EXPLAINER

Why Spain lost power in 90 seconds

On 28 April 2025, Spain and Portugal lost power in about ninety seconds. It was not renewables, not inertia and not a cyberattack — it was a voltage collapse. Here is what happened, and what stopped it happening again.

Watch the 40-second version: Spain Lost Power in 90 Seconds — a Short from the Grid & Silicon channel. This page has no long-form companion video yet; the Short is linked, not embedded, because the site's explainer pages normally embed long-form videos.

Ninety seconds, minute by minute

The blackout was not one event. It was four phases, and the first three took about half an hour to set up and ninety seconds to finish.

12:00–12:30 — oscillations. The grid began swinging. Between 12:03 and 12:07 a forced oscillation at about 0.6 Hz — a rhythmic swing in power and voltage, local to the Iberian Peninsula — was traced to an anomaly in a generation plant. Between 12:16 and 12:22 a second, slower oscillation at about 0.2 Hz set the whole peninsula swinging against the rest of the European grid. The operator damped both oscillations by switching on extra lines and reducing flows across the border. That worked — but it raised the voltage, because less-loaded lines inject more reactive power into the system.

12:32:00–12:33:18 — generation losses from high voltage. From 12:32, voltage began rising rapidly across the whole network. Generators have protection settings that disconnect them automatically if voltage goes too high, to protect their own equipment — and they began tripping, one after another. Three large trips in southwest Spain did most of the damage: 355 MW near Granada at 12:32:57, 727 MW near Badajoz at 12:33:16, and 928 MW across Segovia, Huelva, Badajoz, Sevilla and Cáceres at 12:33:17. Each disconnection removed the plant's ability to absorb reactive power, so the voltage rose further and tripped the next plant. In total, more than 2.5 GW of generation vanished in roughly 78 seconds.

12:33:18–12:33:30 — collapse. Frequency fell and the peninsula lost synchronism with the rest of continental Europe at about 12:33:19. The automatic defence plans — load shedding designed to save a grid in trouble — switched on in Spain and Portugal, but they are built to fight a frequency collapse, not a voltage collapse, and they could not stop this one. The AC lines to France and Morocco disconnected at 12:33:21. By about 12:33:30 the voltage was zero. Spain and Portugal were dark.

12:33:30 on 28 April to 07:00 on 29 April — restoration. The grid came back the way grids always come back: in islands. Operators re-energised the network from the AC interconnections with France and Morocco and from hydro plants that can start without grid power, built stable islands of supply, and joined them together. Half of demand was restored by 22:00 on the 28th. Portugal's grid was fully restored at 00:22 on the 29th, Spain's transmission system by about 04:00, and 99.95% of demand was covered by 07:00. Eight people died in outage-related circumstances — seven in Spain and one in Portugal — mostly from improvised generators and candles.

Why voltage killed the grid, not frequency

A power grid has two things to keep in balance, and they are not the same thing. Frequency is the same everywhere on the grid: in Europe it is 50 cycles per second, and it falls when demand exceeds supply and rises when supply exceeds demand. Voltage is local: it varies from place to place, and it is controlled by moving reactive power — the background energy sloshing around in an alternating-current system — in and out. Inject reactive power and voltage rises; absorb it and voltage falls.

Most blackouts people know about are frequency blackouts: too much demand, not enough generation, frequency falls, and the system sheds load or collapses. Spain was the opposite. It was a voltage blackout. The system had enough generation. What it did not have was enough voltage control, in the right places, fast enough. And once voltage started rising, the system's own protections made it worse: every generator that tripped to protect itself from high voltage also stopped absorbing reactive power, so the voltage rose further and tripped the next one.

