Technology giant Nvidia, known for its cutting-edge artificial intelligence (AI) chips, finds itself at a crossroads. With a market capitalization of $2200 billion, the firm has established itself as the beating heart of the new era of generative AI, powering developers ranging from innovative start-ups to behemoths such as Microsoft, OpenAI and Alphabet, Google’s parent company.
But beyond its revolutionary hardware, it’s nearly two decades of computer code that has cemented Nvidia’s near-incontested position. More than 4 million developers worldwide rely on Nvidia’s CUDA software platform to build AI applications, making competition with the company almost impossible.
However, a coalition of technology companies, including Qualcomm, Google and Intel, is determined to break Nvidia’s stranglehold by targeting the chip giant’s secret weapon: the software that keeps developers in its orbit. These companies are part of a growing group of financiers and corporations taking aim at Nvidia’s supremacy in AI.
“We’re really showing developers how to migrate off a Nvidia platform,” explained Vinesh Sukumar, head of AI and machine learning at Qualcomm, in an interview with Reuters.
Armed with an Intel-developed technology called OneAPI, the UXL Foundation, a consortium of technology companies, is planning to build a suite of software and services that will enable AI to be used in a wide range of applications.is planning to build a suite of software and tools capable of powering various types of AI acceleration chips, executives involved in the group told Reuters. This open-source project aims to make computer code executable on any machine, regardless of the type of chip and hardware used.
“It’s specifically – in the context of machine learning frameworks – about creating an open ecosystem and promoting productivity and choice in hardware,” explained Bill Magro, director and chief technologist of high-performance computing at Google. Google is one of the founding members of UXL and is contributing to the technical direction of the project, Magro added.
The UXL technical steering committee is preparing to finalize the technical specifications in the first half of this year. Engineers plan to refine the technical details to reach a “mature” state by the end of the year, executives said. They stressed the need to build a solid foundation to include contributions from multiple companies that can also be deployed on any chip or hardware.
From one ecosystem to another
Beyond the companies initially involved, UXL will seek to attract cloud computing companies such as Amazon and Microsoft’s Azure, as well as other chip manufacturers.
Since its launch in September, UXL has already begun to receive technical contributions from third parties, including foundation members and external players keen to use the open-source technology. Intel’s OneAPI is already usable, and the next step is to create a standard programming model designed for AI.
UXL plans to focus its resources on the most pressing computing problems, currently dominated by a few chip manufacturers, such as the latest AI applications and high-performance computing applications. These initial plans are part of the organization’s long-term goal of rallying a critical mass of developers to its platform.
Ultimately, UXL also plans to support Nvidia hardware and code. Asked about open-source and venture-backed software efforts to break Nvidia’s dominance in AI, Ian Buck, an Nvidia executive, said in a statement, “The world is speeding up. New ideas in accelerated computing are emerging from across the ecosystem, which will help advance AI and the scope of what accelerated computing can achieve.”
Around a hundred startups
The UXL Foundation’s plans are just one of many attempts to eradicate Nvidia’s hold on the software that powers AI. Backers have invested more than $4 billion in 93 separate initiatives, according to custom data compiled by PitchBook at Reuters’ request.
Interest in dethroning Nvidia via a potential weakness in software has intensified over the past year, and startups aiming to break into the company’s leadership gobbled up just over $2 billion in 2023, up from $580 million the previous year, according to PitchBook data.
Succeeding in the shadow of Nvidia’s group on AI data processing is a feat few startups will be able to achieve. Nvidia’s CUDA is a compelling piece of software on paper, as it is comprehensive and continues to grow, both through Nvidia’s contributions and the developer community.