Standing in Ohio, [Trump] did something that would have been unthinkable for a populist campaign five years ago: he defended the machines. Not the workers who build them, not the towns that host them, but the data centers themselves โ the concrete-and-copper cathedrals of the artificial intelligence era. And he did it while acknowledging, in the same breath, that not everyone in his own coalition agrees.
That last part is the story. A rally is theater. The dissent is data.
Here is the paradox I want to sit with: the same political movement that promises to unleash American energy dominance is now colliding with a grassroots backlash against the single largest new source of electricity demand in a generation. And the crowd that feels this tension most acutely โ the people who already own power contracts, substations, and land near transmission lines โ is not Wall Street. It is crypto. It is the miners.
In the ashes of Terra, we didn't learn to trust the narrative. We learned to read the load curve. So let me read this one carefully, because the headline โ "Trump defends AI data centers amid bipartisan criticism" โ is doing a lot of quiet work, and most of it points somewhere the crypto press has not yet bothered to look.
Context: why an Ohio rally about data centers is actually a story about electricity, not AI
To understand why this matters, you have to understand what an AI data center actually is. It is not a warehouse full of clever software. It is a thermal problem with a power bill attached. A single hyperscale facility can draw anywhere from fifty to several hundred megawatts โ the upper end of that range is comparable to the consumption of a mid-sized city, and it draws that power continuously, at a flat, unforgiving load profile that grid operators hate because it does not breathe with the weather or the clock.
Between 2024 and 2025, the United States entered its first genuine AI infrastructure supercycle. Microsoft, Google, Amazon, Meta, and the Stargate consortium collectively announced capital expenditure plans in the hundreds of billions of dollars, and every one of those dollars eventually terminates in a physical building that needs land, water, transformers, switchgear, and โ above all โ electrons. The siting logic is brutally simple: go where power is cheap, where the interconnection queue is short, and where the state legislature will hand you a tax abatement.
Ohio checks all three boxes. It sits inside the PJM Interconnection footprint, the largest competitive wholesale electricity market in the world, and it has spent the last several years courting data center operators with property and sales tax exemptions. Virginia, Texas, and Georgia are its only serious rivals for the title of "most aggressively subsidized destination for compute."
Here is where the story turns. PJM has been issuing capacity warnings with increasing frequency. The region's reserve margins have tightened as load growth โ driven substantially by data centers โ has outpaced the retirement schedule of older generation and the (slow) addition of new supply. When supply is inelastic and demand spikes, the cost lands somewhere. In a deregulated market, that "somewhere" is the residential ratepayer.
So the political math writes itself. A governor promises jobs and tax revenue. The jobs arrive โ briefly, during construction โ and then the facility operates with a few dozen highly skilled technicians. The tax revenue arrives, but it is discounted by the abatement. The electricity demand arrives and stays forever. And the household that pays the bill notices.
That is the real content of "bipartisan criticism." It is not a philosophical debate about AI safety. It is a fight about who pays for the wires.
I have spent twenty-nine years watching this industry, and I have learned that the most important political signals are rarely the loud ones. They are the coalitions that form quietly, out of self-interest, and then harden into legislation. When I audited the Bitcoin.com token sale in early 2017, the tell was never the press release โ it was the multisig structure buried three clicks deep in the contract. The same discipline applies here. The tell is not that [Trump] defended data centers. The tell is that he felt he had to.
Core: the crypto connection nobody is connecting
Now let me draw the line that the crypto-native reader needs drawn, because it is the entire reason a platform like the one I run would even carry this story.
Bitcoin miners and AI data centers are not two industries. They are two tenants competing for the same asset: a large, interruptible, contractually cheap block of electricity attached to a piece of industrial land. The miner built the relationship with the utility. The miner built the substation, the switchgear, the cooling loop. The miner paid for the interconnection queue position that now takes four to seven years to obtain from scratch.
That asset is suddenly worth far more rented to an AI tenant than used to hash SHA-256.
Run the numbers the way I would run them. A modern generation rig earns a gross margin that swings violently with the hash price and the network difficulty โ a business with genuine, structural volatility and a recurring need to reinvest in hardware that depreciates like a falling knife. An AI inference or training tenant, by contrast, signs a multi-year contract at a rate per megawatt-hour that can be several times the effective revenue a miner extracts from the same power. The customer is investment-grade. The revenue is contracted. The volatility collapses.
