Nvidia, at this point, is so far ahead of AI hardware games where competing companies are doing the most unlikely thing — working together to catch up and beat the hilarious green giant in its own game. 

Google, Intel, Microsoft, Meta, AMD, Hewlett-Packard Enterprise, Cisco and Broadcom have announced the formation of the tagline "Ultra Accelerator Link Promoter Group" with the aim of creating a new interconnection standard for ai accelerator chips.

Nvidia's proprietary NVLink interconnect technology is used to connect between multiple chips for demanding AI tasks and is very fast, especially when stacked on modern AI hardware. Nvidia Blackwell Gpus support up to 18 NVLink100GB/s connections and deliver a total bandwidth of 1TB/s per 1.8 GPU. 

But because it is a proprietary technology, it creates a closed ecosystem. Which link standard is used determines the hardware, and that is what this new group is intended to address.

The goal of the UALink Promoter group is to create a new open standard that will allow multiple companies to develop AI hardware using new connections (via Ars Technica).It's like Compute Express Link, an open-standard high-speed connection that Intel has developed to link the Cpu and devices in the data center. The first version of the new standard, UALink1.0, is said to be based on technologies such as AMD's Infinity architecture, and is expected to increase speed and reduce latency compared to existing methods.

But there is a catch. Products that utilize the new interconnect technology are said to be available in the next 2 years, and Nvidia is off to a pretty good start.

And in the meantime, Nvidia's overall AI hardware dominance has shown no signs of decline. With the previous generation of h100gpus and tens of thousands of the latest Blackwell Gpus already on sale before the AI chips were announced, companies trying to disrupt Nvidia's dominance in the market at any level will have to take off their little socks.

Still, those are some seriously big names. Nvidia is definitely the darling of AI hardware at the moment, but Google, Microsoft, Intel and AMD are not exactly the lightening of the technology and break at least some of Nvidia's grip on the market

Will Zhen Hoon lose any more sleep, but, somehow, I highly doubt it.