Ethiopia has just reduced the electricity promised to its Bitcoin miners by 77%. The facilities are now receiving only 23% of their contracted power after water inflows into hydroelectric reservoirs fell by 20%. The paradox is striking: these same miners generated 35% of Ethiopian Electric Power’s revenue in the last financial year and brought in several hundred million dollars in foreign currency. Addis Ababa is nevertheless choosing to preserve electricity for households and factories. Africa’s Bitcoin mining boom is meeting a very physical constraint here: without water, even the cheapest megawatts disappear.
Miners fall to 23% of their contracted capacity
The reduction did not happen all at once. Ethiopian Electric Power, or EEP, initially limited data mining operators to 75% of their contracted power, before reducing that figure to 50%. With no improvement in reservoir levels, the state-owned operator ultimately cut the allocation to 23%.
The episode is a brutal blow to an industry built around a very simple requirement: running thousands of ASICs almost nonstop. A machine that is switched off no longer consumes electricity, but it also stops producing even a single satoshi. A site that had secured 100 MW is therefore receiving only 23 MW under current conditions.
The measure directly affects 39 companies with power purchase agreements with EEP, 31 of which are already operational. These contracts normally provided for at least 98% availability. Today, they offer only a fraction of that capacity.
This dependence on energy was already at the heart of the sector’s economics. Bref Crypto recently showed that the rise of AI is also pushing Bitcoin miners toward the cheapest sources of electricity. In Ethiopia, they had found them. The problem is that cheap does not always mean available.
Bitcoin was nevertheless generating enormous revenue for EEP
The decision appears even more surprising when looking at Ethiopian Electric Power’s accounts.
In the last financial year, Bitcoin miners provided around 35% of the state-owned company’s revenue. EEP also says it collected $180.19 million from data mining operations during the first six months of the 2025–2026 financial year, with foreign-currency revenue from technology companies rising by 138%. The operator posted a profit of 7.1 billion birr over the same period.
These revenues explain why Addis Ababa had opened the door so widely to international miners. Ethiopia has massive hydroelectric potential and was seeking customers capable of buying large amounts of electricity, including when local demand could not absorb it all.
Mining fits that profile precisely. A farm can be set up near a generation source, consume several dozen megawatts and pay for its electricity in foreign currency without requiring the industrial infrastructure of a steel mill or automobile factory.
The relationship appeared almost ideal: EEP sold megawatts, operators received competitively priced energy, and Bitcoin transformed that electricity into a globally exportable asset.
Then the hydrology changed.
With just 23% of the promised power, the same model becomes much less comfortable.
The real problem lies in the dams
Ethiopia has invested heavily in hydropower. Ethiopian Electric Power still estimated the country’s installed capacity at around 9.7 GW this year, with roughly 96% coming from hydroelectricity. The Grand Ethiopian Renaissance Dam, with its 5,150 MW, now dominates the national power system.
This concentration has a major advantage: the electricity produced is largely renewable and can be highly competitive once the dams have been built. It also creates a direct dependence on water.
El Niño has worsened the current dry season. According to figures provided by EEP’s CEO, inflows into the reservoirs have fallen by around 20%. A hydroelectric plant may have turbines capable of producing several gigawatts; if enough water no longer reaches the dam, that theoretical capacity does not become available electricity.
The Ethiopian case therefore highlights something that installed-capacity statistics can easily conceal. Having 9,700 MW of power plants does not mean having 9,700 MW available at every moment.
Water thus becomes the real raw material of Ethiopian Bitcoin.
And when its flow declines, EEP has to choose where to send each remaining megawatt.
Households and factories come before Bitcoin
Addis Ababa’s choice is fairly clear. Ashebir Balcha, EEP’s CEO, says the restrictions are intended to protect supplies for households and manufacturing businesses.
This priority takes on particular significance in a country where access to electricity remains incomplete. The National Energy Compact prepared by Ethiopia with the World Bank estimates that only 44% of the population benefits from an electricity service level considered basic or higher. The country aims to raise that figure to 75% by 2030.
It becomes politically difficult to justify household power cuts while keeping huge warehouses filled with ASICs running at full capacity.
Industry raises another question. Ethiopia wants to develop its manufacturing activities, create jobs and process more raw materials locally. These factories also require a reliable power supply.
A Bitcoin miner can shut down several thousand machines within minutes and restart them when the grid regains capacity. A production line responds much less well to that treatment.
This flexibility therefore becomes almost a commercial weakness: miners are easy to switch off.
In times of abundance, they buy the surplus. In times of scarcity, they absorb much of the adjustment.
Ethiopia had built a mining paradise
The massive arrival of miners was not accidental. For several years, Ethiopia offered some of the world’s most competitive electricity rates for data mining activities, thanks in particular to its abundant hydropower.
The sector reached considerable scale. International companies, particularly those arriving from China after Beijing tightened its crackdown on mining, installed machines in the country. Phoenix Group, for example, had increased its local capacity to 132 MW in 2025. Several estimates at the time put Ethiopia’s share of Bitcoin’s global hashrate at around 2.5%.
EEP had already begun changing the rules before the drought. Since December 2025, miners have no longer benefited from an extremely low flat tariff. The system now takes consumption times and grid availability more fully into account.
This change was already consistent with a trend tracked by Bref Crypto: mining companies now have to look far beyond the BTC price alone. In our analysis of miners missing out on Bitcoin’s rally, electricity, infrastructure costs and diversification were already emerging as variables as important as the asset’s price.
