No one denied there aren't benefits but... some of this is overblown:
1) Why would it remove hardware research from companies? You think if you run a streaming gaming company you aren't involved in research?
2) You think all streaming systems would be exactly the same, removing the concept of platforms? So you envision a monopoly on backends or something?
3) And how can you claim there are no lifecyles then also claim there is more power available yearly? You think somehow we will update the entirety of a streaming platform every year (not going to happen, there would be staggered updates with multiple levels of power/functionality) but that totally removes the idea that a dev has to target new features?
4) No BC issues? What do you even mean? What if a streaming platform wants to make a major change? Why is this any different than hardware refreshes really, where we already are seeing companies like MS plan for easy BC in the future?
Yes there are benefits; conceptually it's cool for a consumer to have it all in the cloud, no game installs, all kinds of other cool stuff. We've been talking about that for almost 10 years now since OnLive demo'd it. But it's not like we magically get away from publishers having to target multiple hardware sets, they will choose a mix of AMD and nVidia for one; and that's all getting easier anyways with Sony / MS and Nintendo all having far less customized hardware and focusing on making things easy for devs.
Are you serious with this?
Mainstream won't care if when playing a singleplayer game it takes a quarter of a second for a server to get the fact they pushed the "fire button" and another quarter second for the action on screen to represent that? That's sort of ludicrous; essentially with lag for streaming games.. "it doesn't work."
You call everyone else short sighted then post the cost of a single server? A server that you admit isn't even designed for gaming? And then let's pretend the only cost is the CPU/GPUs serving up for hosts? What about networking, the power structure, the human beings it takes to manage these servers?
And you actually think a petaflop for $150k (which is probably way low for what is available today) is actually a good price? That's enough for 200 end users if you give them 5 teraflops. Something like PSN is likely approaching 20 million concurrent users at it's peak (judging by how it's larger than Steam, and Steam maxes out a bit below that.)
And you would need far more than the ability to host 20 million at once because you need to handle the 24 hour cycle worldwide and locate these servers all over the world to handle the peak traffic in THAT part of the world. Let's super low ball it and say you could get away with 8 DCs (no possible way, I'm being generous), and that the peak in each DC is 5 million (again no possible way, low balling) that means you'd need the power to host 40 million concurrent users. At next-gen levels of 5TF (being generous again) that's 200 million teraflops.
That's 200,000 petaflops. So 200,000 of your $150k server number you threw out there.
Roughly $30 billion.
Just for the CPUs / GPUS; let alone networking, paying for bandwidth, the buildings themselves, the power draw, the employees, etc.
And that's going to get upgraded yearly?