Romania without a single active nuclear unit — Anatomy of a vulnerability announced four months in advance

Between May 10 and the beginning of June 2026, Romania will for the first time in its post-2007 history experience a window of ~3 weeks without any nuclear production. Unit 2 automatically disconnected on May 4 after an isolator transformer power evacuation failed. Unit 1 entered planned shutdown at 11:00 AM on May 10, according to a budgeted calendar with 24 months in advance. Lost capacity: ~1,412 MW (2 x 706 MW), ~20% of the national average consumption and an unusually high weight in base load. On May 13, the day when total shutdown occurred, PZU price spiked to 713 lei/MWh (~137 EUR/MWh), peaking at over 1,500 lei/MWh by 21:00 — the third consecutive day with the highest spot price in Europe. The episode was not a technical failure, but an architectural one: CANDU units performed exactly as expected, yet the scheduling of maintenance schedules failed at the most basic level of critical system controls — a stop-loss on programmed decisions when context changes.

1. Chronology, in exact dates

The U2 defect is a routine engineering issue: transformer evacuation units have finite lifespans, an isolator fails, it's replaced. Operation requires isolation, disassembly, relocation, installation of standby transformer, testing and inspection (isolators, resistors, oil quality) — days, not hours. The Cernavoda reactors have a stellar operational record: historical capacity factor above global average, infrequent unplanned outages. This is not the problem. The real issue is that U1 planned shutdown review was postponed exactly to Tuesday, May 10th, after U2 had already been down for six days, without a stop-loss decision made. On Sunday, May 7th, when Nuclearelectrica announced the extension of U2 shutdown, there was a window of three workable days in which the U1 review could have been rescheduled within two to three weeks. The standard argument (“the annual review is budgeted, supplier contracts are signed, optimal hydrological window exists”) fails under opportunity cost testing. The spot market delivered the blame's price less than 72 hours later.

2. Boarding Score of the Grid — OPCOM, May 13, 2026

For comparison, Germany, which exited nuclear power generation in 2023 but with storage and batteries at scale, trades at approximately half the Romanian price on the same day. The arithmetic of this event: for 1,412 MW × 720 hours (three weeks) and a difference of ~80-100 EUR/MWh between the marginal nuclear cost (~25 EUR/MWh) and replacement costs (gas + import at solar time), the combined U1+U2 produces an impact of approximately 80-100 million EUR on the system. Who pays in the end: the market transfers to consumers through the CPC component, hedged portfolio providers, and a transfer from the budget to compensation mechanisms.

3. The Bulgarian Paradox – the vulnerability that doesn't show up in MW

Data Aprilie 2026, încă relevante ca bazal pentru asimetria:

In MWh, raportul este mai echilibrat. Diferența vine din timing: România expunează solar la timp de-mișcare, importând load și flex în seara. Bulgaria a investit în baterii grid-forming și capacitatea de balansare, transformându-se în „bateria regiunii“. Cu Cernavoda fără, asimetria crește. Capacitatea reală de import a SEN este actualizată cu aproximativ 2.500 MW (Transelectrica 2026), cu o fluctuare peste 4.200 MW până în 2030 și 7.000 MW la punctul strategic. Cu aproximativ 2.000 MW importați marti seara, România a funcționat la aproximativ 80% din plafonul curent – nu la jumătate, cum sugerau primele estimări. Sistemul n-a pierdut deplin în continuare, dar perna de siguranță s-a redus dramatic. Orice eveniment N-1 suplimentar (cadere termocentrală, consum atipic, calm anticiclonic fără vân) pune Transelectrica în zona de import maxim.

4. Why It Matters in the Long Term—Three Structural Vulnerabilities

(a) The Gap Between Policy Plans and Physical Reality. Romania has announced: U3+U4 Cernavoda (2 × 720 MW), retrofit U1 with World Bank funding (~1 billion EUR from the Romanian package of 2 billion EUR, including Transgaz), SMR Doicesti (4 modules NuScale × 77 MW). Realistic timelines: retrofit U1 completion ~2029; commissioning of U3+U4 in the optimistic scenario by 2031-2032; SMR Doicesti first module beyond 2030. Until then, the system remains with a single point of failure (U1, U2) and no nuclear redundancy—this visible problem now.

