Code is law, but people are purpose. I have repeated that line in conference halls from Geneva to Denver, usually to explain why a governance vote matters more than a gas optimization. This week it acquired a sharper edge. Over the past seven days, the loudest argument in American technology policy was not about model weights or inference throughput. It was about electricity bills in Ohio โ and about whether a politician standing on a rally stage could lower them by sheer force of endorsement.
That should unsettle anyone who has spent a decade building decentralized infrastructure. We have trained ourselves to think about AI's constraints in terms of silicon: who controls the fabs, who can export the GPUs, who holds the licenses. The Ohio story suggests the constraint has already moved somewhere less glamorous and far harder to route around โ the power grid, the land beneath it, and the permission of the people who live on that land.
I want to be careful here. The reporting on this event is thin. The account I read carried no date, no direct quotation, no named critics, and no project specifics โ a single aggregated dispatch on a crypto outlet about a political rally. I am not going to pretend that gives me ground truth. But the thinness itself is informative, and the industrial logic underneath it is not thin at all. So let me separate the two: what the sparse reporting tells us, and what the energy economics tell us regardless.
Context: Why a crypto desk is covering a power plant
Start with the obvious question. Why is a publication whose readers mostly care about wallets and yield curves writing about AI data centers at a political rally?
The answer is the single most under-reported fact at the intersection of crypto and AI: the physical infrastructure of these two industries is the same. A bitcoin mine and an AI training cluster are both, at the metal level, warehouses full of compute that consume enormous amounts of electricity and care about one thing above all โ the cost per megawatt-hour. When the price of renting out a GPU for inference and training surged past the revenue a miner earns per hashrate, capital did what capital always does. It started walking across the street.
This is not a metaphor. Throughout 2024 and 2025, a growing number of mining operators began converting โ or announced intentions to convert โ their sites into AI data centers. They already owned the land, the substation interconnections, the power purchase agreements, and the cooling. What they lacked was the networking and the customer contracts, and both turned out to be cheaper to acquire than a new grid connection, which can take years.
So when a crypto outlet covers a rally about AI data centers, it is not off-topic. It is reporting on the asset base of a meaningful slice of its readership. The miners are becoming landlords to the AI boom, and the politics of electricity just became their politics too.
That connection is the lens I want to use, because it reframes the whole event. This is not a story about artificial intelligence. It is a story about who pays for the physical layer of the digital economy โ and about the moment a technology stops being a bipartisan national project and becomes a local cost dispute.
Core: The economics of a warehouse that eats a town's power
Let me put numbers on the table, because the debate is usually conducted without them.
A modern hyperscale AI data center draws somewhere between tens and hundreds of megawatts. To make that concrete: a single 100-megawatt facility consumes roughly as much electricity as 70,000 to 80,000 average American homes. When one of these facilities connects to a regional grid, it does not politely sip from surplus capacity. It becomes, almost overnight, one of the largest single loads in the territory.
Now place that on a grid that was not built for it. The Ohio rally sits inside the footprint of PJM, the largest regional transmission organization in the United States, which coordinates the grid across thirteen states. PJM has spent the past two years issuing capacity warnings โ signals that the system is tightening faster than new generation and transmission can be built. Data centers are one of the primary drivers of that tightening. The result is a classic squeeze: demand rises, supply cannot respond quickly, and the clearing price of electricity climbs.
Here is the part that turns a utility spreadsheet into a political fire. Under most regulated rate structures, the cost of upgrading the grid is socialized across all ratepayers, while the benefit accrues disproportionately to the large new load. The data center gets a dedicated substation and a long-term supply contract; the household down the road gets a line item on its bill that went up. This is not corruption โ it is just how cost allocation works when a system designed for gradual growth meets a step-function demand. But it produces a specific and explosive political artifact: a visible, monthly, personal cost, attached to a benefit that residents cannot see.
Capital is concentrated. Cost is distributed. That asymmetry is the engine of every backlash, in energy and in crypto alike.
I have watched this exact structure before, and I do not mean in a textbook. In 2017, when I was auditing early ERC-20 distribution logic for a community-governed wallet project, I found a flaw that quietly tilted token allocations toward whales and away from the retail holders who had done the actual community-building. The mechanism was technical. The consequence was social. A handful of addresses captured a disproportionate share of the upside, while thousands of small participants absorbed the downside โ dilution, weak governance voice, and the slow erosion of trust. I did not just patch the code. I organized three town halls to explain why fair distribution is not a nice-to-have but the load-bearing wall of decentralization. Five hundred people showed up to hear a mathematician talk about game theory, because they intuitively understood that a system which concentrates gains and disperses costs will eventually lose its legitimacy.
A data center raising your electricity bill is the same pattern in physical form. The upside โ construction jobs, some tax revenue, the vague glow of hosting "the future" โ is concentrated in a few hands and a few months. The downside โ higher rates, water consumption for cooling, noise, land use, and the quiet question of whether the tax breaks were worth it โ is spread across everyone who stays. When the arithmetic of benefit and burden runs in opposite directions for two groups of people, no amount of growth statistics will paper over it.
And the employment argument, the one every developer deploys at the zoning hearing, is weaker than it sounds. Yes, construction of a hyperscale facility can create several hundred to a thousand temporary jobs. But the operating phase โ the part that lasts decades โ typically employs only a few dozen to a few hundred highly specialized staff, because these facilities are engineered to run with as few humans as possible. The jobs-per-megawatt ratio collapses the moment the cranes leave. Compare that to a traditional manufacturing plant of comparable capital intensity, and the gap is stark. This is why "we brought jobs" so often curdles into "we brought a bill."
