Report: Meta Plans Shift to Liquid Cooling in AI-Centric Data Center Redesign
As Meta' adapts its data centers for an AI-powered future, the new design will feature more extensive use of liquid cooling, according to new reporting from Reuters.
The strategic shift in Meta's data center design is focused on optimizing new facilities to better support artificial intelligence, and was announced shortly after the release of ChatGPT in November, which highlighted the disruptive potential of AI chatbots that can create content.
In December, Mets paused construction on a number of its data center projects to "rescope" its new data centers with the pending design focused on AI, which is the lynchpin of Meta founder Mark Zuckerberg's vision for an immersive metaverse community.
Meta's new design could provide an additional boost to the nascent market for liquid cooling. The company operates more than 40 million square feet of data centers and has dozens of data centers under construction across its global network. A large buyer like Meta could give a shot in the arm to liquid cooling, which has been focused on high-performance computing (HPC) and supercomputing. Google has already shifted its AI infrastructure to liquid cooling, while Microsoft is testing immersion cooling in its production data centers.
Networking, Cooling May Get a Revamp
Officially, Meta is not revealing any details of it design update. "The new data centers that we're building are intended really for future year capacity," said Meta CFO Susan Li in Thursday's earnings call. "So they won't come online for a few years. So we don't have too much more to share there, except that, that work is ongoing."
But Reuters reports that Met will make more extensive use of GPUs (graphics processing units), which generate more heat than traditional CPUs and will require more high-density cooling capacity and different networking infrastructure.
"The facilities needed 24 to 32 times the networking capacity and new liquid cooling systems to manage the clusters' heat, requiring them to be 'entirely redesigned,' " Reuters reported, citing internal documents from Meta.
Last fall DCF reported on Meta's plans to gradually add more liquid cooling in its data centers. which included adoption of cold plates to provide direct-to-chip cooling for AI workloads on its GPU servers, as well as several designs for managing the temperature of supply water as rack power densities increase.
At the time, Meta Vice President for Engineering Alexis Bjorlin said that " the power trend increases we are seeing, and the need for liquid cooling advances, are forcing us to think differently about all elements of our platform, rack and power, and data center design,”
That effort accelerated in late 2022, according to the Reuters report, which says that Meta has focused on boosting compute power across its infrastructure to adapt to a changing AI landscape. That includes more GPUs in the short term and custom AI silicon down the road, according to Reuters.
The new reporting includes some good news for three U.S. communities where Meta has paused its data center construction. A Meta spokesperson told Reuters that data center construction that was paused while transitioning to the new designs would resume later this year. Local media reports indicate Meta has paused construction at projects in Huntsville, Alabama; Temple, Texas and Kuna, Idaho.
In February, Meta announced plans to implement generative AI in its applications, with a goal of making them available by December 2023.
"As these new models and use cases continue scaling, we're going to need to continue investing in infrastructure, although we'll have a better idea of the trajectory of that investment later in the year once we can gauge usage of some of the new products we'll launch," said Zuckerberg.
