On 28 April 2025, at 12:33 in the afternoon, the Iberian peninsula went dark. It was not a neighbourhood outage or a one-hour cut: it was the complete collapse of the electricity system of continental Spain and Portugal, the most severe incident on the European grid in more than twenty years. Ten months later, on 3 October 2025, ENTSO-E’s expert panel published the factual report; on 20 March 2026 came the final report, with 49 specialists, 16 meetings and a conclusion that makes almost everyone uncomfortable: the central problem was not a lack of inertia, it was a lack of held voltage.
Inertia has become the favourite headline of the energy transition: renewables do not spin, the grid loses mass, and from there to collapse only one push is missing. It is an attractive story and partly true. But the engineering of the Iberian incident says something else, and understanding it explains why half of Europe is right now rewriting the connection rules for inverters.
A grid holds two things at once, and they are not the same
An alternating-current grid keeps two quantities stable, and they behave very differently.
The first is frequency: 50 Hz in Europe, exactly the same 50 Hz at a socket in A Coruna, in a factory in Lyon and in an alternator in Poland. That is no coincidence: every synchronous machine connected to the system turns at the same electrical speed, mathematically tied to all the others. Frequency is an indicator of global balance: if more is generated than consumed, it rises; if more is consumed than generated, it falls. It is shared, continental data: you can measure it at a single point and know what is happening 2,000 kilometres away.
The second is voltage, around 400,000 volts between phases on the transmission grid. And here lies the key difference: voltage is not global, it is local. A voltage problem in Almaraz cannot be fixed from Poland or from France; it is fixed by injecting or absorbing reactive power exactly where it is missing or in excess. Reactive power carries no useful energy and travels badly along lines: the further it has to be moved, the more losses and voltage drops it creates.
That asymmetry is what decided 28 April.
Where inertia comes from (and what replaces it)
A conventional synchronous generator – a gas turbine, a combined cycle, a hydro unit, a nuclear plant – is a rotor of tens of tonnes spinning at 3,000 revolutions per minute. That rotor stores kinetic energy, and its relationship to rated power is summed up in a single number: the inertia constant H, expressed in seconds. A combined cycle sits around 4-6 seconds; a nuclear unit can exceed 6. In practice, H means that if generation suddenly goes missing, the rotor can hand over part of its spinning energy and slow down a little before anyone reacts.
The physics condenses into the swing equation: the rate at which frequency falls is proportional to the unbalanced power and inversely proportional to the total inertia of the system. With an aggregate inertia of 4 seconds on a 25 GW system, the sudden loss of 1 GW produces an initial drop in the order of 0.25 Hz per second. That figure – the RoCoF, rate of change of frequency – is what trips protections: generators start to disconnect on underfrequency around 47.5-48 Hz, and some RoCoF relays act from 0.5-2.5 Hz per second so as not to stay inside a system that is falling apart.
A solar panel, a modern wind turbine or a battery have none of that. Their power electronics – the inverter – do not spin: they switch silicon carbide transistors or IGBTs tens of thousands of times per second and synthesise the wave. With no mass there is no kinetic energy stored in a rotor, and therefore no inertia. Nor is there short-circuit current: a synchronous machine can deliver 5 to 7 times its rated current for a few tenths of a second, while an inverter typically stays at 1.1-1.5 times, limited by its semiconductors.
That has two very concrete technical consequences. First, the system loses “strength” (short-circuit ratio) and converters become more fragile near weak grids, because their control depends on measuring the voltage they see accurately. And second, almost all the power electronics installed so far is grid-following: it behaves as a current source that latches onto the existing wave through a phase-locked loop (PLL) and injects current into it. If the wave distorts or the grid loses a strong reference, the inverter is following a ghost.
Grid-forming: stop following and start holding
The industry’s answer is called grid-forming. A grid-forming inverter does not imitate a current source but a voltage source behind an impedance: it generates its own angle and frequency reference, and synchronises with the rest through power sharing (droop control, the same proportional principle turbine governors use). With that it can start an islanded grid from scratch, operate in very weak grids and – this is the important part – deliver inertial response intrinsically, without waiting to compute RoCoF, something the Australian system operator AEMO even considers preferable: measuring the derivative of frequency is slow and noisy.
