The question you must ask yourself is why are miners renting their gear for other people to use? If it was profitable at the current price, they could simply mine themselves and skip the rental part.
Profit is hardly guaranteed considering capital costs. Disregarding that, I'm not sure shovels are the best comparison. Coin mining is pretty predictable in aggregate, unlike gold mining. If you own mining hardware, you can calculate expected income pretty reliably.
A couple years ago, there were scripts that would watch Github commits for AWS (or other) credentials. If it found them, it would fire up instances and mine as much as possible until they were shut down.
Cloud mining is only profitable if you have a lot of equipment allocated to under mined coins. If a coin has a low enough difficulty you can overflow with hashing power and earn a profit depending on the difficulty and how often the coin re-targets. A coin like Aptcoin or other non-asic mineable coins can create this situation.
But mostly no. It is almost exclusively more profitable to just buy the coins.
No and it never well be. Lets consider AWS. On-demand pricing, even reserve is never even going to be close, simply because there exists Amazons overhead, profit, etc. built into the cost. The only possible chance is spot instances. Spot instances start to approach a "free market". The pricing will adjust based on supply and demand. There still exists AWS overhead, as they aren't going to run them at a loss, just far less profit. If the price of say the p3 family of instances approached a profitable price point, then others would flock to using them, driving up the price, causing them to no longer be profitable. There also exist external use cases, such as AI, which can justify the computation power at a higher price point, as the value of the computation might exceed the value gained from mining.
