This is a new term to me. I think it's bit like PUE in that you want the number to be as close to 1 as possible (i.e. for 1 server's worth of utilisation, you want to have provision 1 server, not 5)

So, I think this is the term you might use to talk about how a hyperscaler datacentre might be full of individually efficient servers, but only have 5% of them serving workloads - the rest is just spare capacity ready to sell, or give away as credits