Meta is reportedly testing in-house chips for AI training

It is reported that AI systems are reported to be examining an in-house chip for training a part of the strategy to reduce dependence on hardware manufacturers like Nvidia.

According to ReutersMater Chip, which is designed to handle AI-specific work pressure, was created in the Taiwan-based firm TSMC part. The company is running a “small installation” of the chip and plans to increase production if the test is successful.

Meta has previously deployed custom AI chips, but only to run models – not training them. As a Reuters Note, several of the company’s chip design attempts have been canceled or otherwise scale after failing to meet internal expectations.

Meta is expected to spend $ 65 billion for capital expenditure this year, most of which will go towards Nvidia GPUs. If the company transfers to in-house chips and manages to reduce a fraction of that expenditure, it will be a big win for the social media giant.