Skip to main content

Meta is working on a new chip for AI

Meta is working on a new chip for AI

/

It’s also working on a ‘next-generation data center design’ that’s optimized for AI.

Share this story

Image of Meta’s logo with a red and blue background.
Illustration by Nick Barclay / The Verge

Meta is building its first custom chip specifically for running AI models, the company announced on Thursday. As Meta increases its AI efforts — CEO Mark Zuckerberg recently said the company sees “an opportunity to introduce AI agents to billions of people in ways that will be useful and meaningful” — the chip and other infrastructure plans revealed Thursday could be critical tools for Meta to compete with other tech giants also investing significant resources into AI.

Meta’s new MTIA chip, which stands for Meta Training and Inference Accelerator, is its “in-house, custom accelerator chip family targeting inference workloads,” Meta VP and head of infrastructure Santosh Janardhan wrote in a blog post. The chip apparently provides “greater compute power and efficiency” than CPUs and is “customized for our internal workloads.” With a combination of MTIA chips and GPUs, Janardhan said that Meta believes “we’ll deliver better performance, decreased latency, and greater efficiency for each workload.”

The MTIA could be a big boon for Meta, especially given increasingly high demand for AI compute power. But the MTIA chip is seemingly a long ways away: it’s not set to come out until 2025, TechCrunch reports.

In addition to the MTIA, Meta is also introducing a new ASIC specifically to help with video transcoding, which it calls the “MSVP,” or Meta Scalable Video Processor. It’s designed to support both “the high-quality transcoding needed for VOD as well as the low latency and faster processing times that live streaming requires,” Meta said in a separate blog post, and “in the future,” it will help bring things like AI-made content and AR- and VR-specific content to Meta’s apps.

Meta is also working on a “next-generation data center design” that will be “AI-optimized” and “faster and more cost-effective to build,” Janardhan said, and the company also touted the power of its Research SuperCluster (RSC) AI Supercomputer, “which we believe is one of the fastest AI supercomputers in the world.” This isn’t exactly new rhetoric about the RSC from Meta; the company has been sharing high praise about the supercomputer since last year. But as the company tries to stand out against an increasing number of AI initiatives from many of the biggest technology giants — including other custom chips —it makes sense that Meta would want to brag about its faith in its own AI hardware.