Would a "charge by power not by energy" model satisfy this?

As in, you pay a lump monthly fee for (up to) single phase 200A 250V, but a higher fee for 3-phase 400A, and a lower fee for a 100A service. Or you buy your own miniature fusion reactor (with its maximum output) and the input fuel is sufficiently common.

@Imuli "Charge by power": yes. I think that would be analogous to a 100mb vs 1gb broadband plan today.

Your own fusion reactor: I think that would also count if the running costs were negligible. It would be the same if everyone had enough solar panels on their roof to have as much energy as they wanted.

The existence of proof-of-work cryptocurrency is one of several reasons this is highly implausible.

@AaronLehmann At some point, the cost of GPUs would be the limiting factor, right? If you give me a free 1TW hookup right now, I need billions of dollars worth of GPUs to fully utilize it. (Quick calculation actually puts it over $1T, but cooling will consume a lot of the power, and is a smaller capital expenditure per watt draw.)

You don't need fancy GPUs. You could use the lousiest scrap computing hardware available since your marginal operating cost is zero.

Generating and distributing electricity is very capital-intensive and these costs need to be paid somehow. It's very hard to imagine this changing by 2050, especially with the investments needed to decarbonize electricity generation and scale the grid to support vehicle electrification. "Too cheap to meter" electricity would encourage inefficient usage that would make these problems even harder to solve.

@AaronLehmann I agree with much of this. However, for wind or solar electricity generation, the marginal costs per additional kWh are pretty low; there are numbers like 1-2 ct for wind out there, and these costs are declining further. If there are relatively high flat fees for industrial consumers, perhaps even differentiated by industry, this would still fall under "too cheap to meter".