Meta is Testing its First In-House Chips With TSMC: Report
Designed specifically for AI tasks, the new chip functions as a dedicated AI accelerator
Meta has begun testing its first in-house chip designed to train AI systems, according to a report by Reuters. This initiative is part of the company's strategy to reduce reliance on external chip suppliers like NVIDIA and cut AI infrastructure costs.
Meta is collaborating with Taiwan Semiconductor Manufacturing Company (TSMC) for the chip’s production. The chip belongs to the Meta Training and Inference Accelerator (MTIA) series. If the tests are successful, Meta plans to ramp up production and expand its use.
The report highlights that Meta's AI-related expenses are a significant portion of its projected 2025 costs, which are expected to range between $114 billion and $119 billion, including up to $65 billion in capital expenditures.
Designed specifically for AI tasks, the new chip functions as a dedicated AI accelerator, offering greater efficiency than the general-purpose GPUs typically used in AI training.
Meta’s chip program has faced challenges in the past. After scrapping a previous inference chip due to poor test results, the company reverted to purchasing billions of dollars worth of NVIDIA GPUs in 2022.
However, last year, Meta successfully deployed a custom chip for AI inference in recommendation systems on Facebook and Instagram.
Other companies investing heavily in AI such as Google, Microsoft, and Amazon, have also developed their in-house chips with TSMC.
