Wind-Powered Mining Pays Only When It Stops Being About Curtailment

A new study from the Technological University of the Shannon has produced a headline that Bitcoin's energy debate will repeat for years. Pairing mining with an Irish wind farm can lift its revenue by 32 percent while absorbing most of the power the grid would otherwise waste. The facts are solid. The reading beneath them is where the interesting work sits.

Two claims travel together in that headline, and they are not the same claim. One is that mining monetizes energy that would otherwise be curtailed, turning waste into value. The other is that mining is a profitable flexible load for a wind farm. The study's own scenarios show exactly where these two part ways, and the gap between them is the whole story.

Why the cleanest model starves

Dispatch-down is wind the grid cannot take, so it is switched off. It arrives in gusts and gaps, not as a steady supply. A mining rig that runs only on that curtailed power therefore sits idle most of the year. And a rig is almost entirely fixed cost. The machines are bought once, and they earn only while they run. Utilization, not the price of electricity, is what decides whether they pay for themselves.

That is why the curtailment-only configuration, S1 in the study, fails. A 20 megawatt rig fed only curtailed wind runs at about 16 percent utilization. A 90 megawatt rig runs at under 5 percent. The authors call it not financially viable, and mechanically the reason is plain. The cleanest possible model, the one that touches nothing but energy that would otherwise be wasted, cannot keep expensive hardware busy enough to earn back its cost.

The paradox at the center of it

Here is the tension the study exposes without quite naming it. The property that makes mining useful to a grid is that it will shut off the instant the power is worth more elsewhere. That willingness to stop is exactly what a system with too much intermittent wind needs from a flexible load.

But that same willingness to stop is what empties the hardware. A miner that yields whenever the grid or the market wants the electricity is, by definition, a miner that runs part-time. For a machine whose cost is almost all up front, part-time is a problem. So the grid-friendly behaviour and the hardware economics pull in opposite directions. The better a rig behaves for the system, the worse it uses the capital sunk into it.

This is not a quirk of mining. Any device built to soak up curtailment meets the same wall, as the study itself points out. Size a battery, or any absorber, beyond the typical length of a curtailment gap and it too sits empty between gaps. What is specific to mining is the way out, which the next sections trace.

What makes it pay is no longer curtailment

To make new hardware run enough to pay, you feed it something other than curtailed power. That is precisely what scenarios S2 through S5 do. S2 mines saleable wind whenever mining beats selling it. The higher scenarios add grid electricity. Utilization jumps to between 31 and 76 percent, and revenue climbs.

But notice what changed. The extra energy is no longer waste. It is wind that had a buyer, or power drawn from the grid. The additional revenue is arbitrage. In a given hour, mining out-earns the market, so the operator mines instead of selling. That is a legitimate business. It is also a different activity from rescuing curtailed power, and the study is honest that the two blur together as flexibility rises. The closer a project sits to pure curtailment mining, the worse it pays. The better it pays, the less it is about curtailment at all.

The hardware the study did not test

The study resolves its own paradox in one direction only. It assumes new, efficient machines with full investment costs, roughly 34.7 million Euro for a 20 megawatt build. Against a bill that size, low utilization is fatal, which is why the curtailment-only model collapses.

Change the hardware assumption and the logic flips. If the machines are cheap or already paid for, there is almost no capital to amortize, so running part-time on free power stops being a problem. That is the population the study set aside. It excluded the older Antminer S9 on efficiency grounds and modeled only new commercial rigs, so it never tested the one setup where curtailment-only mining might actually make sense, which is an operator running depreciated or second-hand machines on power that would otherwise be thrown away.

This is not hypothetical. When MARA bought a 114 megawatt wind project in Hansford County, Texas, it did so to run a behind-the-meter operation on what it called stranded, zero marginal cost renewable energy, and it staffed the site through an "Advanced ASIC Retirement Initiative" that redeploys retired machines beyond their previous economic lives. That is the study's unviable S1, made viable by deleting the capital cost. The tradeoff is real and worth stating plainly. Old rigs are far less efficient, so at any positive power price they lose. They only work on genuinely free or curtailed power, and often only where waste heat or another byproduct adds value. But within that narrow window, the cheap and interruptible machine is the one that fits the grid's needs, precisely because nobody minds when it sits idle.

So "S1 is not viable" is better read as "S1 is not viable as a new commercial ASIC build." For existing miners, small operators, and individuals with old hardware and access to wasted power, the cleanest model is also the one whose economics the study did not measure.

The price regime it cannot escape

Even the profitable end is conditional, and here the study is careful in a way worth preserving. Its headline caveat is subtle. Viability turns less on the Bitcoin price by itself than on the spread between how fast the price grows and how fast the global network hashrate grows. Mining revenue per megawatt-hour is, in effect, the price divided by the hashrate. When both climb together, that revenue barely moves, and the returns come from soaking up curtailment rather than from Bitcoin going up. The danger is not a bear market as such. It is hashrate outrunning price, which quietly thins revenue per machine even if the price holds.

The absolute level still sets a floor, and the study's static payback table is blunt about where it sits. A 20 megawatt build pays back inside six years at 100,000 Euro per coin across every curtailment level, at 80,000 Euro only on the most constrained sites, and at 60,000 Euro nowhere. The model's base price was near 90,000 Euro. In late August 2026 Bitcoin traded near 79,000 Dollar, roughly 73,000 Euro, down from an all-time high of about 126,000 Dollar the previous October, while hashrate has pushed past the model's 780 exahash assumption to around 900. Read against the study's own table, a newly built rig on a moderately constrained farm would struggle to pay back at today's price under static conditions.

