of dollars to compete against China’s most advanced open-source artificial intelligence models and its own American customers.

Nvidia, the world’s largest chipmaker for the AI hardware market, announced a licensing deal with AI startup Poolside that would have the company directly compete against OpenAI and Anthropic to build its own open-source models, according to the Wall Street Journal, citing people familiar with the matter. Open-weight models typically cost less and allow for more customization compared to closed-source models such as Anthropic’s Claude or OpenAI’s ChatGPT.

Nvidia will invest $1 billion in the AI startup and pay $6 billion to license Poolside’s technology and hire many of its top engineers, per the Wall Street Journal.

The startup’s founders said that the deal would ensure that AI “would not be a closed technology controlled by few but one built by many out in the open.”

An Nvidia spokesperson did not immediately respond to a request for comment from the Daily Caller News Foundation.

Over 100 Poolside employees will help develop Nvidia’s Nemotron project, which aims to develop open-weight models that the company released in 2023, the Journal reported.

American AI companies have focused on closed-source AI models instead of developing open-source models, which allow any user to download and alter them at will. China has emphasized developing open-source models, which has created concerns that too many will flock to Chinese AI over American closed-source models, according to the outlet.

One top Nvidia executive touted the benefits of open-source technologies.

“Open technologies are generally safer because there’s more sunlight,” Bryan Catanzaro, the vice president of applied deep learning research at Nvidia, said in August.

The top chipmaker has worked to develop and promote open-source AI models.

The company in July launched the Open Secure AI Alliance, a coalition that includes many leading tech companies such as Palantir, IMB, Crowdstrike and SpaceX.

Nvidia in March launched the Nemotron Coalition — a group consisting of open-source developers such as Mistral, Thinking Machine Labs and Perplexity — to share, data and computing resources. The company at the same time backed another startup whom aims to become the DeepSeek of the West,” a reference to the popular Chinese AI lab that develops popular open-weight models.

China’s open models most advanced AI models, Moonshoot’s Kimi K3 and Z.AI’s GLM models, trail America’s frontier models by only a few months. Nvidia hopes that its licensing deal with Poolside can produce a version of Nemotron that could compete with the most advanced models.

The Poolside-Nvidia deal puts the chipmaker in direct competition with large frontier large such as OpenAI and Anthropic, some of Nvidia’s biggest customers. However, some experts noted that large AI labs continue to design their own AI chips to power their models, which may already put them in direct competition with Nvidia, the Wall Street Journal reported.

The company at the same time backstopped the asset value of an OpenAI data center, which could become one of the world’s largest, according to the New York Times.