Enlarge image
Superlative chips:
Nvidia boss
Jensen Huang
presented new products in San Jose in mid-March
Photo: Josh Edelson / AFP
Founder and CEO
Jensen Huang
(61) created Nvidia, the chip and software giant for the age of artificial intelligence (AI), now worth around $2.2 trillion on the stock market. Without the computing power of its special chips, the boom in generative AI, from which Microsoft, OpenAI and Google parent Alphabet are also benefiting, would be unimaginable. Just last week, Huang introduced the next generation superchip: the "GB200", consisting of 600,000 parts and weighing around 1.5 tons.
Now a coalition of prominent technology companies, which also includes Qualcomm, Google and Intel, is planning a counterattack on Nvidia's previous quasi-monopoly. They want to attack the chip giant's secret weapon: the software that binds development teams worldwide to Nvidia chips. Based on the almost twenty-year-old computer code, the CUDA software platform, more than four million developers are constantly programming new AI and other applications. The software is therefore considered to be just as important for Nvidia as the hardware in order to secure its own market position.
However, the group of financiers and companies looking to challenge Nvidia's dominance in artificial intelligence is growing. As the consortium's attack on the software now shows. "We're showing developers how to migrate from an Nvidia platform," said
Vinesh Sukumar
, Qualcomm's head of AI and machine learning, in an interview with Reuters.
Starting with a technology developed by Intel called OneAPI, the consortium, calling itself the UXL Foundation, plans to develop new software and various tools that can power multiple types of AI accelerator chips, group executives told Reuters. The open source project aims to make computer code run on any machine, regardless of what chip and hardware powers it. So in plain language: even without Nvidia.
"It's about how we can create an open ecosystem specifically around machine learning frameworks and drive productivity and hardware choice," said Google's director and chief technologist for high-performance computing, Bill
Hugo
, in an interview with Reuters . Google is one of the founding members of UXL, which was founded last September. The group is helping to define the technical direction of the project, says Hugo.
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 a "mature" state by the end of the year, it said. In addition to the companies originally involved, the UXL heads also want to court cloud computing companies such as Amazon and Microsoft as well as other chip manufacturers. So far, however, they are not included.
The UXL plans to focus its resources on the most pressing computing problems dominated by a few chip makers, such as the latest AI applications and high-performance computing applications. These early plans feed into the organization's longer-term goal of attracting a critical mass of developers to its platform. In the long term, the UXL also wants to support hardware and code from Nvidia.
At Nvidia, however, they are calm, as Nvidia manager
Ian Buck
explained in a statement to Reuters: "The world is getting faster and faster. New ideas in the area of accelerated computing are coming from the entire ecosystem, and this will help AI and to advance the scope of what accelerated computing can achieve."
The UXL consortium's plans are just one of many attempts to break Nvidia's influence on the software that powers AI. Venture capitalists and companies have collectively poured more than $4 billion into 93 different ventures, PitchBook data shows. Last year the sector experienced a real boom. Startups aiming to weaken Nvidia's leadership position raised just over $2 billion in 2023, compared with $580 million the year before, according to PitchBook data.
However, only a few will probably be successful. Jensen Huang's CUDA software is considered convincing on paper because it has a wide range of functions and is constantly growing thanks to contributions from both Nvidia and the developer community.
lhy/Reuters