-advertisement-

Meta Platforms unveiled details about the next generation of the company's in-house artificial intelligence accelerator chip.

Meta planned to deploy a new custom data center chip, internally referred to as "Artemis," to handle the increasing computing power needed for AI products in Facebook, Instagram, and WhatsApp, reducing reliance on Nvidia's AI chips and cutting energy costs.

"This chip’s architecture is fundamentally focused on providing the right balance of compute, memory bandwidth, and memory capacity for serving ranking and recommendation models," the company wrote in a blog post.

The new Meta Training and Inference Accelerator chip is part of the company's broader custom silicon initiative, which also involves exploring other hardware systems.   Beyond building the chips and hardware, Meta has heavily invested in software to optimize the efficiency of its infrastructure.

The company, spending billions on purchasing Nvidia and other AI chips, intends to buy around 350,000 flagship H100 chips from Nvidia this year, according to CEO Mark Zuckerberg. Along with other suppliers, Meta aims to acquire a total of 600,000 H100 chips this year, he added.

Taiwan Semiconductor Manufacturing Co, which will produce the new chip on its "5nm" process, Meta said it is capable of three times the performance of its first generation processor.

The chip has been deployed in the data center and is engaged in serving AI applications. The company said it has several programs underway "aimed at expanding the scope of MTIA, including support of (generative AI) workloads."

Editor：Lulu

▼▼▼

US blacklists 4 Chinese firms for helping military get AI chips

Google unveils Arm-based data center processor, new AI chip

Western Digital confirms HDD and NAND flash shortages, warns partners of higher pricing

Intel, Dell, and HP face lawsuit in France amid legal battle

Biden to give Taiwan's TSMC $6.6 billion to ramp up US chip production

Samsung wins Nvidia’s 2.5D package order