Not Chips. Not Memory. Data Center Cooling Is the Unsung AI Infrastructure Play Nobody's Watching
Investors should take a closer look at Vertiv, as it plays a crucial role in the AI infrastructure ecosystem.
Overview
The artificial intelligence (AI) infrastructure build-out has supercharged the growth of chip designers, foundries, memory manufacturers, and other companies in the semiconductor ecosystem.
This isn't surprising, as semiconductors are the basic hardware layer of AI data centers. Basic AI tasks, such as processing data and storing parameters, wouldn't be possible without chips. So, it is easy to see why the semiconductor sector has boomed in recent years, given its role as the foundational hardware layer of the AI infrastructure ecosystem.
Details
However, running AI tasks using these chips generates a lot of heat. An AI data center rack equipped with high-end graphics cards usually draws 40-100 kilowatts (kW) of power, which is well above the 5-10 kW power draw of a traditional server processor-based rack. Higher power consumption necessitates efficient cooling to enable AI accelerator chips to perform at their full potential.
Source
Originally published at www.fool.com.