A simultaneous drought-driven shutdown of every operating nuclear power plant in South-East Europe would remove almost 5.9 GW of dependable generation from the interconnected market. On its own, that would be serious but potentially manageable for several days. Combined with weak hydrology, restricted thermal-plant cooling, high air-conditioning demand and reduced cross-border availability, it would become a regional electricity-security event requiring industrial demand reduction, emergency generation and coordinated TSO intervention.
The stress test covers Paks in Hungary, Kozloduy in Bulgaria, Cernavodă in Romania and Krško in Slovenia. Their combined net operating capacity is approximately 5,918 MW: 1,916 MW at Paks, 2,006 MW at Kozloduy, around 1,300 MW at Cernavodă and 696 MW at Krško. The capacity figures are consistent with the latest IAEA reactor data for Hungary, Bulgaria and Slovenia.
The event is not a central forecast. It is a severe common-mode stress test. The plants use different cooling configurations and operate under different river-level, intake and discharge-temperature limits. Yet Paks, Kozloduy and Cernavodă all depend directly on the Danube system, while Krško relies on the Sava. A prolonged regional drought accompanied by extreme temperatures could therefore constrain several plants at the same time, even when the exact timing and extent of their derating differed.
ENTSO-E’s standard 2026 Summer Outlook did not identify a systemic adequacy problem across continental Europe under its central assumptions. That conclusion does not cover an extreme combination in which nuclear, hydro and part of the thermal fleet are simultaneously unavailable. ENTSO-E’s longer-term adequacy work already warns that declining dispatchable thermal capacity cannot automatically be replaced by additional intermittent renewables and calls for more storage, demand response, cross-border infrastructure and firm capacity. ENTSO-E
The direct nuclear deficit
At an assumed pre-event operating level of 90 per cent, the nuclear fleet would normally supply approximately 5.3 GWcontinuously. A total shutdown would therefore produce an energy deficit of approximately 0.90 TWh after seven days, 1.79 TWh after fourteen days and 3.84 TWh after thirty days. Measured against full net capacity, the maximum thirty-day loss would reach approximately 4.26 TWh.
That volume could theoretically be replaced by gas-fired generation. Producing 4.26 TWh in combined-cycle gas turbines with efficiency of 52–55 per cent would require approximately 0.75–0.80 billion cubic metres of gas. The physical problem is that SEE may not have enough simultaneously available gas capacity, pipeline flexibility and power-network capacity to deliver the replacement where it is needed. Some gas plants may also be unavailable, commercially mothballed or dependent on constrained transmission nodes.
Replacing the nuclear output with lignite and coal would increase emissions by roughly 4 million tonnes of CO₂ over one month. Predominantly gas-based replacement would still add approximately 1.4–1.7 million tonnes. These figures exclude the additional fossil generation needed to compensate for depressed hydropower.
The nuclear loss is therefore only the first layer of the stress. The same drought capable of restricting cooling water would also reduce Danube and Balkan river flows, lower reservoir inflows and diminish run-of-river production. Reservoir operators would have to choose between generating aggressively during the first days and preserving water for later evening peaks. Overproduction at the beginning of the crisis could convert an energy shortage into a more severe capacity shortage during its second or third week.
Under a severe regional drought, hydro availability during critical hours could be 4–7 GW below seasonal norms across the wider SEE market. Another 1–3 GW of coal, lignite and gas generation could be derated by cooling-water limitations, high ambient temperatures, equipment failures or interrupted river logistics. Electricity demand during a sustained heatwave could simultaneously rise by 5–10 per cent, adding several gigawatts to the evening peak.
The combined regional stress would consequently be much larger than the 5.9 GW nuclear outage. During the most difficult evening hours, the deterioration in the supply-demand balance could reach 10–15 GW compared with normal summer conditions. That is the number that matters for system security.
Hungary becomes the first pressure point
Hungary would face the sharpest immediate exposure. Paks normally provides close to half of domestic electricity production, and the country already relies materially on imports during many hours. Removing 1.9 GW from Paks while solar production declines after sunset would leave Hungary dependent on gas generation, lignite from Mátra, batteries, demand response and imports through Slovakia, Austria, Croatia, Serbia, Romania and Slovenia.
Import capacity should not be confused with import availability. The physical interconnectors may remain operational, but neighbouring markets must possess surplus electricity at the same hour. In the proposed stress, Romania, Bulgaria and Slovenia would be seeking imports rather than offering them. Austria could also be affected by low hydro availability, while Serbia and Croatia would be managing their own summer balances.