None of this is a forecast, and we make none. The point is narrower and harder to argue with. A result calibrated to a single favorable year describes that year. The same rig lives through a full cycle, and the 32 percent uplift is not a fixed property of the pairing. It is a number that swings with where Bitcoin sits relative to the network's own growth, and the snapshot flatters it.

The tariff question hiding in S3 to S5

There is a second assumption worth pulling on. The scenarios that import grid power charge it a fixed network fee, set in the study at 31 Euro per megawatt-hour to reflect a demand-side tariff that itself rose 46 percent across 2024 and 2025. Whether flexible mining is a system benefit or a hidden cost depends heavily on how that fee is designed. A separate 2026 study in the same journal, evaluating grid tariff structures, reaches a conclusion that travels well beyond its subject. Existing tariffs give weak incentives for flexible behaviour, and no single design satisfies fairness, efficiency and simplicity at once. Every design trades one against another.

Applied here, a miner drawing grid power in scenario S3 or S5 might be helping the system, by soaking up local surplus, or gaming it, by pulling cheap power while shifting network costs onto other users. The difference is not in the mining. It is in the tariff. The wind study is alert to this and proposes guardrails, a cap on how much of a site's output can feed self-consumption so a wind farm cannot quietly convert into a mining plant, and metering duties for sites claiming preferential curtailment treatment. It even notes that Ireland's December 2025 rules for large data centres, which oblige them to bring their own generation, are the mirror image of a miner placed at a constrained wind node. The question is not whether flexible computing loads and the grid interact. It is on what terms, and those terms are still being written.

The bigger picture: a bridge, not a fix

Step back from the single farm and a larger case comes into view, one the study gestures at when it says mining could make grid investment easier to finance. Ireland is building generation faster than it is building wires. That gap is what produces curtailment in the first place. Flexible, interruptible load parked behind the constraint turns some of that wasted energy into revenue, and revenue is what keeps projects bankable while the transmission catches up.

There is a longer arc here too. Cheap, stranded power tends to attract demand over time. Industry follows energy, and mining is simply the most mobile and most interruptible form of that demand, able to arrive first and leave quietly when firmer, higher-value load shows up. Read that way, mining is not the destination. It is an anchor tenant for stranded energy, a way to stop wasting power today so that the grid, and the industry that a stronger grid invites, can arrive tomorrow.

The case only holds under one condition, and it is the same condition as before. The load has to genuinely yield when the system needs the power. A miner that runs flat out and refuses to stop is not grid support. It is just another large consumer. The interruptibility is the whole benefit, which loops back to the hardware point. The rigs that can afford to be interruptible are the cheap ones.

The strongest version of the optimistic case

Before drawing our own line, the positive case deserves its full weight, because the study makes it well and honestly. This is not theoretical. The authors point to real precedent, from 32 Texas wind and solar projects that earned 47 million Dollar from pre-commercial mining, to two gigawatts of registered mining load in Texas that delivered fast demand response during Winter Storm Elliott, earning 25 million Dollar while sparing the system an estimated 100 million Dollar, to Brazil, where more than 32 terawatt-hours of curtailed renewable energy since 2021 has begun to draw co-located mining in with no mandate at all.

The study is also candid that its own figures are cautious in places. It leaves out grid-service revenue, worth perhaps 0.85 million Euro a year to a 20 megawatt load, and it leaves out the price floor that Ireland's renewable support scheme would add on top. On those counts the returns it reports are understated rather than inflated. The honest counterweight, which the authors flag themselves, is that the model assumes perfect foresight of prices, so in that one respect the figures lean high. A fair reading carries both corrections at once. This is a serious, well-hedged case, not a sales pitch, and our caution below is a refinement of it rather than a rebuttal.

How we read it

Curtailment-soaking Bitcoin mining is a real and useful tool. It is flexible load that can absorb wasted renewable energy and put a floor under stranded power. But it is neither a free lunch nor automatically green, and the study is most valuable for showing why.

Its economics are an ASIC-utilization problem first and a Bitcoin-cycle bet second. The configuration that looks cleanest on a slide, mining nothing but curtailed wind, is the one the study calls unviable for new hardware, and the configurations that pay do so by trading power rather than only rescuing it. The resolution to that paradox is not to build bigger. It is to build cheaper, because the machine that can sit idle without bleeding money is the one that actually fits a grid with too much wind and not enough wire. The study lands close to this when it advises deploying early at the most constrained sites, before the free energy pulls in competing rigs. We would only add that the operator who can do that most cheaply is the one holding hardware already paid for. The honest sentence is not "mining makes wind farms profitable." It is "under the right hardware, the right size, and the right point in the Bitcoin cycle, interruptible mining can monetize energy a wind farm would otherwise lose, and whether that reads as an environmental win or an arbitrage business depends on where the operator sets the dial."

What would change our mind

This reading rests on two claims, and each has a clear failure point.

The first is that cheap, interruptible hardware is the real fit. That weakens if new, efficient rigs on curtailment-heavy sites turned a durable profit through a full Bitcoin cycle, not just in a favorable year. If the expensive build pays across a bear market, the case for old hardware loses its force.

The second is that the grid benefit depends on genuine interruptibility. That weakens if flexible mining at scale were shown to raise system costs even when it yields on demand, for instance by distorting network charges enough that other users pay for it. The grid-tariff literature suggests that outcome is possible, not certain. Both are worth watching, and both would make good follow-ups.

The study is careful to say mining cannot replace the grid buildout Ireland needs. That modesty is the right note. Flexible mining is a patch on one specific problem, wasted wind behind a constrained grid, and a patch with real value. Just not the value the headline implies, not for every operator, and not for free. It also sits inside Bitcoin's older environmental debate, where the same rule has always held: location, power mix, and behaviour decide almost everything.