Serbia’s proposed €2.6 billion Iron Gate 3 pumped-storage project is in contracted preparation, but its investment case depends on an operating regime for the Danube that safeguards Serbian and Romanian hydropower plants and downstream nuclear cooling. Locally, the project is known as the Đerdap 3 project. Its final configuration and cost remain subject to design and review.
The current Serbian starting case envisages capacity of up to 1,800 MW. The government aims to complete the project documentation by the end of 2029, after which construction would still depend on technical, environmental, bilateral and financing decisions. A consortium led by the Jaroslav Černi Water Institute, which is controlled by Millennium Team, has won the planning and early technical-documentation assignment.
Bechtel–Enka has qualified for further negotiations in the government’s international supplier-selection process. These steps establish substantive preparation work, but do not demonstrate financial close or an unconditional commitment to build. Contracting studies advances the project; the key investment test is whether it remains financially viable once binding water-management restrictions, construction risks and financing costs are included.
Existing generation and cooling depend on the river
The existing Iron Gates cascade already provides substantial generating capacity on both sides of the Serbian-Romanian border. EPS, Serbia’s state power utility, operates 1,140 MW at Đerdap 1 and 270 MW at Đerdap 2. Romania’s Hidroelectrica operates 1,166.4 MW at Porțile de Fier I.
Porțile de Fier II has about 251.2 MW in its main plant, with a further 54 MW at Gogoșu. Together, the existing hydropower assets account for approximately 2.9 GW of capacity. Their generation depends on natural river flows, hydraulic head, equipment availability and coordinated operation, so installed capacity does not guarantee equivalent output when water levels are low.
Further downstream, Nuclearelectrica’s Cernavodă nuclear station has approximately 1.4 GW of gross capacity across its two existing reactors and relies on the Danube for cooling. The river thus supplies hydropower generation, nuclear cooling and, under the Đerdap 3 concept, the water used to store electricity. These functions can complement one another, but compete for water when supplies are scarce.
The 2026 drought has demonstrated that exposure. In October, both Cernavodă units were kept in a safe shutdown state, pending improved Danube water levels. Nuclear safety remained protected, but electricity production was interrupted.
Nuclearelectrica’s revised 2026 forecast cut expected annual grid deliveries from 9.080 TWh to 6.743 TWh. The reduction also reflected an earlier transformer-related outage and cannot be attributed solely to drought. Đerdap 3 did not cause these interruptions, as it has not been built; they illustrate the operating vulnerability against which its potential additional effects must be assessed.
Pumped storage changes the timing of water use
Đerdap 3 is planned as a pumped-storage development connected to the existing Iron Gate I reservoir. EPS’s established concept would take water near Danube kilometre 1007, pump it to elevated reservoirs at Pesača and Brodica, then send it back through turbines when electricity is needed. An older, full-development concept envisaged 2,400 MW, 578 million cubic metres of reservoir water and an energy equivalent of approximately 484 GWh; those figures are distinct from the current 1,800 MW starting case.
Pumped storage shifts electricity from one time to another: electricity is used to lift water, and some of that energy is recovered when the water is released for generation. The value of the scheme depends on storage duration, charging opportunities, market prices and operational flexibility. Water use also differs between initial filling and routine operation.
During initial filling, water is retained in new upper reservoirs. The downstream effect would depend on the filling rate, river conditions at the time and whether other releases offset the withdrawal. A staged programme during suitable high-flow periods could have substantially different effects from rapid filling during drought.
Routine cycling is different because water pumped uphill is subsequently returned. Across a completed cycle, lasting water losses mainly come from evaporation, seepage and other consumptive uses; electrical efficiency losses do not equate to the same percentage loss of water. Even limited net annual consumption would not prove an absence of downstream impact, since temporary retention can matter when intake levels, navigation depths or agreed releases are close to their limits.
Cernavodă is around 700 river kilometres downstream of the conceptual Đerdap 3 intake, with existing reservoirs, coordinated releases, tributaries and river branches between them. A withdrawal at the proposed site would therefore not cause an immediate, identical reduction in flow at the nuclear station. Assessment requires a hydraulic model tracing effects through the whole system, including how water is distributed towards Cernavodă’s supply route.
Romania is already addressing local cooling-water constraints through dredging and river-channel interventions, while also considering the potential needs of a future expanded Cernavodă nuclear station. A Serbian study limited to water levels near the proposed powerhouse would therefore be insufficient.
Drought support would depend on reserved water
EPS has also presented Đerdap 3 as a possible source of support for downstream users during drought. Water stored beforehand could theoretically be released to help maintain downstream Danube levels and improve cooling conditions for nuclear plants. The mechanism is physically plausible: stored water can later be released, contributing to electricity generation and increasing downstream flow.
The benefit would depend on the volume available, the timing of release and the effect reaching the relevant intake. It would also depend on whether the water had actually been reserved for drought support. As an illustration, a reserve of 100 million cubic metres could provide an additional flow of 100 cubic metres per second for about 11.6 days, or 500 cubic metres per second for about 2.3 days, before losses and downstream routing are taken into account.
