Your mind would be blown if you knew how costly it is serving sufficient bandwidth for millions of people downloading tens/hundreds of gigabytes of data.
Depending on how people are consuming/trying games, xCloud may be the cheaper solution because for shorter games or where people are trying games for an hour and deciding they don't like them, it could well be less overall bandwidth to stream the content than download the whole game.
Indeed. Few companies intend to lose money, even in a business which is predicated on loss-leading economics, but the game subscription business isn't really an established or proven business, Microsoft are trying to forge a business model out of it - and I admire that. That's ballsy as anything and it is great to see Microsoft try to build a new business rather than enter/re-enter an established one and not succeed commercially. (e.g. Zune/Plays4Sure, Mixer, WindowsMobile etc).
As I've said above, this is not will, desire, intent, effort or investment, sometimes the economics just aren't sustainable - or they are but the economics of the parties you rely on to fill you service change and your business is no longer viable.
What Microsoft paid for Minecraft and Nokia is far from frugal! They are not afraid to invest big on acquisitions and are pretty good at giving projects a fair shake at the stick before canning them. Some of their pet R&D projects are kind of nuts but you never know what technological dividends experiments like Surface (the tablet computer), Kinect and HoloLens will pan out down the line.