Ethiopia is now adding another variable: the amount of water that fell months earlier.
Cutting miners also costs Addis Ababa dollars
Giving priority to households and factories does not make the decision cost-free for the state.
Data mining activities had become one of Ethiopian Electric Power’s important sources of foreign currency. When a miner goes from 100 MW to 23 MW, EEP mechanically sells much less electricity to that customer.
The drought is also hitting regional electricity exports. EEP has cut its export-revenue forecast for the current financial year by 40%, to $279 million. If water levels remain insufficient, the company is also considering further reductions in electricity sales to neighboring countries.
The authorities are therefore facing several competing uses for the same hydroelectric output: households, domestic industry, Bitcoin miners and regional exports.
Bitcoin has one advantage: it pays and brings in foreign currency.
It also has a drawback: it creates fewer direct jobs per megawatt than a large factory, and its machines can be switched off quickly without damaging the grid.
Ethiopia’s trade-off is therefore not really “Bitcoin versus the population.” It is the allocation of a resource that has temporarily become scarce among several customers that have neither the same economic importance nor the same ability to accept interruptions.
For now, the ASICs are losing that competition.
Miners can survive a cut of this magnitude
A 77% reduction is extremely severe for the operators affected. It nevertheless does not have the same effect as a regulatory ban.
The machines still exist. The contracts still exist. The sector is not being expelled from the country. EEP is due to review the situation in October, when reservoir trends will provide a clearer picture of the amount of electricity available.
Epic Mining was already confirming to its Ethiopian customers in July that the sector’s allocation had been reduced to around 23% because of hydrological conditions. The company adapted the operation of its facilities without announcing their permanent closure.
This ability to interrupt consumption is one of the distinctive characteristics of Bitcoin mining. An ASIC produces nothing when it is switched off, but its task is not lost. Once electricity returns, it immediately starts calculating hashes again.
At the global network level, the decline in Ethiopia’s hashrate does not prevent Bitcoin from continuing to operate. Mining difficulty is automatically adjusted approximately every two weeks to maintain a block time of around ten minutes.
Miners elsewhere in the world then recover a slightly larger share of the competition.
The network adapts far more easily than the financial statements of Ethiopian companies.
AI further complicates the battle for megawatts
Ethiopia’s experience comes at a time when mining is already facing another energy battle.
Data centers dedicated to artificial intelligence are willing to pay heavily for reliable, immediately available power connections. In the United States, several miners are beginning to use their sites to host GPUs rather than ASICs.
Bref Crypto recently detailed the case of Riot and its $9.1 billion AI contract with Anthropic. The financial logic is clear: when the same megawatt generates more revenue through high-performance computing than through Bitcoin’s SHA-256, shareholders look at the numbers before ideology.
Ethiopia is not yet at that stage on a large scale. Its current problem is more basic: producing enough electricity.
But the question will probably arise eventually. The best hydroelectric sites may attract miners, conventional data centers, AI, industry and electricity exporters at the same time.
Bitcoin has an advantage in this competition: it tolerates interruptions very well. A data center running an AI model prefers an extremely stable power supply.
Ethiopia has just turned that flexibility into energy policy.
The miner becomes the customer that can be switched off when the dam lacks water.
Bitcoin also exposes the limits of the hydropower model
The episode could easily be portrayed as a problem specific to mining. Above all, it reveals a much broader weakness in Ethiopia’s power system.
The country produces very clean energy, but remains extremely reliant on hydropower. Its own energy plan now calls for the share of non-hydroelectric renewables to rise from 5.6% to at least 15% by 2030. Total installed capacity is simultaneously expected to increase from around 9,761 MW to 14,000 MW.
Solar, wind, geothermal power and storage are therefore becoming more important, not because hydropower is being abandoned, but because a system too dependent on rainfall remains exposed to dry seasons and climate change.
For miners, this diversification could be good news in the long term. Solar power sometimes produces more electricity than the grid can immediately absorb. Bitcoin is particularly well suited to this kind of intermittent surplus.
The same farm that is being rationed today could tomorrow be used to absorb surplus power for a few hours each day.
The model would nevertheless change fundamentally. The era when a miner could simply sign up for tens of megawatts of hydropower and expect availability close to 100% appears less certain.
Ethiopian energy remains cheap.
It has just shown that it is not guaranteed.
October will show whether the cut is temporary
Ethiopian Electric Power is due to reassess miners’ allocations in October. Everything will depend primarily on how water inflows into the dams evolve.
Improvement would allow power to be gradually restored to the farms. Continued drought could have the opposite effect, bringing further cuts to mining and electricity exports.
Ethiopia’s Bitcoin sector is therefore in an unusual position. It is neither banned nor rejected by the government. On the contrary, its foreign-currency revenues have become significant enough to appear in the financial results of the national operator.
But 35% of revenue does not confer 35% of the rights to the grid.
When electricity becomes scarce, the country chooses its households and factories first.
That may be the episode’s most interesting lesson. Bitcoin mining has often been presented as a last-resort buyer capable of monetizing energy that would otherwise be wasted. Ethiopia had turned that idea into a large-scale economic activity.
The drought shows the other side of the model: Bitcoin works particularly well with surplus electricity precisely because it can be sacrificed when that electricity is no longer surplus.
For EEP, miners therefore remain valuable customers.
Simply not priority customers.
