Google's already lost China - the Chinese brands sell Android phones with all the Google bits stripped out and replaced with Baidu, WeChat, etc. to the billion plus customers within their borders.
I agree that there isn't much of an alternative to Android. Google learned from street corner drug dealers - give it away free to get them hooked! Now they're stuck with Android, and if Google starts selling 'better' Android phones and steals the high end of the Android market away from Samsung, there's probably nothing they can do about it. Instead of wasting time cloning Google's services with stuff no one uses like S Voice and Samsung Pay the last five years, Samsung should have been working on making Bada/Tizen a credible alternative.
If Google packs unique features in its SoCs it could drive Android handset makers elsewhere to differentiate--but where? Win 10? Tizen? Not many good alternatives, and Android has consumer mindshare.
I think Google launched Pixel in part because OEMs were not being as aggressive as Apple and Samsung in pushing the envelope. Their Nexus phones tried to nudge things forward but provided only a little momentum.
I suppose the big risk is China--will Huawei, Oppo, Vivo and the China handset-OEM-of-the-day go to some clone China mobile OS? Will Google be able to break into the China handset market? These are perhpas Google's biggest risks with a unique Android SoC.
Well, recently there have been good results with image and video compression using deep learning. Assuming it's saves just 25% of bandwidth per video(but possible more, maybe 2x improvement), with wide deployment , the savings from youtube would be huge. And forget savings, this may be used strategically, making Google even stronger in TV.
But it's gotta low power. So there's been some really impressive work about analog neural networks, talking about 5 orders of magnitude improvement in power/cost. That seems something widely deployable even on low-end phones, and really useful for an AI company.
If Google makes SoCs with capabilities not available to other Android OEMs, to go along with software capabilities they've already not made available to other Android OEMs, that's possibly a dangerous path.
Maybe they feel Android is so ubiquitous that OEMs like Samsung have no choice but to keep making Android phones. However, if anything would focus competitors on trying to come up with some sort of alternative, providing themselves a competitive advantage over other Android phones would be it.
At Google IO the company said it is making image recognition available on handsets. It would make sense to accelerate inference tasks on handsets and Google has the technology in its TPUs today.
Done right, Google could make a case it would have a better end-to-end experience on its smartphones for machine learning jobs like image/face recognition and AR...and it could lauch new accelerated services more quickly and easily without the need to cooridnate with third parties.
They'd need to hire a whole CPU design team if they were going to design their own ARM CPU core like Apple, Qualcomm and Samsung. So they're just going to take generic ARM cores and integrate them into an SoC, like Apple did with the A4 and A5 for the iPhone 4 and 4S.
The question is, do they want to design something to sell to Android OEMs or something they will keep to themselves - like they designed Pixel with a few features Android OEMs couldn't get.
It is also possible this wouldn't even be targeted at phones, but rather Chromebooks, or maybe something smaller like a watch or a second try at Glass.
An OS made plenty of sense (i still remeber the day they announced it and how excited everybody was) but today they don't quite need to control the hardware since they have the OS and the app store.
SoCs for glasses, server, cars make sense but not for phones.
Maybe if the angle is connectivity. The main problem with phones are the carriers.. Only half of the global population has a smartphone and those that do can barely use them due to the absurd data prices. If Google has a solution, great but bringing it to market with their own SoC might be less than ideal. Yeah, getting rid of crarriers would be a large business opportunity but something like that is not easily achieved,.
Now I come to think of it, is it specifically for phone/tablet or mobile in general?
"Google is ramping up a team to build mobile SoCs for tablets and smartphones. The Web giant posted nearly 200 job openings in its hardware group including at least half a dozen specifically for mobile SoC designers."
-> The job openings could as well be for IoT SOCs in general ranging from Google Home to developping an open source IoT platform including hardware/software/networking services/cloud services/analytic services...
Again, they could just give it for free too the community as one unified package for the world of IoT leaving manufacturing, distribution and sales to vendors. Ensuring very strong integration of all componends would ensure IoT cloud service (revenue) domination and worldwide control of the *very* rich vast amount of data that IoT generates by having plenty of devices at most public and private locations. By giving the entire stack away for free it would kill off competition before it even started as developping such whole stack including security is a complex and very expensive nightmare.
So many years ago, I thought the same about Google making a phone OS. I did think:
A) Aren't they far too late with it? They'll never get a decent market share.
B) Are they just doiing this as a nice side project? They'll never be able to make a penny on developping an open sourced mobile operating system.
. Not only did they become the dominant player on the market, they also added "features" that enabled them to hiddenly collect plenty of additional information from their user base. Think about mapping all wifi devices while developping Google maps for example. It gave them the ability to pinpoint the location of Android devices almost in realtime with great accuracy even if the user turned of their GPS. That feature, and the collecting of this information by Google only recently became clear to a growing group of users. Think of their agressive push for users to give the Gmail app access to body sensors (you can't type five words in the app without beiing nagged about it to turn it on if you have turned it off; why does a mail app absolutely need acces to my phones body sensors???). Both "features" work well for Googles pockets.
So I would not be surprised to see Google add a few features to their phone/tablet/Chromebook SOCs. Like for instance:
A) A unique device identifyer that *can't* be blocked from leaking to Google by using no cookies, an ad-blocker, script blocker,...
B) An ability to make it harder to install "non-authorised" software on devices like malware, adware but also ad-blockers, tools monitoring your own network trafic...
Will they have sufficient market share to make profit on it? The answer may be they don't need to sell the hardware themselves. They could use the same strategy as they used with Google analytics , Google tag manager... etc. There they gave code that was convenient to use for free to developpers. The result? Today it is hard to find a website that doens't link to both giving Google a treasure trove of information about your detailed surfing behaviour, for free. More and more websites break if you block Google tag manager so it becomes compulsory.
Potential profit model? Develop the hardware with "added" features locked firmly in it. Make it perform well. Give it away for free. It'll end up in the vast majority of phones. When/if giving it away for free would be violating the Arm license deal, then develop RiskV cores. Optimise Android to run well (better then on Arm?) on RiskV. Problem solved, market dominated, profit trough more free information and ability to control basic functionallity of the device. Too far fetched? Maybe, but what is happening today surpassed my imagination five years ago more than this one.