With the ability to scale databases so easily with cloud services, is the art of performance-tuning databases losing value?

I am of the opinion that, the value it is loosing applies to small scale performance and tunning. No brainers and low hanging fruit are already automated by Cloud PaaS. Thins like creation of indexes, execution plan analysis, statistics gathering, orders of execution by selectivity and storage reorganization based on queries. They even recommend you action plans based on your queries (Azure, Oracle Cloud control 12, at least that I know of).

The problem arises when the analysis you need to do, is more complex or has many combinations before you even come to a possible optimization plan or strategy. I am 100% sure that this king of performance tuning will still be human made. At least for a couple of years.

For example, imagine a 450 table join with 300 where clauses that projects 250 columns (some banks and telecoms need queries like this). How in hell can you start analyzing that automatically without increasing your background processing costs (or idle time)? How can you bruteforce all execution plans and provide the best if they are more than you can represent in memory? I think the bigger the size of the query and the data, the less likely is cloud optimizing it efficiently.

If anything, DBAs of today will have a more difficult job focused on really hard to solve challenges instead of regular index creation and query rewriting. Which I personally like.