Three Blackout Problems, Not One Three-Month Collapse

A line of heterodox energy commentary now treats a regional blackout lasting three months or more as the realistic face of grid fragility: long enough that diesel, spare transformers, bottled water, and institutional patience are all spent. That duration is an analytical frame. It is not an event in the recent record of wealthy grids. What the record does contain is three different failures, each documented by official or mainstream sources, none of which is a bulk transmission system in a rich region staying down for a fiscal quarter. Keeping them separate is what makes the fragility claim interesting rather than inflated.

The first is a control failure on a modern European system, reversed the same day. On 28 April 2025 the power systems of continental Spain and Portugal experienced a total blackout. ENTSO-E, the association of European transmission operators, calls it the most severe incident on the European system in more than 20 years and the first of its kind. Contemporary reporting put the peninsula’s population near 60 million and said barely a corner escaped. Restoration of the bulk system was measured in hours, not months: accounts of the sequence put Portugal’s transmission restoration around 00:22 on 29 April and Spain’s around 04:00, with about 99 percent of Spanish demand back by morning. The October 2025 factual report described a cascade of generation disconnections accompanied by rising voltage — the first time that pairing had produced a blackout in the Continental Europe synchronous area. The final Expert Panel report, published on 20 March 2026 and not in October, attributed the collapse to interacting factors: oscillations, gaps in voltage and reactive power control, rapid output reductions, and generator disconnections, producing fast voltage increases and a cascade. Coverage of that final report added a more specific claim: some inverter-based resources disconnected before reaching the voltage limits in their connection requirements, and some conventional units delivered less reactive power than the operator had requested. No fallen tree and no official cyber finding were required. The system was running, and a voltage-control mismatch was enough. What the event does not show is destroyed hardware. It came back because the plant was still there.

The second failure mode is a supply chain, not an outage. Reuters reported in August 2025 that Wood Mackenzie expected a 30 percent U.S. supply shortfall for power transformers and 10 percent for distribution transformers that year, as demand — increasingly from data centers — outran domestic manufacturing, with utilities placing some orders three or four years ahead. Industry accounts of Wood Mackenzie’s second-quarter 2025 survey put average lead times near 128 weeks for power transformers and about 144 weeks for generator step-up units. Those week-counts are widely repeated; the underlying survey dataset was not retrieved for this review. The official corroboration is plainer. The Department of Energy’s Office of Electricity has described critical grid equipment as facing lead times of two or more years, citing import dependence, thin domestic capacity, and excessive customization, and has tied new funding to shortening that queue. The inference that a wrecked large transformer stays wrecked for years is a reading of those lead times. It is not a log of such an outage. It is the part of the fragility argument with the firmest hardware basis and the least operational proof.

The third is a chronic generation shortfall that has already hardened into a social condition, without matching the three-month total-blackout definition. Cuba’s grid collapsed on 4 December 2024 after a failure at the Antonio Guiteras plant. Reuters reported the island that evening producing roughly a sixth of typical peak demand, with millions still dark — one in a string of nationwide failures that year. Interruptions continued through 2025 and into 2026, tied to fuel shortages, breakdowns, and an aging thermal fleet. Daily notes from the state utility Unión Eléctrica, as carried by Cuban outlets, repeatedly projected peak deficits on the order of 1,000 to 1,800 megawatts against demand often near 3,000 megawatts. An annual average of roughly 1,600 megawatts is a summary claim; it was not independently averaged here from the utility’s archive. A 3 December 2025 transmission failure produced a multi-hour blackout in greater Havana. It is not clearly documented as another island-wide collapse on the December 2024 scale. Cuba is evidence of what rolling scarcity does to hospitals, water, and work. It is not evidence that a bulk system in a wealthy region stays down for ninety days.

North American adequacy sits between these modes, and the official assessment is more careful than the commentary built on it. NERC’s 2025 Long-Term Reliability Assessment, released in January 2026, finds that 13 of 23 assessment areas face resource adequacy challenges over the next 10 years, with results summarized for elevated- and high-risk areas. Summer peak demand is forecast to grow by 224 gigawatts over the decade, most of it from new data centers; planned retirements total 105 gigawatts of peak seasonal capacity over ten years, about 10 gigawatts less than the year before. Trade-press readings of the same assessment put MISO, PJM, ERCOT, and parts of the Pacific Northwest at high risk of insufficient reserves or excess unserved energy at some point inside five years. Grid Strategies’ March 2026 review accepts the direction and disputes the severity: the analysis, it argues, undercounts generation that already holds interconnection agreements and gives too little credit to interregional transmission, so the forecast may be pessimistic even while delays in interconnection and transmission remain real constraints. That critique is a documented institutional response, not a rebuttal from the fringe.

One figure in the commentary does not survive cleanly. The claim that utilities have announced more than 55,000 megawatts of coal retirements inside the same five-year window is in the neighborhood of numbers coal-industry advocates have used. It is not the figure NERC published. What the Energy Information Administration documented is a slowdown at the margin: U.S. coal retirements in 2025 were 2.6 gigawatts, the least since 2010, because operators delayed or cancelled closures. The Department of Energy has also used emergency authority to keep specific coal units running past planned retirement dates. Whether those orders meet the legal test of an emergency is in court. That they are being issued is not in dispute. Announced retirement and realized retirement have diverged.

Held apart, the three modes imply something narrower than a season in the dark, and more concrete. A voltage cascade on an inverter-heavy system can darken a modern peninsula in seconds and still be reversed in hours if the transformers survive. Spain is the existence proof. A physical loss of large transformers cannot be reversed on that clock; the supply chain says so, and no recent wealthy-grid blackout has tested the point. Chronic shortfall, as in Cuba, degrades daily life without requiring the bulk system to stay dark for a quarter. NERC’s metrics describe a fourth, intermediate risk: shortfalls under stress, counted in hours, concentrated where data-center load, retirements, and slow interconnection meet. Duration is the variable that changes the social object. A ten-hour Iberian outage already stopped trains, card payments, and some water service. Multi-day shortfalls would be a different condition. A three-month regional blackout remains a scenario.

The connection easy to miss is that the fast failure and the slow failure do not cause each other. Inverter behavior did not create the transformer shortage, and transformer lead times did not trip Spain. What they share is a thinning buffer: less dispatchable capacity held back, fewer spare large transformers on the ground, and demand growth that NERC no longer treats as a rounding error. The officially worried position — NERC’s, the Department of Energy’s, ENTSO-E’s — is not that society is about to spend a season without power. It is that the margin for error is thinner, a new electrical failure mode has already occurred, and the hardware required to recover from a worse physical one is on a two-to-four-year clock.