All of Us Against You Alone : Tech Giants Unite to Develop AI Chip to Rival NVIDIA

Intel, Google, Microsoft, Meta, and other major tech companies have joined forces to establish a standard for next-generation AI accelerator chips.
Nvidia Chip

A multitude of major tech companies, including Intel, Google, Microsoft, Meta, AMD, HP, Broadcom, and Cisco, have announced the formation of a group called UALink. This group aims to establish standards for next-generation AI accelerators. Through their efforts, the new generation of AI platforms, which are entirely composed of GPUs, will be standardized and developed collaboratively.

Generally speaking, AI accelerators are systems used to provide the computational capacity needed for models to perform their tasks, encompassing everything from GPUs to sophisticated hardware. The effort to establish a new standard inevitably touches on NVIDIA. AMD General Manager Forrest Narrod’s statement, “The industry needs an open standard that is not constrained by a single company,” highlights this situation.

A new standard for a 1024-GPU pod is emerging.

A new standard for a 1024-GPU pod is emerging_0

According to the UALink 1.0 proposal, pods consisting of at least 1024 graphics processing units (GPUs) installed in a single server system will establish the initial standard. This standard will be compliant with “open standards,” meaning they will be able to accommodate structures like AMD’s Infinity Fabric. With this system, data processing will be faster, and latency will be reduced.

A new standard for a 1024-GPU pod is emerging_1

The group stated that they will establish a consortium for this standard and developments will be monitored by this team. It remains to be seen what NVIDIA, which is thought to control between 80% to 95% of the sector, will do in response to this initiative.

Scroll to Top