Free power doesn’t mean free production.
There’s also labor, wear on the machines and the lost opportunity of using your money to do something else. Building such crushing machines and only use them x% of the time (for, for now, fairly small values of x) may not be a good investment.
So, energy storage should be lucrative and incentivized for us to solve the Duck Curve, and Alligator Curve, and overcharging the grid forces the price to zero or unlucratively below problems.
For firms with depreciating datacenter assets that are underutilized, a Research Coin like Grid Coin or an @home distributed computation project can offset costs
Datacenters scrap old compute - that there's hardly electronics recycling for - rather than keep it online due to relative cost in FLOPS/kWhr and the cost of conditioned space.
Doesn't electronics recycling recover the silica?
Can we make CPUs out of graphene made out of recycled plastic?
Can we make superconductive carbon computers that waste less electricity as heat?
Hopefully, Proof of Work miners that aren't operating on grants have an incentive to maximize energy efficiency with graphene ASICs and FPGAs and now TPUs