(1) Please don't
over-quote. It wastes space and causes confusion if I edit something.
(2) I used the term "infinitely-scalable" in a poetic human-scale sense, not the objective mathematical one - and it's a separate argument whether concepts like "infinity" are ever really applicable in the real world.
(3) Yes, as basic math will tell you, the amount of time and/or computing resources needed to crack a 128-bit key by brute force is (by our present-day standards) astronomical - billions of top-notch CPU's for billions of years. Unfortunately future technology can make what's astronomically complex today child's play in the future, and some entities may have access to future tech breakthroughs sooner than others. The commercial cloud computers would be very expensive for that sort of thing - but if NSA really does have its own underground semiconductor manufacturing and secret atomic-powered supercomputer mega-labs, all bets are off.
(4) Cracking estimates will depend not just on the size of the key, but the encryption algorithms used, whether there are any shortcuts to using brute force, and the computational complexity of recognizing a success case (i.e. how certain are you that you have a decrypted file and not gibberish). There's always an unknown risk factor that the governments, which recruit many of the world's top mathematicians, know a shortcut to cracking a particular algorithm but are keeping it secret.
(5) If you can recognize the success case then you can crack anything - it's just a matter of time and money. Today it might not be worth-while to spend the CPU power to decrypt something, but after many years of exponential growth in (post)quantum computing - who knows. (And don't even get me started on them
stellar-scale computing.) So don't trust any kind of encryption for things you don't want your great-grandkids to know about.