Meta Platforms (META) is reportedly testing its first in-house AI training chip, giving it the power to design its own custom silicon and rattle current external suppliers like Nvidia (NVDA).
As reported in Reuters today, Meta, alongside Taiwan chip manufacturer TSMC (TSM), has begun a small deployment of the chip and has plans to ramp up production for wide-scale use if the test goes well. Meta shares were up around 1% in pre-market trading with TSM up by a similar amount.
More Power Than GPUs
The drive to develop its own chips is part of a long-term Meta plan to bring down huge infrastructure costs as it strives to develop AI tools in a competitive market both in the U.S. and abroad, particularly China. It would also reduce its reliance on external suppliers like Nvidia.
Again, according to the report, Meta’s new training chip is a dedicated accelerator, which means it is designed to handle only AI-specific tasks. This can apparently make it more power-efficient than the integrated graphics processing units (GPUs) generally used in AI. Being at test stage means that the work is making good progress. It means that Meta has finished the first “tape-out” of the chip, which is simply sending an initial design through a chip factory.
Meta executives have said they want to start using their own chips by 2026 for training, or the compute-intensive process of feeding the AI system reams of data to “teach” it how to perform. The goal for the training chip is to start with recommendation systems and later use it for generative AI products like chatbot Meta AI.
Challenge for Nvidia
Meta has been here before, however, previously scrapping an in-house custom inference chip after it flopped in testing and instead placing more orders for Nvidia GPUs back in 2022. It is one of Nvidia’s biggest customers, using its GPUs to train its models, including for recommendations and ads systems and its Llama foundation model series.
However, the emergence of low-cost models such as China’s DeepSeek has, analysts say, questioned the value of these GPUs. It has also put more focus on so-called inference – the computing resources needed when a user makes a request – than training. This, analysts say, plays into the hands of companies such as Meta rather than Nvidia.
Is META a Good Stock to Buy Now?
On TipRanks, META has a Strong Buy consensus based on 44 Buy, 3 Hold and 1 Sell rating. Its highest price target is $935. META stock’s consensus price target is $764.61 implying an 27.86% upside.
