It usually does, yes, as long as the operating system can emulate ARM instructions from the x86 software. Microsoft Windows and Apple macOS are both capable of this.

There are even solutions for Linux ARM distros but work is ongoing there.

Expensive machines with CISC processors will run slow because whoever spent that money did not account for cooling. With some very notable recent exceptions in Intel’s latest generation of chips that I must concede you’re very much right about, the heat spreaders on virtually all CPUs are more than capable of removing thermal energy from the chips themselves at a rate sufficient to have stable performance at maximum load for extended periods of time.

Between the heat spreader and the outside environment is another story. Thermal paste, cooling solution, and its correct installation all play a huge role in the thermal performance of all processors.

Instruction set of a processor does not increase its speed to any significant degree, at least to the casual observer. There’s a slight edge in favor of CISC because of in-memory operations but that just makes it use fewer clock cycles to get certain things done. It’s not very noticeable unless performing very heavy tasks that rely on constantly modifying data stored in RAM.

The server world is a very complex one. You have everything from edge servers to live-service broadcast, to streaming, to encoding, to huge AI clusters. In most use cases, data centers have one of two priorities: Performance vs. Economy.

RISC processors bring economy in droves without sacrificing any significant performance, so it’s likely that more data centers are going to be using these chips as cost-saving measures as well as reducing their carbon footprint.