Starting with Intel Meteor Lake (14th Generation) CPUs, now part of the new Intel Core Ultra Processor (Series 1) brand, an integrated Neural Processing Unit or NPU is built right into the SoC (system-on-chip) and is optimized for low power AI Inferencing.
I found this article from Chips and Cheese about the new Intel Meteor Lake NPU to be super insightful, definitely recommend a read if you are new to NPUs.
While you can already consume the Intel integrated graphics (iGPU) in platforms like the Intel NUC with ESXi for workload inferencing, I was curious on whether this new Intel NPU could actually be used by ESXi? 🤔