Meta has announced the development of four new computer chips aimed at bolstering generative AI features and content ranking systems within its applications. These chips, part of the MTIA series, mark Meta’s venture into producing its own AI hardware alongside its existing collaborations with industry leaders like Nvidia. The new semiconductors, crafted in partnership with Broadcom and based on the RISC-V architecture, are being manufactured by Taiwan Semiconductor Manufacturing Corporation, a prominent player in the chip industry.
The MTIA 300 chip, already in production, focuses on training algorithms used to rank and recommend content to millions of users across Meta’s platforms such as Facebook and Instagram. Additionally, Meta plans to roll out the MTIA 400, 450, and 500 chips between early and late 2027 to support inference tasks, which involve executing trained AI models to generate outputs like text and images. Meta’s vice president of engineering, YJ Song, emphasized the iterative approach taken in developing these chips to adapt rapidly to evolving AI workloads.
Meta’s foray into designing its own hardware signifies a significant shift for the social media company, traditionally reliant on external suppliers for its computing infrastructure. This move underscores Meta’s commitment to enhancing its AI capabilities and highlights the growing importance of specialized hardware in powering advanced AI functionalities.
Source: WIRED