A customer can bring billions into a power company and still be the first one asked to stop buying.
Bitcoin miners in Ethiopia are discovering exactly how that works.
Bitcoin miners arrived with machines, capital, and an appetite for electricity. They became a major source of revenue for Ethiopian Electric Power, the state utility known as EEP. Then the water feeding the country’s hydroelectric system became scarcer, and the relationship changed.
Bloomberg reported that miners accounted for approximately 35% of EEP’s revenue in the previous financial year. Their electricity allocation has now fallen to around 23% of contracted supply, following successive reductions as conditions deteriorated.
That is a spectacular reversal for an industry that sells itself as an ideal electricity customer.
It also exposes something more consequential than the fate of one mining hub.
For years, miners have promoted their ability to shut down when the grid needs relief. The machines can stop hashing and resume later. Electricity consumption falls. Bitcoin continues operating elsewhere.
The engineering makes sense.
The financial consequences deserve much more attention.
If your greatest value to the grid is that you can be switched off, your business has to survive the grid exercising that option.
Ethiopia is putting that proposition under pressure. The outcome matters wherever miners hope to turn abundant electricity into a durable business.
Most industrial development requires an enormous amount of coordination. Factories need transport, suppliers, workers, customers, financing, and dependable infrastructure. Electricity is one part of a much larger system.
Bitcoin mining has an unusual advantage: its customers do not need to live nearby.
A mining facility can participate in the same global network from many different locations. With comparable hardware and operating conditions, its hashes compete on the same terms. Bitcoin does not pay an extra reward for proximity to a financial center.
That makes mining a potential buyer where generation is available before local demand or transmission has fully caught up.
For a utility, the attraction is straightforward. It already has electricity to sell. A miner can bring substantial demand to a suitable connection point without waiting for an entire manufacturing ecosystem to develop.
Ethiopia’s experience demonstrates how substantial that demand can become. Ethiopian Business Review reported that data mining generated 50.37 billion birr for EEP in the last fiscal year, exceeding every other customer category.
Revenue on that scale makes mining a serious commercial relationship. The utility receives payment for power; the miner uses that power to compete for Bitcoin rewards.
But a large bill does not tell us whether every unit of electricity was otherwise unusable, whether the tariff covered every relevant cost, or whether the arrangement produced the greatest possible benefit for the country.
Those questions require separate answers. They become unavoidable when electricity gets scarce.
Hydropower can look reassuringly solid. The concrete is enormous. The turbines have a rated capacity. The investment horizon stretches across decades.
Water availability still determines what the system can deliver.
EEP attributed the restrictions to weaker reservoir inflows, reported at roughly 20% below the relevant comparison level, amid dry conditions associated with El Niño. Bloomberg described the utility reducing miners’ allocations first to 75%, then 50%, and eventually 23%, while prioritizing households and manufacturing.
The distinction between installed capacity and available electricity is easy to overlook in a bullish investment presentation.
A nameplate capacity tells you something about the equipment. It does not promise that enough water, fuel, transmission capacity, or operating headroom will be available whenever a customer wants power.
For a hydro system, decisions also extend beyond the current hour. Water released through a turbine today may have been available for generation later, subject to reservoir limits and other water-management obligations.
That makes the idea of “surplus electricity” more complicated than it first appears.
Power can look abundant during one period while the system still needs to conserve resources for another. A utility planning for a dry season must weigh tomorrow’s needs alongside today’s sales.
A miner’s spreadsheet can assume continuous production. The reservoir has no obligation to cooperate.
USDC made Circle a giant of digital dollars—and left its revenue tied to interest rates. Arc is its bid to capture the payments, markets, and machine commerce built around them.
The reported figure concerns the share of contracted electricity being supplied. It should not automatically be read as every machine operating for exactly 23% of every day.
A reduction can be implemented through a smaller continuous allocation, scheduled shutdowns, different treatment across sites, or some combination. A facility might keep its most efficient machines running while leaving the rest idle.