These are illustrative calculations, not forecasts for Đerdap 3. They show why installed megawatts alone reveal little about the project’s ability to support the Danube during a prolonged drought. A credible arrangement would specify minimum reserve volumes, release triggers, responsibilities and compensation, as well as when downstream water support takes priority over electricity trading.
The same stored water cannot at once be treated as fully available for unrestricted market trading, Serbia’s strategic reserves and guaranteed Romanian drought support. The most challenging case would be a prolonged regional shortage, when hydropower output falls, nuclear cooling is constrained and electricity prices rise. Đerdap 3 could generate using water stored earlier, but replenishing that water would require electricity and a permitted opportunity to withdraw water.
This links hydrology directly to revenue. Rules designed to protect downstream users could restrict recharging just when the regional electricity system is under the greatest strain.
Romanian rights could shape the operating regime
Romania’s strongest point of influence comes from its rights within the existing shared Iron Gates system. Bilateral arrangements require each side to avoid actions that interfere with the other’s use of its installations. They also require agreement on certain modifications that could change the operating parameters of the jointly managed system.
The implications for Đerdap 3 would depend on its final design. If the scheme required changes to shared facilities or agreed reservoir operations, Serbia could not assume domestic approval alone would be sufficient. Romania could withhold agreement from a configuration that materially altered the shared system’s functioning.
Existing operating rules also limit the ability to retain water. For the relevant downstream reach, the bilateral regime sets minimum discharge conditions and restricts daily operational water-level variations. A new storage plant could not simply hold water whenever electricity prices made pumping attractive.
Environmental assessment brings separate obligations. Projects that may have significant transboundary impacts require notification, assessment and consultation with neighbouring states. This does not necessarily give Romania an automatic veto over every Serbian authorisation, but environmental objections, hydraulic effects and rights related to the shared system could delay the project, require redesign or lead to operating restrictions.
Romania could therefore influence the project without demanding its cancellation. Conditions covering initial filling, pumping, reservoir fluctuations, drought reserves, minimum releases or compensation could change its viable scale and commercial value. For lenders, an unresolved operating regime would be a major financing risk: a technically buildable power plant may still be unattractive if restrictions materially reduce the hours in which it can operate profitably.
Černi and Bechtel–Enka are involved in preparation
The Jaroslav Černi Water Institute will help establish the hydrological assumptions, environmental baseline, technical alternatives and cost framework used in subsequent decisions. Its contracted assignment includes a pre-feasibility study with general design and a special-purpose spatial plan with strategic environmental assessment. Cross-border impacts and the need to agree key parameters with neighbouring countries are central to this work.
Millennium Team controls the institute through its ownership structure. That relationship does not in itself show that the technical work is biased or predetermined. It does strengthen the case for independent technical and financial review, given that an institute within a wider construction group is helping define a project that could later generate substantial construction work.
A financeable process should independently test the preferred design, hydrology, reservoir assumptions, construction costs and projected market revenues. It should also compare the proposal with smaller, phased or alternative storage configurations. The assessment must be able to conclude that the project should not proceed if its economics fail once realistic constraints are included.
Bechtel–Enka has qualified for further negotiations on additional project and technical documentation, giving the consortium a strong prospective position. Qualification, however, is not a completed construction award. Coordination between the Černi-led work and any later Bechtel–Enka assignment will matter, with reservoir design, geotechnical assumptions, Romanian conditions, grid requirements and cost contingencies based on a common technical foundation.
Without that shared basis, later changes could lead to redesign costs, delays and contractual claims. An experienced international contractor could strengthen execution capacity, but would not resolve water-management rights or establish the project’s commercial viability.
Economics must reflect operating restrictions
Đerdap 3’s commercial rationale is to absorb or buy lower-value electricity and sell electricity back when its value is higher, potentially alongside balancing and ancillary services. Because of efficiency losses, the selling price must cover more than the initial purchase cost. At an illustrative 80% round-trip efficiency, electricity bought at €50/MWh would cost €62.50 for each MWh later discharged, before operating expenses, network charges, financing costs and capital recovery.
The project could also affect the prices on which its business case depends. A large pumped-storage plant could raise prices while charging and lower peak prices while generating. The feasibility study therefore cannot simply apply historical price spreads to all future output.
The economic model would need to assess hydrology, nuclear availability, renewable generation, electricity prices and transmission constraints together. Drought can simultaneously reduce hydropower production, constrain nuclear generation and push wholesale electricity prices higher. That is when storage may be most valuable, but restrictions on water use could prevent or limit recharging.
The decisive feasibility assessment must therefore use a constrained operating scenario, rather than an unrestricted technical maximum.
Preparation is under way, but no investment decision has been made
Đerdap 3 has advanced beyond a political announcement, with contracted technical preparation, government involvement, Serbian-Romanian engagement and interest from major international engineering and construction groups. This gives the project greater substance than earlier conceptual announcements, but it is not a financed construction project. The next milestones include a consistent final design, independently reviewed hydrological modelling, an operating regime agreed with Romania, realistic capital costs and a financing model based on the restrictions the project would actually face.
Serbia can commission studies and negotiate with contractors, while Romania can materially influence the conditions governing the shared Danube system. The river will determine how much water is available. Whether Đerdap 3 becomes a regional storage asset or remains in feasibility work will depend on how much commercial value is left once these factors are assessed together.