Hungary’s large solar fleet would reduce the daytime deficit, particularly between late morning and mid-afternoon. It would not resolve the evening adequacy problem. High temperatures reduce photovoltaic efficiency, while demand can remain elevated after sunset. Batteries would help bridge the solar ramp but, at present deployment levels, could not replace Paks through a succession of windless nights.
Romania loses two pillars at once
Romania would lose around 1.3 GW from Cernavodă, normally representing roughly one-fifth of its electricity generation. More importantly, a drought severe enough to affect Cernavodă would almost certainly also weaken production across the Romanian hydro portfolio.
Romania’s diversified fleet provides more resilience than Hungary’s. The country can draw on hydro, gas, coal, wind and solar. The difficulty is correlation. Hydro and nuclear could fall together, solar would disappear during the evening peak, and periods of summer heat can coincide with limited wind output. Domestic gas plants would then become the principal balancing source.
Romania could move rapidly from occasional exporter to sustained importer. Congestion between generation areas and major consumption centres would add an internal grid problem to the regional energy deficit. Available cross-border capacity with Hungary, Bulgaria, Serbia and Ukraine would become commercially valuable, but those interfaces could not substitute for several gigawatts of domestic firm generation continuously.
Bulgaria’s export position disappears
The shutdown of Kozloduy would remove approximately 2 GW, close to one-third of Bulgaria’s normal electricity production. Bulgaria would lose the generating asset that ordinarily supports both domestic baseload supply and exports into Greece, North Macedonia, Serbia, Romania and Turkey.
The country could increase lignite production from the Maritsa East complex, subject to unit availability, coal stocks, cooling conditions and emissions economics. Some of that capacity is ageing, however, and cannot be assumed to perform at nameplate output through a prolonged heatwave. Higher Bulgarian lignite generation would also raise EU ETS exposure sharply.
Bulgaria would probably retain adequate supply during many daytime hours because of solar generation and thermal capacity. Evening and overnight conditions would be tighter. The market would cease functioning as one of the region’s principal export anchors precisely when neighbouring systems needed Bulgarian electricity most.
Krško creates a dual-country deficit
Krško’s 696 MW may appear small beside Kozloduy or Paks, but its ownership structure spreads the consequences across two systems. The plant is jointly owned by Slovenia and Croatia, with output shared between them. A shutdown would therefore remove approximately 348 MW from each country’s supply portfolio.
Slovenia would lose a large share of dependable domestic production while Alpine and Sava hydrology was already weak. Croatia would lose its contractual nuclear entitlement while facing high coastal cooling demand and reduced hydro availability. Both countries would turn more heavily towards imports and gas-fired generation.
This would tighten the Slovenia–Italy, Slovenia–Austria, Slovenia–Croatia and Croatia–Hungary corridors. Italian prices could increasingly influence the western Balkan market, while congestion would prevent full price convergence.
Serbia faces a supply crisis without owning a reactor
Serbia has no nuclear plant, but it would be among the most exposed non-nuclear markets. The country sits between Hungary, Romania, Bulgaria, Bosnia and Herzegovina, Montenegro and North Macedonia and is both an importer and a transit route during stressed periods.
Low Danube flows would weaken production at Đerdap 1 and Đerdap 2, while low reservoir inflows would constrain Drina and Lim system flexibility. High temperatures could affect cooling and performance at lignite-fired plants. Coal quality, mining continuity and unit reliability would then determine whether EPS could support the system or would need sustained imports.
The loss of Paks, Cernavodă and Kozloduy would remove approximately 5.2 GW around Serbia’s northern and eastern borders. Electricity might still be technically available from Greece, Turkey, Italy, Austria or Central Europe, but it would have to cross multiple congested interfaces. Serbia could therefore face scarcity prices even during hours when total European generation remained sufficient.
The immediate Serbian risk would be concentrated in the evening ramp. Solar generation would assist during the day, but current storage capacity would not carry the surplus far enough into the night. Industrial consumers with interruptible contracts, behind-the-meter generation or batteries would acquire considerable system value.
Albania, Montenegro and Bosnia lose the value of hydropower flexibility
Albania and Montenegro are highly exposed to hydrology. Bosnia and Herzegovina combines hydro with lignite and coal generation, but ageing thermal units and mine-related constraints limit dependable availability. During a deep drought, these markets could not perform their traditional balancing role.