Five weaknesses lined up behind that feedback loop. First, the margin between the normal voltage and the voltage at which generators must disconnect was thin on the 400 kV network — Spain's upper limit is 435 kV, Portugal's 420 kV — and voltage had been drifting upward since mid-morning. Second, renewable generators in Spain operated at a fixed power factor, meaning their reactive power moved in step with how much power they were producing, not with what the voltage needed — so when voltage rose, they did nothing to pull it down. Third, the conventional plants that did have a voltage-control obligation did not meet it: some diverged from the reactive-power output required under operating procedure P.O. 7.4, and there was no financial consequence for doing so. Fourth, the shunt reactors — the equipment that absorbs reactive power and pulls voltage down — were switched by hand, and at 12:32 only about 58% of Spain's shunt-reactor capacity was connected; the rest could not be switched in time. And fifth, many generators tripped at voltages below the thresholds the rules set — they disconnected before they were obliged to, removing their voltage support exactly when it was needed.

None of these five alone would have caused a blackout. That is the point the investigators kept returning to: the grid is designed to survive the loss of any single element — the N-1 criterion — and this sequence exceeded it. A local problem, handled a step at a time, escalated into a system problem in ninety seconds because each step made the next one worse.

What it was not

Three explanations dominated the first days after the blackout. All three were wrong, and the investigations said so plainly.

It was not the renewables' fault. At the time of the blackout, 82% of Spain's generation was renewable and 18% was not. That fact was widely quoted as if it were the explanation. It was not. The expert panel convened by ENTSO-E, Europe's association of transmission operators, concluded the root cause was an overvoltage cascade made worse by insufficient voltage control — and its president summarised it as: the problem is not renewable energy, but voltage control, regardless of the type of generation. Wind and solar did not cause the blackout; the panel found they helped restore supply afterwards. The real distinction was not between renewable and conventional generation, but between plants that were required to control voltage dynamically and plants that were not. In Spain, thermal plants had that obligation. Renewable plants did not yet have the same one.

It was not an inertia problem. Inertia is the kinetic energy stored in the spinning parts of big generators: when frequency changes suddenly, the spinning mass carries on and slows the change, buying the control room seconds to react. Wind turbines and solar panels do not provide it, so it was natural to blame the blackout on too little inertia. Spain's grid operator, Red Eléctrica, said explicitly in its June 2025 report that the incident was not caused by an inertia problem — and that on 28 April the system's inertia was higher than ENTSO-E recommends. The panel found that even with significantly higher inertia, the loss of synchronism would not have been avoided, because the failure was in voltage, not frequency. Inertia buys time against a frequency fall. It does nothing against a voltage rise.

It was not a cyberattack, and not a single failure. The Spanish government's investigation analysed 133 GB of data and ruled out a cyberattack, in the system operator's facilities and in the power plants analysed. An early suggestion from Portugal's grid operator — a rare atmospheric phenomenon — was ruled out when Spain's meteorological agency reported no unusual weather. And there was no single point of failure to find: the panel identified fifteen interacting factors, and no one of them alone would have collapsed the system.

How the power came back — and what changed

Restoration took the rest of the day and most of the night, which is fast for a blackout of this size and slow for everyone living through it. Trains stopped, traffic lights went dark, card payments failed, and mobile networks faded as their batteries ran down. Hospitals ran on backup generators. By 22:00, half of demand was restored; by 07:00 the next morning, 99.95% of it was.

What changed afterwards falls into three parts. First, the rules. On 12 June 2025, Spain's regulator CNMC approved operating procedure P.O. 7.4, which extends dynamic voltage control to renewable generation: plants of 5 MW or more connected to the transmission network can now volunteer to follow real-time voltage setpoints from the operator and be paid for it, with a market for procuring reactive power. The procedure had been ready for some time before the blackout and was approved shortly after it; the setpoint mode began operating with financial effect in March 2026. Second, the operator. Red Eléctrica's own report made fifteen recommendations, above all implementing dynamic voltage control, and in the months after the blackout it scheduled extra synchronous plant — mostly gas-fired — to hold voltage while the new rules took effect. Third, the European level. ENTSO-E's expert panel set out recommendations covering voltage control, oscillations, protection settings, system defence plans and restoration, and observed that regulatory frameworks have to adapt to a power system that no longer looks like the one the old rules were written for.

