Equinix is preparing for the generative-AI boom by arming more than 100 of its data centers with advanced liquid-cooling technologies.
The data-center operator has begun work on building the infrastructure necessary to support new technology, such as direct-to-chip cooling and rear-door heat exchangers, in more than 100 of its data centers in 45 locations around the world, including London, the San Francisco Bay Area, and Singapore.
Equinix is one of the largest data-center companies in the world. Its 250 data-center facilities rent space, hardware, and broadband to other companies to store data and run workloads in the cloud.
As companies start to experiment with AI applications, industry experts have raised concerns over the vast amounts of power those workloads require to run. Generative-AI applications run on GPUs, which are more powerful than chips used to run regular cloud services — but also heat up much faster.
"It is absolutely essential that data-center providers enable more businesses to use the most performant cooling technologies to support their AI and other compute-intensive workloads," Jon Lin, the executive vice president and general manager of data-center services at Equinix, said.
The data centers will include new liquid-cooling technology, which helps lower the temperatures of Equinix's machines. This lets their fans to work at lower speeds, saving energy and money.
In addition, the data centers will use direct-to-chip technology powered by cold plates on top of chips inside servers, as well as rear-door heat exchangers, which work like radiators and allow cool air to flow into a rack of servers as hot air flows out.
Lin said the rise of generative AI, along with increasing customer interest in energy efficiency and sustainability, drives the push for "standardization" of the latest technologies.
Equinix is working quickly, Lin said. The first of the company's data centers to host the new technology will be completed in the first quarter of 2024. Lin said Equinix already offers in-rack heat exchangers at most of its locations and has typically provided more advanced technologies on a "bespoke" basis.
"The availability of highly efficient data-center capacity will be foundational to mitigating the carbon impacts of advanced computing," Lin said.
