Meta shares details on AI progress and data center redesign

After announcing that it was canceling or pausing data center projects that were in mid-development last year, Meta has finally revealed some details on its AI-optimized data centers. 

The giant has shared that it is working on a custom chip "family" to bolster the efficiency of its data centers, will use liquid cooling to support a "significant percentage" of its A.I. hardware, and use less carbon-intensive materials in construction. It is set to break ground on its first facility this year, with the new design 31% cheaper and build twice as quickly as its older data centers. 

According to Meta CEO Mark Zuckerberg, the company has been building advanced infrastructure for A.I. "for years": "This work reflects long-term efforts that will enable even more advances and better use of this technology across everything we do," he said.

​​In a recent blog post 'Reimagining our infrastructure for the A.I. age', Santosh Janardhan, VP and head of infrastructure, said: "Our artificial intelligence (A.I.) compute needs will grow dramatically over the next decade as we break new ground in A.I. research, ship more cutting-edge A.I. applications and experiences across our family of apps, and build our long-term vision of the metaverse."

Janardhan added: "These initiatives aim to facilitate the development and deployment of larger and more sophisticated A.I. models at scale."

Aparna Ramani, VP of Engineering, Infrastructure at Meta said that the company is creating new hardware, including its own silicon. Adding: "Thousands of engineers are innovating on this large-scale infrastructure that's built specifically for A.I."

Many of these announcements were made at the recent AI Infra @ Scale event, where it announced the completion of the second-phase buildout for its Research Supercluster (RSC). This is an AI supercomputer that includes 2,000 DGX A100 systems, with 16,000 A100 GPUs, for  large language models (LLMs) like the LLaMA project.