note the scalable in there. Even under best case assumptions, it's only a good fit for edge coprocessing where latency is super important. You're not going to power the world's compute with a self healing Dyson sphere (but if you think that, I've got something to sell you).
The guy has to rent out his terrestrial data centers because he can't make use of them.
We probably don't need to pollute space with dead GPU's.