According to reports, Facebook is developing a chip that can power machine learning (ML), which will allow it to recommend content to users.
According to The Information Facebook is currently developing a range of custom chips for its data centers.
One processor is said to be used to power ML tasks such as the algorithm that handles Facebook content recommendations. Another processor would help in transcoding livestreamed video to improve quality.
According to Engadget, Facebook’s move to custom silicon will help reduce its dependence upon outside chipmakers and could also lower its carbon footprint for its data centers,
The company’s current third-party processors are still being used in its servers. According to reports, the new chips will work alongside them and not replace it completely.
Engadget was told by a spokesperson that Facebook is constantly looking for ways to improve compute performance and power efficiency through its silicon partners and our own internal efforts.
The spokesperson stated that they had nothing to share about future plans.
It was also reported that semi-custom chips had been tried by the company.
It announced in 2019 that it was developing an application-specific integrated chip (ASIC), for video transcoding and other inference work.