Post #1839819

Is there any actual use case for orbital data centers? They are no good for training (need loads of power (=heat) and network latency would cripple training times). Inference - might as well build small distributed DCs on the ground. And the cost is 3-5 times (minimum) what it would cost to build on the ground, leaving aside the pollution of all those 10s of thousands of launches. Why are people even considering this? #AI #datacenters