I'm not in any way in favour, but your reasoning is incorrect. Efficiency is about doing work with an amount of energy, before it becomes heat. Not about how much heat that can be dumped.
On the energy supply side of the equation, it's very much limited by the area of solar panels. And conveniently, the larger those collectors are the larger the area of shade is available for heat dumping as well.
As for speed of light considerations. All of the proposals will be for LEO. Starlink being the poster child here.
I can't see how doing data-centres like this could ever be economical but that's the thinking.
Being a smokescreen of talk is highly likely. Same as building new fission nukes.