Google wants to dump Qualcomm, launch smartphone SoC as early as next year

A new report from Axios claims that Google has "made significant progress toward developing its own processor to power future versions of its Pixel smartphone" and that a Google-made SoC could debut in a phone as early as 2021. Google is apparently teaming up with Samsung, which is providing design support and manufacturing for the project, codenamed "Whitechapel."

The report says the Google SoC is an eight-core ARM processor with hardware "optimized for Google's machine learning technology" and the always-on capabilities of the Google Assistant. The chip would be built at Samsung's foundries on the firm's upcoming 5nm process, and, in addition to being aimed at the Pixel, the report says that "subsequent versions" of the chip could be used in Chromebooks.

Google has been building custom smartphone silicon for a while now. It debuted a custom camera SoC—not a main system SoC—in the Pixel 2, called the "Pixel Visual Core," which was built in collaboration with Intel. The Pixel 3 and 4 have had similar photography-focused chips, now called the "Pixel Neural Core." Since the Pixel 3, the phones have had Google's "Titan M" security module, an isolated chip that handles the phone's verified boot and cryptographic key storage. In the Pixel 4, there's also Project Soli, a radar system that was shrunken down to a tiny piece of silicon. You can see how Google building its own system SoC could be a natural step after all this other silicon work. The company has been hiring chip designers from Intel and Qualcomm for some time now.

Google developing its own phone processor would mean dumping the Qualcomm SoCs it usually uses. Of course, you can never truly be rid of Qualcomm: Google would presumably still need to use Qualcomm modems, something that even Apple still needs to do. There are other modem manufacturers out there—Samsung, Huawei, Mediatek—but Qualcomm's combination of patents and strong-arm licensing techniques has effectively locked its competitors out of the US and other markets.

An SoC division would give Google some much-needed flexibility when it comes to hardware. The report doesn't mention anything about smartwatches, but the real problem Google's hardware ecosystem has right now is a lack of smartwatch chips. Qualcomm has an SoC monopoly but has decided to basically ignore the smartwatch market, so for years there have been no modern options for a smartwatch chip. This isn't just about smartwatches, ether. If Google wants to build any kind of new form-factor wearable, it needs to get Qualcomm on board first. Right now, if Qualcomm says no, like it did for smartwatches, the market ceases to exist.

The most successful in-house SoC division is easily Apple's, which regularly produces chips that wipe the floor with Qualcomm's smartphone and smartwatch SoCs. Apple's SoC division lets it produce chips for whatever hardware form factors it wants—the Apple Watch owes its entire existence to Apple's fast and battery efficient "S" SoCs. Airpods run on the "H" SoCs and can pick up whatever features Apple wants. Apple can choose its own destiny instead of assembling puzzle pieces from other vendors.

Similarly, Huawei has had all US hardware stripped away from it via executive order and keeps on trucking thanks to the protection offered by its in-house "HiSilicon" SoC division, which produces smartphone and smartwatch SoCs. Samsung's Exynos SoC division isn't as successful as Apple's and Huawei's, but it lets it build viable smartwatch SoCs for its Galaxy Watch line. The company hasn't been as successful with smartphone SoCs, which are generally worse than Qualcomm's options.

Listing image by Ron Amadeo / Intel