Europe has moved. On 4 November 2025 ENTSO-E published its Phase II technical report on grid-forming capability, tied to the draft of the new network code on requirements for generators (NC RfG 2.0), which turns grid-forming capability into an obligation for new non-synchronous generation and storage above 1 MW. The compliance metric is not a response curve but the mechanical starting time, the formal equivalent of the inertia constant of a synchronous machine. Germany had already published a guideline in July 2024 aimed at an inertia market, and in Spain Red Electrica published its technical specification for grid-forming capability in November 2025. The simulations accompanying this work point to a threshold: in weak-grid scenarios, roughly 12% of generation would have to be grid-forming to keep the system stable.
And yet the blackout was not about inertia
All of the above is necessary for what comes next. ENTSO-E’s final report is explicit: the sequence that brought down the peninsula was not a frequency drop caused by a lack of spinning mass, but an uncontrolled rise in voltage, with generation disconnecting in cascade due to overvoltage. The panel’s simulations indicate that having more inertia would not have prevented the loss of synchronism with France. The trigger was something else.
The ingredients, according to the report, are a catalogue of control failures:
- In the half hour before the collapse two oscillations appeared: one at 0.63 Hz, forced in nature and associated with converters in the Almaraz-Carmona area, and another at 0.2 Hz, an inter-area mode that swept across all of continental Europe.
- The measures taken to damp those oscillations, such as cutting exports to France, pushed the voltage profile in southern Spain even higher: the medicine made the disease worse.
- Reactive power absorption was slow and manual: shunt reactors used to absorb voltage were switched in by hand, with decision times incompatible with an event that unfolded in seconds.
- Many renewable plants ran at fixed power factor, that is, blind to voltage. Worse still: when they reduced active power they also reduced their reactive absorption, pushing voltage further up. Several conventional synchronous units also failed to reach the reactive power reference they were obliged to meet.
- The Spanish operating margin reached up to 435 kV, against the 380-420 kV harmonised in the rest of Europe: less cushion before protections called it a day.
When voltage ran away, more than 2.5 GW of generation dropped off in cascade, along with an unexpected actor, both for its volume and its invisibility: thousands of solar installations below 1 MW connected to the distribution grid, with protections set to disconnect on overvoltage. The report found a clear correlation between peaks in transmission-distribution flows and those mass trips. Nobody saw them coming because, quite simply, there was no real-time telemetry from those rooftop inverters: the system operator had no visibility over a growing part of its own grid.
What is being changed (and it is not just adding flywheels)
The panel’s recommendations and the measures already in the pipeline point in a different direction from the easy headline:
- Voltage control instead of fixed power factor, so that renewable plants hold voltage as a reference rather than injecting reactive power in proportion to their output.
- Harmonising voltage limits at 380-420 kV and removing national exceptions.
- Observability: phasor measurement units (PMUs), early warning of forced oscillations and aggregated self-consumption data. In Spain, Order TED/82/2026 removes exemptions for self-consumption installations, which must now ride through voltage dips and meet robustness requirements.
- Adaptive defence plans, able to shed load taking all that distributed generation into account, plus mandatory black-start tests every three years, with communications that survive 24 hours without a grid.
- Synchronous condensers and STATCOMs – machines that spin or electronics that regulate voltage without producing useful energy – alongside grid-forming batteries, to hold voltage where the grid needs it.
An unresolved market problem remains: how inertia gets paid for. The United Kingdom has been separating inertia as a remunerated service, and Australia pays for very fast frequency response. But measuring the inertia an inverter provides when it reacts in 100 milliseconds is not trivial, which is why the metric chosen in Europe is a parameter equivalent to that of a synchronous machine. A figure that is measured, certified and can be penalised if it fails to show up when needed.
The Iberian blackout leaves a lesson that goes beyond renewables: the larger and more interconnected a system is, the more its balance depends on fast, local controls that cannot be governed from the top. Inertia is the visible part of the problem. Voltage, which is solved at district level, is the part that took down 60 million sockets.
