On July 13, Bloomberg spotted an interesting development: more evidence that social media giant Facebook (NASDAQ:FB) is getting serious about developing its own chips.
Apparently, Facebook hired Shahriar Rabii to serve as a vice president leading the company's chip-development efforts.
As Bloomberg observed, Rabii's LinkedIn profile indicates that he worked at Alphabet's Google from October of 2014 to July of 2018 as a senior director of engineering, heading up Google's "silicon engineering, product/program management, production and Technology Engineering." His profile says he worked on products such as "Pixel Visual Core for ML [machine learning] and computational photography, Titan family of secure elements, VP9 and AV1 video transcoders and others."
Bloomberg also says that Rabii will "work under Andrew Bosworth, the company's head of virtual reality and augmented reality," citing "people familiar with the matter."
Let's take a closer look at why Facebook continues to beef up its chip ambitions.
Doubling down on VR and AR?
The Facebook chip team mentioned here seems to be focused on building chips for its consumer-oriented hardware -- in other words, its Oculus family of virtual reality headsets.
Back in October of 2017, Facebook announced a virtual reality headset called the Oculus Go. This headset differed from its other virtual reality headsets in that it incorporated a chip that handled all required processing tasks, eliminating the need to connect it to a computer to handle that processing.
That headset didn't use a Facebook-designed chip, but one developed by wireless-chip giant Qualcomm (NASDAQ:QCOM). The chip inside of the Oculus Go is a relatively dated Snapdragon 821 processor, but Qualcomm has since announced a more powerful chip dedicated to virtual and augmented reality applications, known as the Snapdragon XR1.
Although I wouldn't be surprised if several future iterations of Facebook's Oculus Go headsets used newer, more powerful Qualcomm chips, there now appears to be a risk that Facebook is planning to develop its own chips for such applications, potentially cutting Qualcomm out longer-term.
Why would Facebook go in-house?
One big reason that device manufacturers move from third-party-merchant chips to in-house chips is to enable a much greater degree of control over the capabilities of their chips. A third-party vendor needs to build chips to satisfy a wide range of customers, with potentially different product ambitions and cost structures.
A company that develops its own chips has the luxury of tailoring them to the needs of the devices they'll be going into.
There's room for differentiation as well. Even if Facebook develops its own custom chips for its virtual reality headsets, it'll likely license key intellectual properties like CPU and GPU cores, digital signal processors, and more from third parties. Facebook can differentiate by, say, choosing to use a larger GPU configuration than a merchant vendor might use, or it could use a different CPU configuration.
On top of that, Facebook might choose to develop its own dedicated intellectual properties to augment the basic ones that every chip might have. Those intellectual properties would be unique to Facebook, and could prove a competitive advantage.
Facebook's investment in Oculus seems like one it's willing to stick with over the long term. Given that Facebook seems to see significant value in Oculus, and that Facebook is so enormously successful with the vast amounts of development resources it has available, it wouldn't surprise me to see an Oculus headset roll out in three to five years powered by a fully customized, Facebook-designed chip.