Those details affect the financial damage.
What the figure establishes is a severe constraint on the electricity available to the sector relative to its contracts. It does not establish an identical reduction in revenue, profitability, or Bitcoin output at every company.
Operators have different machines, different debt burdens, and different contractual arrangements. Some can absorb a prolonged interruption. Others may have much less room.
Imagine having paid for the equipment, the connection, the buildings, and the cooling system on the expectation of substantially more power than you now receive.
You can reduce electricity consumption immediately. You cannot reverse the investment just as quickly.
That is where the physical flexibility of mining meets the financial rigidity of a mining company.
Switching off a miner saves electricity. It may also reduce some cooling expenses and wear.
It does not automatically suspend rent, security costs, loan payments, essential staffing, or the capital already committed to the facility.
The equipment’s economic life keeps moving, too. A machine sitting idle may face more efficient competitors by the time it returns to work. Future mining conditions may offer less revenue for the same computing power.
The operator has lost an opportunity to earn during a period it can never recover.
That makes the duration of an interruption crucial.
A short shutdown during a valuable grid event can be a manageable operating decision. A prolonged restriction can undermine the assumptions used to finance the entire site.
An operator that budgeted for occasional curtailment has not necessarily budgeted for a season with most of its expected electricity unavailable.
The distinction also changes which machines make sense to buy. Highly efficient new equipment can make better use of scarce power, but expensive hardware still needs enough productive time to justify its purchase. Cheaper older machines may carry less capital at risk, yet consume more electricity for each unit of computing work.
There is no universal answer. The right fleet depends on the actual pattern of power availability, the purchase price, and expected mining revenue.
The ability to stop a machine tells you very little about how long its owner can afford to wait.
Consider a deliberately simplified example.
Two hypothetical sites each have 10 megawatts of mining capacity, identical equipment efficiency, and $1 million in annual fixed costs and capital recovery. They buy only the electricity they use, receive no curtailment payments, and convert each kilowatt-hour into the same amount of computing work.
One pays 3 cents per kilowatt-hour but receives only 23% of the energy needed to run at full capacity throughout the year. The other pays 5 cents but receives 98%.
These are illustrative assumptions, not estimates of Ethiopian operators’ accounts or a claim about their measured annual uptime.
The site with the cheaper tariff ends up with the higher cost per unit of productive electricity because it spreads the same annual fixed burden across much less output.
This is not a complete profitability model. It excludes other variable expenses, changing Bitcoin rewards, equipment differences, and many real operating decisions. It isolates one mechanism: low utilization can overwhelm a power-price advantage.
Cheap electricity remains valuable. Its value depends on how much profitable work the business can actually perform with it.
A tariff is a price for an input. It is not a complete measure of competitiveness.
“Flexible load” can describe arrangements with radically different economics.
In one arrangement, a miner agrees in advance to reduce consumption during specified events. The compensation may come through a discounted tariff, explicit demand-response payments, or contractual rights to realize the value of unused power.
The interruption is part of the product being sold.
In another arrangement, electricity becomes unavailable for an uncertain period because the system cannot support the expected demand. The miner loses production, and whatever relief exists depends on the agreement’s terms.
Both arrangements reduce load. Their consequences for the operator can be worlds apart.
A useful historical example comes from Riot Platforms. For August 2023, Riot reported approximately $31.7 million in combined power and demand-response credits. Its disclosure distinguished credits associated with its power contracts from participation in ERCOT demand-response programs.
The important feature is the commercial structure. That example shows how a miner can have an economic reason to reduce consumption under particular market conditions.
It does not establish that another grid can offer equivalent compensation, or that every shutdown creates the same value.
Ethiopia’s restrictions demonstrate that mining demand can be cut substantially. Demonstrating a sustainable business around that flexibility requires more: workable prices, manageable interruptions, and a clear allocation of risk.
The shutdown is only half the story. The contract determines who pays for it.
The contractual reporting deserves careful reading.
