Former Binance CEO, Changpeng Zhao (CZ), recently stated that the UAE generates surplus power in order to cover “three days” of high demand each year, making Bitcoin a buyer of last resort for energy that would otherwise go unused.
Stripping away the specifics, the logic holds: mining turns curtailed or stranded electricity into revenue when no other offtaker wants it.
The question for 2026 isn’t whether surplus can be mined, but whether that surplus is structural enough to contract, and whether miners can hold their position as AI and high-performance computing push up the clearing price for firm supply.
The economics are straightforward. Electricity accounts for more than 80% of miners’ cash operating expenses, according to Cambridge’s Digital Mining Industry Report.
The same report cites a median electricity-only cost of around $45 per megawatt-hour and notes that surveyed miners curtailed 888 gigawatt-hours of load in 2023, roughly 101 megawatts of average withheld capacity.
That curtailment figure supports the flexible-load thesis: miners can switch off when grids need relief or when prices spike, making them useful to utilities managing intermittency or congestion.
Geography tells the rest of the story. While imperfect in methodology, the Cambridge Bitcoin Electricity Consumption Index Mining Map tracks where hashrate concentrates, though the data carries caveats, such as estimates lagging by one to three months, and VPN or proxy routing can inflate shares in countries like Germany and Ireland.
Country attribution relies on geolocating IP addresses, a method that is sensitive to routing behavior and subject to other inference limitations.
Within those constraints, the map shows mining distributed across jurisdictions with one thing in common: access to power that’s either cheap, stranded, or both.
Pakistan turns overcapacity into policy
Pakistan made the most explicit bet. The government announced plans to allocate 2,000 megawatts in the first phase of a national initiative split between Bitcoin mining and AI data centers, with CZ named strategic adviser to the Pakistan Crypto Council.
The Finance Ministry framed it as a way to monetize surplus generation in regions with excess energy, turning underutilized capacity into a tradable asset.
Two thousand megawatts running continuously would generate 17.52 terawatt-hours annually. With modern mining fleets operating at 15 to 25 joules per terahash, that power could theoretically support 80 to 133 exahashes per second of hashrate before accounting for curtailment, power usage effectiveness, or downtime.
The scale matters less than the structure.
What type of contracts will miners sign, interruptible or firm baseload? Which regions get selected, and how durable is the policy if tariffs rise or IMF pressure intensifies?
Pakistan’s initiative signals that “extra electrons” can become a national export, but execution will determine whether 2,000 megawatts materializes as a hub or just a headline.
Surplus by design, not accident
The UAE’s opportunity isn’t perpetual surplus, but it’s surplus-by-design.
Peak demand in Dubai reached 10.76 gigawatts in 2024, up 3.4% year-over-year, concentrated in summer months when cooling dominates load.
The International Energy Agency (IEA) projects that cooling and desalination will account for close to 40% of electricity demand growth in the Middle East and North Africa through 2035, with data centers explicitly named as another rising load source.
That creates a specific opening for miners: utilities build systems to handle high summer peaks but need year-round monetization, normalization, and grid stability during off-peak periods.
Miners win where they can offer more flexibility than AI or HPC buyers, such as curtailment-ready loads that absorb power others can’t take because of location, congestion, or dispatch constraints.
Bitcoin miners can switch off in an instant, whereas datacenters require continuous operation, making curtailment and grid management much more difficult.
The region’s buildout trends favor baseload capacity that outpaces seasonal demand, but the same IEA outlook that flags data centers as a driver of demand means miners face direct competition for the electrons they need.
The hub case depends on whether utilities value dispatchable load enough to price it attractively, or whether firm offtake contracts with AI buyers crowd out mining altogether.
When surplus becomes contested
Paraguay illustrates what happens when surplus power attracts miners, only to trigger a backlash.
The country’s hydro capacity attracted operators seeking cheap electricity, but tariff changes repriced that advantage. Miners now reportedly pay between $44.34 and $59.76 per megawatt-hour plus taxes, and local industry sources cited 35 companies ceasing operations after the increase.
Law No. 7300 tightened penalties for electricity theft linked to unauthorized crypto mining, raising maximum sentences to 10 years and allowing the confiscation of equipment.
Nevertheless, real capital still flows in. HIVE completed Phase 1 infrastructure at a 100 megawatt facility backed by a fully energized 200 megawatt substation, signaling that some operators see durable economics even after repricing.
The tension is clear: hydro surplus creates the initial draw, but once miners scale, the state re-prices power when it realizes they’re a concentrated, taxable offtaker, or local grid constraints and noise externalities build political pressure.
Paraguay’s trajectory shows how a hub can flip if social license breaks, making policy durability a first-order variable in any site-selection model.
What actually makes a hub
Mining hub viability in 2026 comes down to a formula: delivered cost per megawatt-hour times contract flexibility times policy durability, measured against what AI and HPC buyers are willing to pay, grid scarcity, and foreign-exchange or import friction.
Three scenarios play out across those variables.
