Facebook is working on a chip that will form the basis of a machine learning accelerator, according to The Information, citing two people familiar with the project. In particular, the chip should help create content recommendations for users. According to the report, the company also developed a video encoding chip to enhance the video experience for its applications.
Apparently, Facebook has decided to follow the lead of other big tech companies like Apple, Amazon and Google, which are increasingly ditching publicly available chips in favor of in-house designed products to lower costs and improve productivity.
Back in 2019, Facebook announced that it was working on creating chips specifically designed to handle artificial intelligence and video transcoding tasks. In doing so, the company wanted to improve the performance, power and efficiency of its infrastructure, which at the time served 2.7 billion users across all of the company’s platforms.
Facebook has also said it will work with semiconductor companies such as Qualcomm, Intel and Marvell Technology to create their own specialized chips, as general purpose processors will not be enough to handle the amount of workload that Facebook systems handle. However, a report from The Information says that Facebook is developing these chips on its own.