Think of a data center situation though where you can get a ~50% power requirement difference for marginal performance decreases. The cost savings would be huge even if you require more servers to make up for the performance difference.
For most home users I think you're right for now but for businesses it could be huge.
Of course, you've got to remember that which chip has the best performance per watt depends on the job. A sufficiently smart data center could determine this automatically and take that into account in its job scheduling.
For most home users I think you're right for now but for businesses it could be huge.