Do any of these links actually show the exact method that datacenters are "socializing the costs" for energy production? They tend to pay industrial rates like any other industrial consumer would, and pay for their local interconnect to the grid. Why they should be on the hook for long distance transmission/etc. is unclear to me given that every user on the grid takes part of that and has enjoyed under-replacement electrical rates for decades. I have seen the typical local tax abatement deals which I totally agree need to be outright banned - but that's really the only thing that's jumped out at me. Otherwise it's largely opinion pieces parroting the same few background sources.

I'm not sure it's really a correct mental model to put 50+ years of lack of investment maintaining and expanding transmission and grid capacity onto the industry that laid that fact bare. From where I stand and the underlying studies/reports I've personally read the root cause always seems to be we're running out of power on a grid built by the Greatest Generation and more or less zero work has been put into expanding anything since. There were plenty of people sounding alarm bells a decade ago before the latest datacenter boom.

Then you get into NIMBY stuff, pie in the sky wishful thinking re: dispatchable vs. intermittent power generation, etc.

In the end if we want wealth, we need industry. Industry needs cheap and abundant power to be competitive - AI datacenter or Aluminum smelter. Energy consumption correlates very strongly with wealth, and we've spent decades in the margins messing around with efficiency gains vs. actually investing in anything substantial. When your power company is willing to give you a credit for a more efficient appliance so they don't have to upgrade the grid to your neighborhood you know things have jumped the shark and we are in malinvestment territory.

I certainly agree with the argument that the datacenters might not end up panning out as profitable investments for society, but I'm at least hopeful that when the dust settles we'll actually have augmented our electric grid and finally started to take that looming problem seriously. We might be left with something useful that lasts another 3 generations when all said and done.

We were going to get here either way with population growth and older power generation facilities not being replaced faster than they are reaching end of life. Datacenters simply brought it forward maybe a decade or so. Eventually you run out of the previous generation's energy investments.