This is why, over the past two years, we have watched a steady migration: publicly listed miners rebranding as "energy and compute" companies, signing hosting deals with AI clouds, and quietly de-emphasizing their hash rate in investor decks. The market has rewarded them for it. The hash rate, once the only metric that mattered, has become an asterisk.
The hidden insight โ and this is the part the rally coverage completely misses โ is that the miner-to-AI pivot is not a technology story. It is an energy arbitrage story, and it is the cleanest illustration of a structural shift I have been tracking for a while: the binding constraint on American AI leadership is migrating from chips to electricity and permitting.
Think about what that means for the entire geopolitical narrative. For three years, the policy conversation around AI and China has been dominated by export controls โ the assumption being that whoever controls the flow of advanced accelerators controls the race. That assumption is now being undercut by physics. You can manufacture every H100 you want; if you cannot connect it to a substation, it is an expensive paperweight. The bottleneck is not the fab. The bottleneck is the interconnection queue, the transformer supply chain (which has lead times measured in years, not months), and the local zoning board.
I watched this same pattern of misplaced confidence during the DeFi summer of 2020, when I ran live workshops dissecting the Uniswap V2 automated market maker. New users were terrified of impermanent loss and obsessed with the flashiest new pool, while the thing that actually determined their outcome โ liquidity depth and the cost of being early or late โ sat in the footnotes. The AI buildout has the same shape. Everyone is watching the chips. The chips are the footnotes. The grid is the pool depth.
And here is the part that should make every crypto reader sit up: the miners who understood this first are now the counterparties that the largest technology companies in the world must negotiate with. The stranded, contracted power that was once a liability in a post-halving world is now a strategic asset. That is not a small reversal. It is the revaluation of an entire class of physical infrastructure, and it is happening under a political spotlight that is about to make it more expensive, not less.
Let me be precise about the mechanism of that expense. Political attention on data center economics does three things. First, it slows permitting โ every project in a contested jurisdiction faces longer hearings, more environmental review, more community negotiation, and therefore a later revenue start date. In a capital-intensive business where the return is back-loaded, a twelve-month delay can erase several points of internal rate of return. Second, it threatens the tax abatements that made the siting math work in the first place; strip the abatement and the effective cost of power rises without a single electron changing price. Third, it invites rate design โ special tariffs for large loads, or the reversal of the cost-allocation rules that let data centers pay wholesale while households pay retail.
All three of those are political variables, not engineering ones. And political variables are, by definition, unstable. That instability is the real risk premium the market has not yet priced.
The contrarian angle: the criticism is the signal, and the crypto press is reading it wrong
Now for the part I think almost everyone is getting backwards.
The reflexive crypto-media read on a story like this is celebratory. "Trump backs AI data centers" gets translated, in the aggregator economy, into "bullish for compute, bullish for miners, bullish for anything with a substation." That read is not wrong on the first-order effect. But it misses the second-order effect, and the second-order effect is the one that lasts.
The genuinely underreported signal is not the endorsement. It is the fact that AI infrastructure is crossing the line from bipartisan national consensus into partisan dispute. For the last several years, "we must lead in AI" was one of the vanishingly few propositions that both parties could agree on. Data centers were a rare technology topic that did not immediately become a culture war. That consensus is now fracturing โ not over AI safety, not over job displacement, but over electricity bills and tax abatements. And when a technology project becomes a partisan issue, its cost of capital rises, its permitting timeline stretches, and its political risk premium becomes permanent rather than cyclical.
Here is where my skeptic's instinct kicks in, and I want to be honest about it, because I have watched this exact pattern before. I have spent years arguing that "liquidity fragmentation" in DeFi is not a real problem โ it is a manufactured narrative that venture funds use to justify launching the next product, because a problem that does not exist is the easiest thing in the world to sell a solution to. Watch for the same playbook here. As AI infrastructure becomes politically contested, expect a wave of "solutions" โ energy-marketplace tokens, decentralized grid protocols, compute-credit schemes โ that claim to solve the social-license problem without touching the actual constraint, which is that a substation takes years to build and a ratepayer votes every two years. If a pitch deck promises to "democratize AI compute" without mentioning the interconnection queue, you are looking at narrative, not engineering.
The deeper contrarian point is about who actually wins. The consensus is that the winners of the AI buildout are the chipmakers and the model labs. I think that is only half right, and increasingly the smaller half. The durable winners of a supercycle defined by power scarcity are the suppliers of the physical bottleneck: transformer manufacturers, switchgear and high-voltage equipment vendors, liquid-cooling and precision-thermal specialists, and the regional utilities whose load growth is now structurally assured. These are the businesses that get paid whether or not the model architecture of the moment turns out to be the right one. They are the picks-and-shovels of an electricity-constrained world, and they are far less exposed to the political backlash than the data center operators themselves, who are the ones signing the abatements and facing the zoning boards.
