In an attempt to challenge market leader Nvidia’s dominance, major tech companies, including Meta, Microsoft, Advanced Micro Devices, and Broadcom, said on Thursday that they had created a new industry standard for networking in AI data centres.
An open standard for communication between artificial intelligence accelerators, or systems that may assist in processing the enormous volumes of data used in AI jobs, is being attempted to be established with the “Ultra Accelerator Link.”
Hewlett Packard Enterprise, Intel, Cisco Systems, and Google—all controlled by Alphabet—are among the other members. Not included in the grouping is Nvidia, the industry leader in AI chips with a market share of almost 80%.
Giants in the IT industry like Google and Meta are eager to wean themselves off of Nvidia, whose networking division is a crucial component of the package that allows for its dominance in AI. Marvell Technologies, Broadcom’s main competitor in the networking and custom chip markets, is likewise excluded from the group.
Important quote
“For the next generation of AI data centres and implementations, an industry specification becomes critical to standardise the interface for AI and Machine Learning, HPC (high-performance computing), and Cloud applications,” the businesses stated in a statement.
Context: The demand for AI data centres and the chips they run on is rising as a result of tech businesses investing billions of dollars in the hardware needed to enable AI applications.
Specifications controlling connections between various accelerators in a data centre have been designed by the Ultra Accelerator Link group.
In the third quarter of 2024, companies that become members of the Ultra Accelerator Link (UALink) Consortium will have access to the specifications.Big Tech Develops AI Networking Standard—But Without Chip Leader Nvidia
In a surprising move, major tech giants have come together to establish a new AI networking standard—without the participation of Nvidia, the undisputed leader in AI chips. This development signals a shift in the AI industry, as companies seek to create open, interoperable systems that reduce reliance on Nvidia’s proprietary technologies.
The initiative, backed by industry heavyweights like Google, Microsoft, Meta, and Intel, aims to create a universal framework for AI model training and deployment. As AI workloads grow increasingly complex, efficient networking between GPUs, CPUs, and AI accelerators is crucial for performance. The new standard focuses on improving scalability, reducing bottlenecks, and allowing seamless integration across diverse hardware ecosystems.
Nvidia, which dominates the AI chip market with its CUDA software and InfiniBand networking technology, has notably been left out of this collaboration. Some analysts believe this exclusion reflects growing concerns over Nvidia’s market control, with competitors pushing for a more open AI infrastructure. Others speculate that Nvidia’s proprietary approach does not align with the open-standard vision championed by the consortium.
By developing an independent AI networking framework, Big Tech companies hope to lower costs, increase competition, and foster innovation. However, Nvidia’s absence raises questions about adoption, given its dominant position in AI hardware. Whether this new standard can challenge Nvidia’s ecosystem remains to be seen, but it marks a significant step toward a more diversified AI industry.
Group Media Publication
Construction, Infrastructure and Mining
General News Platforms – IHTLive.com
Entertainment News Platforms – https://anyflix.in/
Legal and Laws News Platforms – https://legalmatters.in/
Podcast Platforms – https://anyfm.in/
%20/23%20cimr%204.jpg "India is selected to host the ICCC Conference in 2027.")