Ethiopian Business Review quoted EEP’s chief executive referring to an agreement to deliver 98% under the power-purchase arrangement. Capital’s September 20 report, however, attributed to EEP officials the position that climate-related interruptions do not trigger mandatory supply guarantees or penalty payments.
Those descriptions leave an important question: how does the expected supply commitment interact with the contract’s exceptions?
The published accounts do not provide a basis for deciding every operator’s legal rights. They do show why the headline availability figure is insufficient for understanding the commercial risk.
A company needs to know what happens under the conditions most likely to threaten its project.
Who can order a shutdown?
For how long?
Under which exceptions?
What compensation, notice, or termination rights exist?
For an investment tied to a hydro system, the drought provisions belong near the center of the analysis.
A promise of abundant power during ordinary conditions is useful. The business may ultimately depend on the clauses governing extraordinary ones.
It would be a mistake to imagine the utility can remove a major customer’s electricity without consequences of its own.
When consumption falls, sales can fall. If operators eventually move their machines, restoring that demand may require new investment, new negotiations, and renewed confidence.
The relationship therefore contains a difficult trade-off. Preserving water and prioritizing essential uses can be necessary today while still carrying a financial cost tomorrow.
A utility also needs to think about who bears that cost. Lower revenue can affect the resources available for maintenance and expansion unless it is offset elsewhere. That is a possible consequence to examine, rather than evidence that a particular investment program has already been cut.
My reading of Ethiopia’s experience is that a large mining customer base can create two forms of dependence at once. Miners depend on the utility for electricity. The utility can become exposed to the fortunes and continued presence of the mining industry.
A sound arrangement has to survive stress on both sides.
One of the strongest arguments for mining is its ability to buy electricity with few immediate alternative uses.
But that description belongs to a particular location and period. It is not a permanent property of the electricity.
Demand can grow. Transmission can improve. A new industrial customer can arrive. An export connection can become available. A dry season can make stored water more valuable.
Any of those changes can alter the opportunity cost of serving the mine.
This creates a development scenario that deserves more attention. A country builds generation before it has enough connected demand to use it fully. Miners buy some of that early output. Later, households and businesses gain access, and the electricity available to mining shrinks.
Under the right commercial terms, that could be a successful lifecycle for the power system.
The miner would have provided useful demand during an earlier phase, while accepting that the grid’s needs would evolve.
But the arrangement has to be financed accordingly. A temporary opportunity can be valuable. Financing it as if the original conditions will last indefinitely creates a very different risk.
This is why national development and mining growth cannot always be measured with the same yardstick. A successful electrification strategy might eventually reduce the surplus available to miners.
That possibility should be part of the agreement from the beginning.
Electricity generation and electricity access are related, but they are not interchangeable.
A household needs a connection, local distribution infrastructure, affordable service, and reliable delivery. Available generation somewhere on the system does not establish that the household can use it.
That is why a large mining load can coexist with unmet electricity needs. Geography, network bottlenecks, and investment requirements matter.
It is also why shutting down a mine does not automatically connect a new village. Reducing demand may help the existing grid operate within its limits, while new connections still require physical work and funding.
The strongest case for a mining arrangement would show how it supports those needs: what the customer pays, what costs it imposes, where the revenue goes, and what happens when supply tightens.
Revenue is evidence that someone bought a product. The wider development benefit requires tracing what that transaction made possible.
The same discipline applies to environmental claims. Hydroelectric power has a different emissions profile from fossil generation, but the relevant assessment still includes the actual power source, water-management constraints, infrastructure impacts, and competing uses.
A serious argument for Bitcoin mining can handle those questions. It becomes more credible by answering them.
For Bitcoin itself, the scale of the disruption matters.
Hashrate Index’s Q3 2026 heatmap estimated Ethiopia at approximately 2.4% of global hashrate, or around 23 exahashes per second. That is an estimate of geographic distribution, not a real-time measurement of the impact of these restrictions.
It provides useful perspective: severe local curtailment can be painful for operators without amounting to a comparable re…
