Bankruptcy or Reinvention? AI’s Energy Crossroads
Tripling power demand means higher bills, strained water, and political backlash. But it could also create a smarter, cleaner grid.
I’m just going to assume we’ve all lived the moment where our reckless tomfoolery resulted in knocking that thing your mom cherished off the mantle. And any of us with a healthy dose of fear transformed at warp speed into automatic crisis prevention mode. Well, I got some news: U.S. data center power use is projected to triple by 2030. That’s not growth, or disruption, that’s a three-alarm shattering of your grandma’s lamp from roughhousing.
Tripling demand doesn’t just mean more turbines. It means rewiring transmission lines, draining water tables, and rewriting the social contract of electricity. If nothing changes, households will see higher bills, grids will strain, and emissions goals will be shredded.
It’s the kind of projection that should spark a national conversation about how to build the digital infrastructure of the future. Instead, we get headlines about ribbon cuttings and “economic development,” with mayors beaming as they welcome a new hyperscale campus. The glow fades as fast as your parents taking a look at your credit card spending at college.
We’ve seen this bit before, the company-town problem. Coal towns. Steel towns. Logging towns. The Company (with a capital C) towns, and like a cattle drive through wet clay, they leave the ground so torn up that no one wants to walk it anymore.
It always seems like everyone wants to move west to find their gold, and this is no different. Except this time, it's not gold or a timer; it’s cheap power and scarce water for a large room of computers bought on credit. A $10 billion box that hums and spits out maybe a few dozen permanent jobs, while drinking the resource and economic wells of small towns dry with a long straw, just as Danial Plainview (There Will Be Blood) so eloquently demonstrated (in what was one amazing cinematic experience).
Just like the lure of new mines or factories, these facilities arrive with headlines full of promises. But once the contracts are signed, the leverage shifts. Tech giants lock in sweetheart utility deals for decades. Local regulators are left holding the bag when costs inevitably rise. And residents, whose bills creep higher to subsidize power transmission and data center cooling, have the least say in the process.
We don’t have to accept the enviable—the formula by which many in rural areas are being worked over. However, this system is fragile, and local leaders risk significant political backlash. Because nothing turns neighbors into activists faster than skyrocketing utility bills they can’t explain.
It doesn’t have to be this way. There’s a version of this story that can end on a high note. The same facilities that risk bankrupting the aquifers and town councils could, with a different framework, become centers of economic prosperity in a smarter, more resilient system.
Thinking of this in two categories, the first being policy and the second being related to responsible engineering. Starting with policy, we can take a note from initiatives like California’s standards of efficiency and required use of energy consumption, paired with mandates that require data centers to leverage green-energy onsite, making them not only self-contained, but contributors to the larger grid during off-peak hours (SF Chronicle) (FT).
Moving on to engineering objectives, we can use AI to optimize power grid usage and cooling (read: water consumption) demands. For example, accepting we don’t need every answer right now, it would be easy to shift workloads to off-peak times, thus smoothing out the CPU curve (Goldman Sachs)..
I guess every engineering decision IS a policy decision and vice versa, so while we’re at it, requiring new facilities to invest in new energy storage technologies would also be a good dividend. We’re already seeing promise with new underground thermal storage technology (NREL).
This isn’t sci-fi. It’s a design choice. Europe already requires certain efficiency disclosures. China is building massive renewable data center hybrids. And U.S. utilities are starting to pilot AI systems that anticipate spikes and balance load automatically.
The difference is whether regulators have the backbone to make these the default instead of the exception.
We’re not facing a technical challenge, but rather a policy one. By this point, the big issue with the system isn’t the “how do we do it”, but rather the “how do we go about it.” We already know how to smooth out demand curves to maximize energy costs and store excess energy. We know how to recycle waste heat into district heating and how to site facilities near renewables, rather than straining fossil-heavy grids.
What’s missing is political will. For decades, the U.S. allowed utilities to play the role of middleman, cutting long-term deals with whoever promised the largest upfront investment. That worked when the worst outcome was a stranded natural gas plant. It won’t work when trillion-dollar tech companies can bulldoze their way into preferential rates and leave everyone else holding the bag when the inevitable happens.
This isn’t about preventing growth. It’s about matching scale with responsibility. If AI is going to soak up terawatts of power, then the companies behind it should be building the storage, renewables, and resiliency to match.
There’s also a cultural dimension. Tech leaders love to cast themselves as visionaries. They boast about the potential for curing cancer with machine learning, automating the drudgery of tedious tasks, and developing general intelligence. But it all rings hollow if the foundation is brittle grids and hollowed-out communities.
The architects of the future must prove their worth by creating value beyond quarterly reports. Data centers can make rural communities more resilient if we have the gumption to negotiate with the same brass knuckle tactics that the tech oligarchy uses. Providing locals an equity stake in the resources being consumed, publishing energy-use numbers in plain English, and making efficiency a core performance metric should be the standard, not the exception.
Because here’s the reality: the backlash is coming. If the AI boom raises household bills by $40 a month or drains local aquifers to keep servers cool, the pretty glow of innovation will dim quickly. And it won’t matter how many Nobel-level breakthroughs the algorithms deliver if ordinary people see AI as the reason their lights flicker and their water bills spike.
So where does this leave us? We can treat AI as a new form of digital colonialism—extracting resources, exporting profits, and leaving behind liabilities. Or we can demand a model of shared equity, shared responsibility, and shared benefit. Stripping local communities of their local resources so a college kid can ask a robot to generate their English 101 essay is a human crime on so many levels.
I hope we can choose a path where we end up with something surprisingly hopeful. Rural towns could become leaders of clean energy innovation. Data centers that double as community batteries, feeding green power back into the grid when and where it’s needed most. Imagine an AI future built not on extraction but on resilience.
The first wave of AI data centers looks kinda like company towns. The next wave doesn’t have to.
It’s not a technical question. It’s a political one. Do legislators, regulators, and zoning committees have the will to force the masters of the future to build responsibly? Or will we muck this up?
The future of AI’s energy demand isn’t written yet, but the bill will be, and the only question is who signs it and who will bear the expense.
This is gonna require more than an elaborate story involving mental gymnastics; this will require us figuring this out before Mom gets home.
This is Part 4 of a 4 part series, if you read it from the start, thank you, and I’d love to know what you think. If you’re just now finding this, check it out from the start.