Facebook developing in-house machine learning chips

Facebook is designing a range of chips focussed on machine learning workloads such as content recommendations and video transcoding, The Information reported yesterday.

The social media company joins the likes of Google and Apple in the move away from chip suppliers to save on costs and increase server performance.

Citing two people familiar with the project, the report said that the one chip in development would improve its apps’ video-streaming experience.

The other would improve the precision of content recommendations on the social media platform.

According to the report, implementing inexpensive, more powerful chips could help Facebook reduce its environmental impact at its continually expanding data centres.

In 2019 Facebook indicated that it would be developing its own custom chip designs for AI tasks and video transcoding in conjunction with major custom silicon players such as Qualcomm, Intel, and Marvell.

According to a Facebook spokesperson, the company “is always exploring ways to drive greater levels of compute performance and power efficiency with our silicon partners and through our own internal efforts.”

Main Image: web