Qualcomm Teases New Data Center CPUs and AI Inference Accelerators in the Works

Closing out their hour-long, AI-focused Computex keynote, Qualcomm CEO Cristiano Amon ended his presentation with a brief teaser of what’s in store for the server side of Qualcomm’s hardware business.

The company has repeatedly dabbled in the server market with products such as their Cloud AI 100 inference accelerators. But thus far Qualcomm has not gone all-in on the field, despite the hefty profits that Intel (and more recently, AMD) have been reaping from the market. This is despite the fact that they’ve even acquired server hardware designers in the past, most notably Nuvia, who was originally working on developing their custom Phoenix CPU core for servers before being re-tasked with developing the equally custom Oryon core for use in client and smartphone processors.

But, as teased by Qualcomm at the end of their Computex keynote, the company is getting ready to make its long-awaited incursion into the data center market. Speaking briefly on the subject and backed by a single slide, Amon confirmed that Qualcomm is working on both server-grade CPUs and a new AI inference accelerator for the data center market. Noting the similarities between breaking into the server market and breaking into the client market – as Qualcomm just performed – Amon noted that he believes there’s room for Qualcomm if they have something “unique and disruptive.” In this case, the company seems to be heavily focused on energy efficiency, a similar tack that they’ve taken with Snapdragon X for PC and mobile.

At this point Qualcomm is not sharing any concrete details on their hardware other than that they’re busy working on the hardware. But a press release from NVIDIA earlier in the day for NVLink Fusion included a choice quote from Qualcomm, who is participating in the program and is developing an NVLink Fusion-compatible server CPU – the same upcoming CPU server mentioned here.

So, as Amon said in his keynote, stay tuned for more details.