The general lesson is the one the panel put last in its report: the power system is governed by physics, and market mechanisms, regulatory frameworks and energy policies have to stay aligned with the physical limits of the system. Spain's rules required voltage control from some plants and not from others, and paid no penalty when the obligated plants did not deliver it. The blackout was what that gap cost.

What it means for a data center

A data center in Madrid on 28 April had the same experience as everyone else: the grid went down, and the building had to keep itself alive. That is what backup power is for — batteries carry the load for minutes while diesel generators start, and the generators carry it for as long as the fuel lasts. Most facilities rode it out that way. UpCloud reported its Madrid data center had no visible interruption: power transferred automatically to backup generators and the site ran self-sufficiently for the remaining hours of the blackout.

But the blackout also showed the limit of that arrangement. A peer-reviewed study of the event found that internet traffic at the peninsula's largest exchange points fell by 33–70% within minutes, and then revealed a second, previously unreported outage: hours into the blackout, a partial power failure inside a major data center, when its backup diesel generators began to fail. That second failure had a larger effect on global internet routing than the nationwide blackout itself. Backup power is a bridge, not a destination — it works until the fuel, the batteries or the machines themselves give out, and a blackout that lasts ten hours tests all three. Our explainer on why backup power isn't enough sets out what that bridge can and cannot do.

There is a second lesson, about concentration. A regional blackout hits every data center in the region at once — which is one reason data centers gather where infrastructure already exists, and why a successful hub can become harder to protect. And there is a third, about the connection itself: a data center does not just need power, it needs a grid connection that survives the events the grid is designed to survive. Our explainer on how data centers connect to the grid follows that connection from the transmission line to the rack.

Spain has also started writing the lesson into rules. In December 2025 its digital ministry opened consultation on a draft decree that would require electronic communications and certain digital infrastructure — including some data centers and internet exchanges — to keep operating for 24, 12 or 4 hours during a power interruption, depending on the tier. It is a draft, not law.

Questions readers ask

Was it the renewables' fault?

No. At the time, 82% of Spain's generation was renewable — but the investigation found the cause was an overvoltage cascade made worse by insufficient voltage control, not the generation mix. The plants that were supposed to control voltage did not do it, and renewable plants were not yet required to help. ENTSO-E's president put it as: the problem is not renewable energy, but voltage control, regardless of the type of generation.

Could it happen in the United States?

The physics is the same everywhere: any grid can suffer a voltage collapse if voltage control fails in the wrong places at the wrong speed. Whether it can happen in a given grid depends on that grid's rules — who is obliged to control voltage, how fast they must respond, and what happens when they do not. Spain's failure was in those rules as much as in the equipment. A grid whose rules cover all generation, with consequences for non-delivery, is harder to break this way.

Why did it take hours to bring the power back?

A blacked-out grid cannot simply be switched back on. There is no power to start the power plants with, so operators have to build the grid back in islands: start a hydro plant that can start without grid power, energise a small area around it, add generation and demand in blocks, and join the islands together once each is stable. In Spain that meant starting from the interconnections with France and Morocco and from black-start hydro, and it took until 07:00 the next morning to cover 99.95% of demand.

What is inertia, and why didn't it save the grid?

Inertia is the kinetic energy in the spinning parts of large generators. When frequency changes suddenly, the spinning mass slows the change and buys the control room time. It protects against frequency falls. Spain's blackout was a voltage rise, not a frequency fall — and inertia does nothing against rising voltage. Spain's grid operator confirmed the system's inertia on the day was above ENTSO-E's recommendation.

Did the data centers stay up?

Mostly, on backup power — batteries, then diesel generators. But backup power is a bridge, not a destination: a study of the event found a major data center suffered a partial outage hours into the blackout when its generators began to fail, and that failure disrupted global internet routing more than the blackout itself. The industry minimum is measured in hours of fuel on site, and a ten-hour blackout tests it.

Sources and further reading

Glossary — plain definitions for words used here.