The signal speed in fiber is some 200k km/s. If you go 10 miles outside of a city, you are adding 0.1ms ping, and you will have dramatically fewer pitch forks.

The ping between New York and LA is ~70msec. Chatting with a LLM is not latency sensitive.

But if it’s a colo it’s unlikely to have many LLMs running there. The big AI datacenters are mostly purpose built.

Is this DC even for LLMs? I've not seen any sign of that.

Considering that the initial build out was only 10MW, it seems unlikely that was the primary usecase.