Their reservoirs should not be dispatched simply because spot prices are high at the beginning of the emergency. Water would need to be allocated against a regional scarcity curve, preserving hydro energy for the hours with the highest probability of involuntary load shedding. That requires coordination between TSOs, utilities and water-management authorities rather than isolated commercial optimisation.
North Macedonia and Kosovo would also face elevated risk because of limited domestic flexibility, ageing lignite assets and import dependence. Greece would have greater capability to respond through gas generation, LNG supply, solar and interconnections, but its own air-conditioning load could absorb much of that flexibility. Turkey could provide limited support through Bulgaria and Greece, although cross-border capacity would cap its contribution.
Imports cannot be the regional solution
Every national emergency plan tends to treat imports as the first balancing resource. The approach works for an isolated plant outage. It fails when several importing countries experience the same weather shock.
The realistic gross import contribution from outside the affected region might reach several gigawatts through Austria, Slovakia, Italy, Ukraine, Greece and Turkey. The firm amount available during every critical hour would be much lower. Commercial flows would be limited by cross-border net transfer capacity, internal network congestion and the need for exporting countries to preserve their own reserves.
The SEE market would likely split into several scarcity zones. Hungary and the central Balkans could become the tightest area. Romania and Bulgaria would compete for eastern supply. Slovenia and Croatia would increasingly price against Italy and Austria. Congestion income would rise sharply, but high congestion revenue would be evidence of insufficient infrastructure rather than successful supply protection.
Regional day-ahead prices could move into the €250–500/MWh range for sustained periods, with individual scarcity hours exceeding €1,000/MWh. Price caps would not create physical electricity. Where demand response and emergency imports remained insufficient, TSOs would have to move from market measures to controlled industrial curtailment.
A one-month nuclear deficit priced at an incremental scarcity premium of €150–300/MWh would add approximately €640 million–€1.28 billion to regional wholesale procurement costs. Including the simultaneous hydro deficit, balancing actions and congestion, the direct electricity-market cost could exceed €1.5–3 billion before industrial production losses, emergency fuel procurement and government compensation.
The first response must be operational
During the first 24–72 hours, the region would need a single coordinated emergency dispatch process. Nuclear safety limits could not be relaxed merely to preserve generation. The available flexibility would instead come from postponing non-essential plant maintenance, activating strategic reserves, maximising safe thermal output, suspending discretionary pumping loads and shifting industrial consumption away from the evening peak.
Large industrial users should be divided into critical, flexible and interruptible categories before the crisis. Aluminium, steel, cement grinding, electrolysis, cold storage, pumping, water treatment and some refining processes can provide controlled reductions, although their technical limits differ. Contracted demand response is materially cheaper and safer than improvised disconnection.
A region-wide reduction of 2–3 GW for three to five evening hours would materially reduce the probability of uncontrolled outages. Compensation would be expensive but still lower than the economic cost of unplanned industrial shutdowns. Hospitals, water systems, telecommunications, district cooling, mines and safety-critical industrial loads would remain protected.
Reservoir hydro should be treated as strategic capacity rather than ordinary energy. Batteries should be fully charged during solar-rich hours and reserved for the evening ramp, frequency response and congestion management. Pumped-storage plants should avoid pumping during tight periods even when their usual trading algorithms indicate an opportunity.
TSOs would need to release all technically available cross-border capacity while retaining sufficient remedial-action margins. Redispatch, countertrading and emergency assistance should be coordinated by regional security coordinators. National restrictions on exports would be counterproductive: they could strand generation on one side of a congested network while increasing the likelihood of cascading failures elsewhere.
Gas becomes the emergency fuel, but only temporarily
Over a two-to-six-week event, gas-fired generation would be the main controllable replacement for nuclear and hydro. Greece’s LNG infrastructure, the Trans-Balkan system, TurkStream-related routes, the Bulgaria–Serbia interconnector and Romanian domestic production would all become part of the electricity-security response.
A severe thirty-day scenario involving both nuclear and hydro replacement could require an incremental 1–1.5 billion cubic metres of gas, depending on thermal availability and demand reduction. This volume should be secured through seasonal options and strategic contracts before summer rather than purchased entirely during the emergency.
Gas alone is not a long-term adequacy strategy. It introduces fuel-price, pipeline and carbon exposure and can create a second infrastructure constraint. It is nevertheless the most realistic multi-day bridge until the region develops sufficient storage, demand flexibility and geographically diversified renewable supply.
