Facebook Is Developing Its Own Artificial Intelligence Chips For Its Servers

Tech giants with large servers are trying to design their own chips to reduce both cost and dependency.

Facebook is getting ready to be among the companies that design its own custom ARM-based silicon. According to The Information’s report, a family of proprietary chipsets is currently being developed to speed up machine learning tasks.

Giants like Google, Amazon and Microsoft spend millions of dollars to manufacture their own chips to save financially and get better performance from servers.

The social media giant has been scrambling to recruit engineers with experience designing FPGAs and ASICs, and their efforts actually stretched back to 2018. A little later, the company announced its plans to develop a project for data centers with the help of partners such as Intel, Qualcomm, Marvell, Esperanto and Habana.

But the new report shows that the social giant has changed its mind and is developing the new chips entirely in-house. A company spokesperson said:

“Facebook is always exploring ways to achieve higher levels of computing performance and power efficiency through our silicon partners and our own internal efforts.”

Facebook is also developing a chip for video transcoding to enhance the infrastructure that enables video and live streaming in its applications. This project looks pretty similar to what Google is doing with “Argos” Video Encoding Units (VCUs) to speed up transcoding of videos uploaded to YouTube.