Meta Ventures into Chip Design: A Step Towards AI Self-Reliance

Meta, owner of Facebook, is testing its first in-house AI training chip in a bid to reduce dependence on external suppliers like Nvidia. The initiative involves a partnership with TSMC and marks a crucial step in Meta's broader plan to cut infrastructure costs and enhance AI capabilities.

Meta, the tech giant known for social media platforms Facebook, Instagram, and WhatsApp, is venturing into chip design with its first in-house chip for AI training systems. This strategic move aims to lessen the company's dependence on external suppliers such as Nvidia, sources revealed to Reuters.

The decision to develop in-house chips is part of Meta's long-term strategy to decrease infrastructure costs as it bets heavily on AI tools to fuel growth. Should the testing phase prove successful, Meta could expand production substantially. The company's projected expenditures for 2025 are between $114 billion to $119 billion, with a significant portion dedicated to AI infrastructure.

Partnering with TSMC, a Taiwan-based chip manufacturer, Meta has already commenced a small-scale deployment of the chip following a successful tape-out. The program, under the Meta Training and Inference Accelerator series, aims to replace reliance on Nvidia GPUs and potentially lead innovations in recommendation systems and generative AI products.

(With inputs from agencies.)