Why don't Amazon and Google utilise the collective mobile device CPU grid/pool?

both Amazon and Google are big into client-facing cloud computing with AWS and Google Cloud Platform respectively. And they both offer "big data" compute resources, Amazon EMR and Compute Engine respectively. Which means they have thousands (millions?) of servers grinding away, processing data as I type. Which costs a lot of money.

Both also are big into mobile, with Android and the Amazon Fire Phone (OK, OK, so Google's big into mobile and Amazon is trying, bless 'em). And both have very popular online stores, Google Play and of course amazon.com.

So, why don't either of them offer a "goodies for cycles" deal:

• when your device is charging and on Wifi and has low CPU utilisation, it receives and completes small jobs - packets of self-contained work
• it grinds through said jobs until no longer charging or not on Wifi. At which point it "parks" its work until it's charging and on Wifi again.
• each job completed earns the device owner credit on Google Play / amazon.com

Just think of all the unused computing power out there sitting on the kitchen counter or nightstand overnight doing nothing... And it's only going to go up as the number of cores and RAM increases.

I don't think it'd be that hard to set up either since there are open source Grid computing solutions readily available such as BOINC upon which API's for Android and Fire OS (basically Android) could be built.

So, assuming all of this is vaguely logical - why haven't they done it yet? I guess one limitation is that this form of distributed computing is only effective for processing parallelizable workloads and not all problems fall into that category. What else?