None of this is unique to Ohio, and that is precisely the point. Ohio is simply an early and visible case of a national structural shift. The state has aggressively courted data center investment, and it has simultaneously become a venue for debates over data center tax exemptions โ sales tax and property tax relief designed to attract exactly the projects now under scrutiny. When you hand out incentives to win a facility and then watch residential rates climb, you have built a machine for producing political criticism from both directions at once. Fiscal conservatives ask why the tax base is being given away. Ratepayers ask why they are subsidizing the bill. And the two complaints, though they come from opposite ends of the spectrum, land on the same target.
That convergence โ criticism from both flanks โ is the signal worth reading. It marks the moment AI infrastructure stops being a bipartisan national asset and becomes a partisan cost center.
Contrarian: The bottleneck moved, and almost nobody repriced it
Here is where I want to push against the comfortable narrative, because the comfortable narrative is that AI's constraint is chips.
It is not, or at least it is no longer only that. The binding constraint on American AI expansion has been migrating from silicon to social license โ the informal permission a community grants to a project that affects it. Export controls on advanced chips matter enormously, and I do not dismiss them. But an export control is a problem you can engineer around: you redesign the chip, you find a workaround, you substitute. A community that does not want your substation is a problem you cannot compile your way out of. You either persuade it, compensate it, or you build somewhere else โ and "somewhere else" is a location decision with a multi-year clock attached.
This is the insight I think the market is underpricing. For three years, the AI investment story has been told as a chip story: the fabs, the GPUs, the memory, the advanced packaging. The supply chain that benefited most was the semiconductor supply chain. But if the true constraint is the grid and the permit, then the beneficiaries re-rank. The transformer manufacturer, the switchgear supplier, the liquid-cooling vendor, and the regional utility with a growing load base start to matter more than the GPU reseller. Capital expenditure flows to whatever is scarce, and if what is scarce is a megawatt of firm capacity, then that is where the returns will pool.
I have a bias here that I should name. I have spent years arguing that most decentralized finance interest rate models are essentially arbitrary โ elegant curves with parameters plucked from the air, only loosely tethered to any real market for liquidity. Energy markets have the opposite problem and the opposite virtue. The price of a megawatt-hour is not arbitrary; it is discovered, auctioned, and enforced by physics. When the AI industry collides with that market, it is colliding with the one pricing mechanism in the stack that cannot be fudged by governance vote. You cannot propose a temperature parameter to lower the cost of electricity. Physics is the ultimate oracle, and it does not respond to governance proposals.
There is a second contrarian angle, and it is the one that should interest anyone who follows crypto. The backlash is being reported, even by sympathetic outlets, as an "AI threat" story โ as if the residents of Ohio are afraid of machine learning. That framing is almost always wrong, and it is a media artifact, not a grassroots one. The people showing up to oppose a data center are rarely motivated by existential dread of neural networks. They are motivated by water. By noise. By the fairness of a tax break granted to a trillion-dollar company. By the fact that their bill went up while the promised prosperity stayed on the other side of the fence. The abstraction "AI" is a headline; the lived reality is a utility bill and a cooling tower.
This matters because it changes who can fix it. If the problem were public fear of AI, the solution would be communication โ explain the technology, win hearts and minds. If the problem is cost allocation and local consent, the solution is structural: negotiate genuinely, share the upside, price the externality honestly. The first is a PR campaign. The second is governance. And governance, as anyone who has ever tried to run a DAO knows, is the hard part.
Which brings me to a discomfort I cannot fully resolve. Most decentralized autonomous organizations have the legal status of a shrug โ no real corporate form, no clear liability shield, and when something goes wrong, members can find themselves exposed in ways no one signed up for. I have watched communities discover this the hard way. An energy project has the opposite problem: it is over-legalized, wrapped in tariffs, rate cases, and utility commissions. Yet both are wrestling with the same underlying question โ how does a group of people who share a physical resource decide who bears the cost and who takes the benefit? The DAO answers it with a token vote that is often captured. The utility answers it with a rate case that is often opaque. Neither has solved it.
The Ohio story is, at bottom, a governance failure dressed as an energy story. And governance failures do not get fixed by better semiconductors.
Takeaway: Social license is the new consensus mechanism
So where does this leave us, and what should we actually watch?
I think the honest answer is that we are watching the emergence of a new kind of consensus mechanism โ one that runs not on hashpower or stake, but on the consent of the people who live near the load. Call it social license. It is slow, it is messy, it cannot be purchased in a spot market, and it can veto a project that every spreadsheet says should be built. For an industry that has spent a decade optimizing for speed, this is an uncomfortable arrival. Trust, verify โ but also, connect. The verification of a data center's economics is trivial. The connection to the community that has to host it is the part we keep skipping.
For those of us who came up through crypto, there is a constructive role here that has nothing to do with speculation. The mining operators who own the interconnections are sitting on an asset that is suddenly valuable precisely because of its scarcity. But the way they deploy that asset will determine whether they are welcomed or resisted. A miner who converts a site into an AI data center and negotiates a real revenue share, a real water plan, and a real local benefit agreement is doing something the hyperscalers have largely failed to do. That is a competitive advantage, and it is one that can only be earned, not bought. Resilience beats hype every time, and in this arena resilience is measured in community trust as much as in uptime.
The forward-looking judgment I will leave you with is this. For three years the AI trade was a silicon trade. I think the next phase is a permission trade. The companies that win will not be the ones with the fastest GPUs but the ones who can get a substation approved in a county that is tired of being asked to subsidize the future. Watch the state-level data center tax legislation, because a tightening wave there would reprice projects overnight. Watch the regional capacity auctions, because they are the closest thing this industry has to a truth serum. And watch the miners, because they hold the physical keys to a door the AI boom now needs to walk through.
Community is the new central bank โ and it does not print the megawatts for free.