The UXL Foundation, a consortium consisting of technology companies such as Google, Intel and Qualcomm, wants to break Nvidia’s AI dominance.

Today, Nvidia is worth $2,200 billion and has a near-monopoly position with its extremely powerful AI chips. This is due to the unique position that Nvidia finds itself in. CUDA, the software platform trusted by more than 4 million developers to build AI and applications, only works on Nvidia hardware.

A consortium of technology companies wants to put a stop to this. Google, Intel, Qualcomm and other parties are joining forces under the name UXL Foundation. Speaking to Reuters, Qualcomm said it is already “actively reaching out to developers to show them how they can migrate from the Nvidia platform.” Intel CEO Pat Gelsinger wants to shake the tree: “The entire industry wants to eliminate the CUDA market”

The UXL Foundation’s endowment is provided by Intel. Called OneAPI, the company aims to develop a suite of software and tools that support multiple types of chips that accelerate AI applications. OneAPI is open source and should eventually run on every machine, regardless of chip or hardware.

The technical team within the UXL Foundation aims to finalize the technical specifications in the first half of 2024. The engineers will then work on working out the technical details by the end of 2024. Google, Intel and Qualcomm emphasize that the collaboration must be solid and include diverse contributions from multiple companies. “This is the only way we can ensure that it runs on any type of chip or hardware.”

The UXL Foundation was founded in September 2023 and has already received several technical contributions from external parties, including both the founders and outsiders, all open source. In the long term, UXL also wants to support Nvidia hardware and code. The report says nothing about AMD, whether it is part of the UXL Foundation or not. Today you can unofficially run Nvidia CUDA on AMD hardware via the open source tool Zluda.

Intel’s OneAPI can already be used today. The next step is to create a standard programming model specifically for AI. Nvidia reacts neutrally to the news. “The world is speeding up. New ideas in the field accelerated computing come from across the ecosystem and that will help drive AI and the scale of it accelerated computing can achieve in order to move forward.”