Meta has begun testing its first in-house AI training chips.

According to a report from Reuters, the company has deployed a small number of chips and, depending on the outcome of the tests, plans to increase production to support their wide-scale use.

Citing two sources, the report said the training chip is dedicated to handling AI-specific tasks. It has reportedly been manufactured by TSMC, with the test deployment following a successful tape-out, the final process before a semiconductor is manufactured.

DCD has reached out to Meta for comment.

It has long been known that Meta has been looking to develop its own chips in order to reduce its reliance on Nvidia hardware.

First reported to be in development in 2023, the company’s in-house chips, dubbed the Meta Training and Inference Accelerator, are based on 7nm nodes and provide 102 Tops of Integer (8-bit) accuracy computation or 51.2 teraflops of FP16 accuracy computation. The chips run at 800 megahertz and are about 370 millimeters square.

Meta was originally expected to roll out its chips in 2022 but scrapped the plan after they failed to meet internal targets, with the shift from CPUs to GPUs for AI training forcing the company to redesign its data centers and cancel multiple projects.

In February 2024, according to documents seen and reported on by Reuters, the company was planning to deploy the second generation of the MTIA chip.

The following year, it was reported that Meta was in talks to acquire South Korean AI chip startup FuriosaAI.