The thing about AWS that is so impressive is not what it does. What AWS does was previously done by every single IT department at every company across the globe. What is impressive is the scale at which they do it. When you move to a world of every product hosting their own AWS equivalent, you've removed the challenge and impressiveness of AWS since you no longer need their massive scale, you only need enough to host your own product.
Setting up your own servers was not a bottleneck to the generation of internet companies that preceded AWS, and time has only made hardware cheaper, bandwidth more affordable, and OSS higher quality and more plentiful. It is easier than ever to host your own services, AWS just makes it cheaper up front thanks to their massive scale.
Point being, people aren't choosing AWS because AWS solves a problem they're incapable of solving, people are choosing AWS for pricing flexibility and faster time to market.
Setting up your own servers was absolutely the bottleneck. Providing on demand scalable computing and storage untethered a generation of garage startups from venture funding. Prior to 2007, if you had a great idea you could build it and then start renting, or building and colocating, servers to run that app for you, at considerable up front expense. Plus, you got to do all of your own system administration, up to and including driving down to the colo when you couldn't adequately diagnose issues remotely.
So when you say "what AWS does was previously done by every single IT department at every company across the globe" yes sure but (1) it was done worse in many cases, with data loss or significant downtime, and (2) there were fewer such companies by magnitudes, because of the cost barrier.
What AWS did was take a costly process, done inconsistently and to varying degrees of correctness across the business world, and make it available to everyone at a very high quality, with a innovative pricing schedule.
There are a dozen different contractors and companies that can build out electricity supply for you in a relatively sensible timeframe. Most of the major components are off-the-shelf.
S3 has eleven 9s of durability. That alone is bonkers. Sure, let's say you replicate that and staff it up to keep it working ongoing.
I didn't say "replicate S3", I said "replicate a significant portion of AWS". A lot of the value of AWS doesn't come from using _a_ product, there are alternatives for most of what they offer, it comes from having the whole ecosystem of tools integrated and available in one place.
So now you need to go build out a highly available, redundant queuing service like SQS that supports FIFO delivery and up to 20k inflight messages.
And a highly available, redundant notification service with integrations not only with the web/email/etc but also SMS.
And a geographically redundant database service with multi-master, instant snapshots, point-in-time restore, etc, etc.
And... well, pick whatever other handful of AWS services you're using in your specific use-case.
And wrap it all behind tools for management. And hire a whole wackload of ops staff to keep it all going. And pay 10x as much because you don't get the economies of scale that AWS does.
I'm not ignorant of the difficulties in setting up generation and storage for electricity. But I'm also not ignorant of the absolutely massive task it would be to replicate AWS if you're it for more than a really expensive VPS hosting service. I would 100% choose to work on off-grid electricity generation before rebuilding AWS.
I am sorry but this read like a report from fantasyland. In the past 3 years I have seen several AWS failures but not a single powercut.
You entire take is based on the idea that AWS replacement has to be superbly reliable and scaleable, while your grid replacement does not.
Lets compare like for like, your grid replacement has to have redundancy so that generators can be repaired without power loss, it has to support megawatt scale spikes in demand in case several friends with electric cars come to visit, it has to have mean time between failure measured in years, withstand extreme weather, and be renewable. Also you need to have black start capacity and logistics to replace equipment promptly when it fails.
How is the cost for that going to compare to using the grid?
In the end AWS S3 is just distributed file storage.