Existing coal and lignite assets may also be required as strategic reserve during an extreme event. Their retention should be structured as time-limited capacity insurance with clear availability testing, environmental conditions and retirement pathways. Paying ageing plants merely to exist without proving their start-up reliability would create accounting capacity rather than physical security.
Storage must be sized for the event, not the average day
A few gigawatt-hours of batteries can reduce price volatility but cannot replace 5.9 GW of nuclear generation for weeks. A four-hour battery fleet providing 5 GW would store only 20 GWh, equivalent to less than four hours of lost nuclear output.
A credible first regional resilience target would be approximately 8–12 GW of battery power with 30–50 GWh of energy, distributed around major load centres, renewable clusters and congested substations. This would not cover a multi-week deficit, but it could manage the daily solar-to-evening transition, protect frequency, reduce peak imports and prevent short periods of scarcity from becoming system emergencies.
At an installed cost of approximately €220–350 per kWh, the battery component would require roughly €7–15 billion, depending on duration, connection scope and local content. The investment case would need capacity payments, balancing revenue, congestion services and availability obligations rather than dependence on spot-market arbitrage alone.
Closed-loop pumped storage has greater value for longer events because it is less exposed to natural river flows than conventional reservoir hydro. Projects should be screened around existing mines, industrial reservoirs and suitable elevated sites. A regional target of 3–5 GW with 30–60 GWh of usable storage would strengthen multi-day resilience, although development and construction would generally require five to ten years.
Grid reinforcement is the limiting investment
New generation cannot protect the region when transmission capacity is unavailable. SEE needs reinforcement of both international interconnectors and the internal 400 kV, 220 kV and 110 kV networks that deliver imported electricity to load centres.
Priority corridors include Hungary–Serbia–Romania, Romania–Bulgaria, Bulgaria–Serbia, Serbia–Bosnia and Herzegovina, Serbia–Montenegro, Montenegro–Italy, Slovenia–Croatia–Hungary and the Greek connections towards Bulgaria and North Macedonia. Transformer capacity and internal congestion are as important as the overhead lines themselves.
A regional programme adding 4–6 GW of effective cross-border transfer capability, backed by internal substations, phase-shifting transformers, dynamic line rating and remedial-action schemes, could require €5–10 billion. Lead times of three to seven years mean that projects must enter permitting, procurement and financing well before the next severe drought exposes the same weakness.
Serbia’s connection restrictions and long development timelines are especially relevant. A new overhead line, transformer bank or high-voltage substation normally requires several years even when the corridor has already been identified and financing is available. Grid resilience cannot be created through emergency procurement after the hydrological event has begun.
Nuclear adaptation remains part of the solution
The stress test does not support abandoning nuclear generation. It shows that cooling-water resilience must become part of nuclear availability planning. Site-specific measures include deeper or redesigned intake structures, improved debris and sediment management, cooling-tower upgrades, hybrid wet-dry cooling, larger heat sinks, recycled industrial or municipal water and more accurate river-temperature forecasting.
Dry and hybrid cooling reduce water dependence but impose higher CAPEX, auxiliary consumption and efficiency penalties. Retrofitting an operating nuclear station can cost several hundred million euros and requires careful safety assessment. The investment should be compared with the economic cost of even one prolonged shutdown, particularly at Paks and Kozloduy, where a single site represents a very large share of national firm generation.
Outage planning should also be coordinated regionally. Planned maintenance at one reactor fleet should avoid the periods when hydrological forecasts indicate elevated cooling risk elsewhere. Nuclear, hydro and thermal outage schedules can no longer be optimised independently.
The broader resilience package would require approximately €18–30 billion across storage, grid reinforcement, cooling-system adaptation, demand-response platforms and flexible reserve capacity. It should combine EU funding, EIB and EBRD lending, regulated network investment, capacity-market revenue and private storage capital. The financial envelope is large, but it is comparable with the cost of repeated scarcity episodes, emergency imports and interrupted industrial production over the assets’ operating lives.
The central danger is not the isolated loss of one reactor. SEE systems are designed to withstand individual equipment failures. The real vulnerability is the simultaneous loss of several forms of supposedly independent flexibility: nuclear cooling, hydro inflows, thermal cooling, import availability and evening renewable output. A drought capable of aligning those failures would turn national import strategies into a regional competition for the same limited megawatt-hour. Only contracted demand response, firm fuel-backed capacity, multi-hour storage and stronger transmission can prevent that competition from ending in industrial curtailment.
