“Long-term blackout” gets used loosely enough that people end up arguing past each other, and the confusion deepens because a week without power is a bad camping trip for somebody in rural Maine with a generator and a wood stove while being an outright catastrophe for a family on the fourteenth floor of a Miami condo with an insulin-dependent nine-year-old. Both readings are accurate, and neither is the subject here.
For the rest of this piece the definition stays blunt and fixed, meaning three months or more of a bulk transmission system being down across a whole region rather than a single neighborhood or one feeder. Three months is roughly the point at which every buffer has been spent, fuel and batteries and goodwill and savings and patience, the diesel in a hospital’s belly tank, the last pallets of bottled water at the back of a supermarket, and it is also long enough for an emergency to harden into a condition, which is what people stop reacting to.
The duration is not theoretical either, because Cuba has been living inside a rolling version of it since late 2024, with daily generation deficits averaging roughly 1,600 megawatts in 2025 on the state utility’s own published figures, which on a grid that size means half of national demand going unmet on the worst days. Havana logged a single blackout longer than twenty hours, the whole island went dark again in December 2025 with ten million people affected and no fuel reserves left, and the obvious question is what the rich version of the same failure looks like once the money and the plant inventory are better and the preparation is no better at all.
The reliability filings answer more of that than most people realize, and almost nobody reads them. In January 2026 the North American Electric Reliability Corporation published a 181-page long-term assessment that put 13 of 23 assessment areas at elevated or high resource adequacy risk inside the next five years, while utilities have announced more than 55,000 megawatts of coal retirements inside the same five-year window the assessment worries about. Demand is climbing for the first time in twenty years, pushed hard by data centers, and the interconnection queue moves at the speed of paperwork. Grid Strategies has argued the danger is overstated because the analysis ignores projects already holding interconnection agreements, and the critique is fair, since the forecast may be pessimistic while the margin is still thinner than it was.
Hardware makes the picture worse, given that standard power transformer lead times in the United States averaged about 128 weeks in 2025 and generator step-up units roughly 144, with some specialized units approaching 36 to 48 months. The Department of Energy described critical grid equipment in August 2026 as facing lead times of two or more years, and Wood Mackenzie models a 30% supply deficit for power transformers across the national fleet, which leaves a two-to-four-year gap between deciding to fix something and holding the replacement in your hands and forces restoration planning to assume that whatever breaks mostly stays broken.
The failure mode is shifting as well. The April 28, 2025 blackout across Spain and Portugal cut power to something on the order of 55 million people and took most of a day to reverse in the bulk and longer in pockets, and ENTSO-E’s final report that October identified it as the first known blackout driven by overvoltage, a cascading chain of generation disconnections fed by plant whose reactive power performance did not match what the system operator expected, made worse by manual voltage control too slow to catch the swing. No lightning strike, no fallen tree, just a control problem in a system running exactly as designed.
That last detail is the one worth carrying forward, because the Spanish grid was modern, well maintained and full of inverter-based renewables with certified voltage capability, and it still went dark in under a minute when a distributed control problem outran a human one. Everything below happens inside that gap.