(b) SNN’s Budget Already Reduces Risk. For 2026, the company has prudently budgeted: -9% sales, -5% revenues, -14% profit, -42% dividends compared to 2025. SNN's stock price trades around ~67.5 RON, and analyst consensus (Hold) indicates a target of ~46-49 RON in the next 12 months—implying downside relative to spot. Investors already expect a difficult 2026; this episode is within the guidance. A second incident on U2 after restart or a discovered issue in U1 review extending shutdown for another 4-6 weeks standard would shake consensus and rekindle discussions about partial privatization/strategic partnerships for new units.

(c) Lack of a Real Battery Strategy. Romania's solar capacity has surpassed the nominal peak at 2,500 MW (~1,500 MW operational by lunchtime). Without batteries on scale, cheap off-peak power is exported to Bulgaria, stored there, and returns as high as 1.500 lei/MWh in the evening. Cernavoda lowers this inefficiency chronic problem into a sudden hemorrhage. An investment in 1-2 GW of utility-scale batteries would reduce reliance on off-peak supply by 30-40% and transform episodes like today's tariff crisis into operational challenges. Comparative cost: ~1.5-2.5 billion EUR for 2 GW/4 GWh storage—about half the cost of a single new nuclear unit, but with complementary functionality, not substitutive.

5. Strategic Diagnostic

The episode is not a technical failure, but rather one of system architecture. The CANDU units have operated exactly as they were designed to: high availability, automatic shutdown on defect, planned maintenance within the optimal hydrological window. What failed was the upper level – coordination of maintenance schedules, backup capacity for baseload power generation, and a lack of a "stop-loss" mechanism for decision-making during revisions in the preceding 6 days.

Three unanswered questions on May 13, 2026:

1. Why did Unit U1 not go into service two weeks after SNN's announcement on May 7 regarding the extension of Unit U2? Who made the decision and in which meeting with what written justification?

2. What is the total estimated cost for final consumer consumption of U1+U2 – ANRE is obligated to publish the ex-post estimate by components (CPC, imbalances, transmission)?

3. When did Transelectrica update its N-1 / N-2 plan for the June 2026 window and what backup capacities are contracted for gas and imports?

In the absence of these answers, the episode remains in the realm of reactive communication – announcements on the stock exchange (SNN is obligated to publish MAR) and generic statements from the ministry. The ~3-week period Romania lives without any active nuclear unit represents symbolically just over two decades. The direct cost (~80-100 million EUR at system level) is manageable. The reputational cost – for a system that promotes itself as a regional exporter of clean baseload power – is harder to quantify, but it is not zero. And the opportunity cost, measured in storage and interconnection projects that were not started when they still could have been, will be paid out steadily with each subsequent review of U1+U2.

Methodology

Data sources: HotNews — timeline U1/U2 and compensation; Antena 3 CNN — SNN announcement extending U2 shutdown; InvestEnergy — CERNODAU U1 planned shutdown timeline; OPCOM PZU price data on 13 May 2026 — average, minimum, maximum cost per kWh; Focus Energetic — Bulgaria import/export report in April 2026; Transelectrica — current import capacity of 2,500 MW and forecast peak of 4,200 MW in 2030; Nuclearelectrica IR — CANDU 6 (706 MW U1, 705.6 MW U2) installed power. Cost calculation: 1,412 MW × 720h × average spread ~80-100 EUR/MWh = cumulative cost of 81-102 million EUR over a three-week window (assumption: constant spread at median observed price on May 12-13; higher volatility expected during the period). Limits: resynchronization of U2 to June 1 remains an estimate, not confirmed yet; review of U1 takes standard duration of 4-6 weeks — any delay increases cost by ~25-35 million EUR per additional week. Data cutoff: May 13, 2026, 14:00 UTC.