In the first, curtailment gluts persist: renewables add faster than grids can absorb, curtailment rises, and miners win as flexible offtake. Hydro- or seasonal-surplus jurisdictions with weak transmission, such as Paraguay, or countries explicitly monetizing overcapacity, such as Pakistan, are the likeliest hubs.
In the second, AI outbids miners for firm power. Data centers seek long-term firm supply, pushing miners into interruptible, congestion-prone, or stranded pockets. Hubs emerge where miners can access interruptible pricing or “can’t-export” energy rather than prime firm capacity.
In the third, political repricing or backlash reshapes the landscape. Governments raise tariffs once miners scale or when households see shortages or noise. Paraguay becomes the template: a hub flips when the economics that attracted miners get recalibrated by the same state that built them.
The IEA’s framing matters here. Global electricity demand is forecast to grow at a roughly 4% annual rate through 2027, driven by industrial output, air conditioning, electrification, and data centers.
Renewable capacity additions are accelerating, but grid integration lags. That lag creates the curtailment and congestion that miners can monetize, but it also means surplus is a moving target.
The hubs that survive 2026 aren’t just cheap-power jurisdictions, but also places where curtailment or congestion is likely to persist, regulation tolerates mining as dispatchable load, and miners can compete with or complement AI and HPC for electrons.
The checklist
Six variables determine whether a jurisdiction becomes a mining hub or just a headline.
Surplus type is the first. Is it hydro seasonality, stranded gas, flare mitigation, or nuclear baseload off-peak? Each has different persistence and contractability.
The delivered cost and contract structure follow as the second variable. What’s the all-in price per megawatt-hour, and is the contract interruptible? Who bears congestion risk, and is there compensation for curtailment?
Then comes the ASIC import and logistics, such as customs duties, shipping lanes, spare parts availability, and capital controls, all of which affect speed-to-market and operational risk.
Policy durability is the fourth variable: tariff repricing risk, licensing requirements, sudden bans, and theft enforcement determine whether a hub stays a hub.
Climate, cooling, and water also play a part. Air-cooling limits, immersion feasibility, and heat or noise externalities constrain where large-scale operations can operate without triggering local opposition.
The last variable is offtake competition: AI and HPC demand growth is now explicitly reflected in electricity demand forecasts. Hubs must assume competition for “good electrons,” not just cheap ones.
Pakistan’s 2,000 megawatt plan is the clearest signal that governments see surplus electricity as an exportable asset class, with mining as one monetization path.
| Jurisdiction | 1) Surplus / curtailment type | 2) Delivered $/MWh + contract structure | 3) ASIC import/logistics + FX | 4) Policy durability | 5) Climate/cooling + water | 6) Offtake competition (AI/HPC) |
|---|---|---|---|---|---|---|
| Pakistan | ⚠️ Overcapacity framed as policy (“regions with excess energy”), but persistence/seasonality not yet proven | ⚠️ Price & terms TBD (headline MW ≠ delivered $/MWh; key is interruptible vs firm + curtailment comp) | ❌ FX/import friction likely (capital controls, shipping/customs uncertainty) | ⚠️/❌ Execution risk (tariff politics + IMF scrutiny could force repricing or slow rollout) | ⚠️/❌ Hot climate → higher cooling load/PUE unless sited in cooler regions | ❌ Direct competition (initiative explicitly includes AI data centers; firm power may get bid up) |
| UAE (Dubai/GCC lens) | ⚠️ “Surplus-by-design” (systems built for summer peaks → off-peak monetization potential) | ❌/⚠️ Published tariffs are high; mining needs special contracts/curtailment-ready pricing to work | ✅ Best-in-class logistics (ports, spares, finance; low friction scaling) | ✅ Generally stable permitting environment (but energy pricing is the swing variable) | ❌ Extreme heat makes cooling a first-order constraint; water/heat externalities matter | ❌ High competition (data centers expanding; miners likely pushed to interruptible/constrained pockets) |
| Paraguay | ✅ Hydro surplus draw (Itaipú-style abundance is the core “hub” attractor) | ⚠️ Still competitive but repriced (tariff hikes + taxes; economics depend on contract specifics) | ⚠️ Landlocked/logistics add time/cost; manageable but not “plug-and-play” | ❌ Durability risk (tariff repricing + enforcement pressure = hub can “flip”) | ✅ More forgiving climate than GCC; easier cooling profile | ✅ Lower AI/HPC bidding pressure vs major metro markets (for now) |
Whether that path leads to 2026’s next major hubs depends on execution, including contract terms, site selection, and whether the political consensus holds as miners start consuming gigawatt-hours at scale.
CZ’s thesis about Bitcoin as a buyer of last resort is correct in principle. The practice is messier, contingent on grids that can’t absorb renewables fast enough, states that tolerate flexible loads, and miners who can stay competitive as data centers bid up the price of firm power.
The hubs that emerge will be the ones where those conditions align long enough to build infrastructure and sign contracts that survive the first tariff revision or the first summer blackout.