Bill Gates infamously said sometime in the 80's "640K is more memory than anyone will ever need" , and for the scope of computer applications at the time he was correct.
The counter argument is that this is over-, premature, or unnecessary optimization. If we can put 2GB of memory in a computer for $20, why spend millions on R&D that will be rewritten to make it work on 128 MB?
There is also an environmental argument for your point, but everything will always be a tradeoff to companies trying to build and sell devices quickly.
Why would we expect value to scale linearly with memory and transistor count? I would in fact expect it to be logarithmic: each doubling would lead to a roughly linear increase in cool stuff we can do.
I think that's what we see. Yes, our computers could be radically better experiences if it weren't for the bad habit of squandering improvements with layers of unnecessary abstraction and laziness. But the only thing standing in our way is those unnecessary and that laziness, so maybe we should stop doing that, now that real gains in speed are slowing down.