Major technology companies including Meta, Microsoft, Advanced Micro Devices and Broadcom announced Thursday that they have developed a new industry standard for networking in AI data centers, the latest effort to break the dominance of the leader in Nvidia market.
The “Ultra Accelerator Link” is an attempt to establish an open standard for communication between artificial intelligence accelerators – systems that can help process the large amounts of data used in AI tasks.
Other members include Alphabet, owned Google, Cisco System, Hewlett Packard Enterprise and Intel.
Why it matters
Nvidia, the largest player in the AI chip market with a share of around 80%, is not part of the group.
Tech giants like Google and Meta want to reduce their dependence on Nvidia, whose networking business is a critical part of the package that enables their AI dominance. Broadcom’s main rival in the networking and custom chip market – Marvell Technologies, is also not part of the group.
Key quote
“An industry specification becomes essential to standardize the interface for AI and machine learning, HPC (high performance computing) and cloud applications for the next generation of data center and AI implementations,” they said. the companies said in a statement.
Context
Technology companies are investing billions of dollars in the hardware needed to support AI applications, increasing demand for AI data centers and the chips they run on.
The Ultra Accelerator Link group has designed specifications governing the connections between different accelerators in a data center.
Specifications will be available in Q3 2024 for companies joining the Ultra Accelerator Link (UALink) consortium.
The answer
An Nvidia spokesperson declined to comment. Marvell did not immediately respond to a Reuters request for comment.
