Is it power efficient to have 1 (small|medium|large) instance per workload? Or more realistically >=2 instances for redundancy. No? Maybe we could get cost, en

Is it power efficient to have 1 (small|medium|large) instance per workload? Or more realistically >=2 instances for redundancy. No? Maybe we could get cost, energy and other efficiencies by aggregating disparate smaller and larger workloads over a system that scales up and down with demand, so every instance is processing X million TPS. Cool idea. We could label that scalable, because it’s vastly more efficient!