The miners sit in a strange middle. They are asset-rich and politically exposed. They own the thing everyone needs โ contracted power โ and they are also the most visible, most resented symbol of the very load growth driving the backlash. That is a genuinely two-sided position, and I would caution anyone treating the pivot to AI hosting as a clean win. It is a trade, not a guarantee.
And I would extend the skepticism one step further, into territory that crypto readers will not like. I have argued for years that governance tokens are, functionally, non-dividend stock โ the only path to a return for the holder is that a later buyer pays more, which is a structural resemblance to a Ponzi that no amount of elegant tokenomics dissolves. The AI-infrastructure trade is now sprouting the same fruit: tokenized compute credits, tokenized power contracts, tokenized data-center revenue shares. Before you accept any of them, ask the only question that matters. Where does the cash actually come from, and who is contractually obligated to pay it? If the answer is "the next buyer," you already know what you are holding.
The ethical layer nobody names: this is a justice problem, not a safety problem
I want to spend a moment on the human dimension, because it is where I think the discourse is most impoverished.
When people hear "AI ethics," they reach for model safety โ alignment, bias, the sci-fi scenarios. But the Ohio story is not about any of that. It is about the distributional justice of technological expansion. The benefits of an AI data center โ the capital expenditure, the construction jobs, the prestige โ are concentrated and visible. The costs โ the electricity demand, the water for cooling, the noise, the land use, the tax revenue foregone โ are diffuse and land on people who did not sign the contract and cannot move the substation. That asymmetry, between concentrated benefit and dispersed cost, is the oldest recipe for political backlash in the book.
When Terra collapsed in May 2022, I did not spend my time tallying the dollar losses, because the dollar losses were the least interesting part. I built a peer-support network instead, because I had learned that a market is a group of people, and when you break a group of people, you get panic, and panic is a mechanism, not a mood. The same lens applies here. The backlash against AI data centers is not irrational. It is a predictable human response to a cost that has been externalized onto the community. Treating it as mere NIMBYism, or as an obstacle to be routed around with better lobbying, is exactly the kind of institutional arrogance that turns a manageable friction into a structural crisis.
The mature response โ and this is where the industry's institutional-ethical synthesis has to actually mean something โ is to internalize the externality before the regulator does it for you. Rate structures that compensate the host community. Water stewardship that is verifiable, not aspirational. Abatements that are performance-contingent rather than front-loaded. None of that is charity. It is the price of a social license to operate, and the operators who pay it early will be the ones still building when the ones who didn't are stuck in hearings.
I drafted a governance standard for autonomous trading agents back in 2026 precisely because I believed then what I believe now: the technology will outrun the ethics unless someone deliberately builds the guardrails first, and the market will not do it spontaneously because the market discounts the future. The same logic holds for energy. The grid is not a background condition. It is the product.
Takeaway: what to watch, and why the load curve is the real headline
So where does this leave the reader who is not interested in political theater but is very interested in where capital should be positioned?
The rally is noise. Here is the signal I am watching instead.
First, the PJM capacity auctions. Every clearing price is a data point about how tight the eastern grid really is, and by extension how much political pressure is building on data center loads. Second, state-level tax legislation. Ohio is not unique โ it is a bellwether. If its data center exemptions start to tighten, watch for a cascade across the states that copied its playbook, because tax competition is a race, and races reverse when the voters notice. Third, and most important, the capital expenditure guidance from the hyperscalers. Listen for a new word entering their language. When you start hearing "power constraints" and "permitting timelines" in the same sentence as "capacity guidance," you will know the bottleneck has officially migrated from silicon to copper.
And keep one eye on the miners. They are holding the most interesting asset in this entire saga โ contracted power at a moment when contracted power has become scarce โ and they are also standing in the crossfire of a backlash they did not start and cannot easily escape. That tension is the trade.
In the ashes of Terra, we didn't learn that markets are safe. We learned that the infrastructure underneath the narrative is the only thing that survives it. The chips will be obsolete in three years. The substation will still be there in thirty. When the next cycle arrives and the story changes again, ask yourself which one you actually own.
The load curve does not care about the rally. It only cares about what you can keep powered.