A peninsula went dark in five seconds
The April 2025 Iberian blackout is the clearest live demonstration that grids are now changing faster than the models used to run them.
We were not in the room. Every reading on this desk is an outside analysis of public reporting, written to show how the method is applied — not to allege what any organisation knew or decided. Where we go beyond the public record we say so in the text.
On 28 April 2025 a cascading failure removed power across mainland Spain and Portugal within seconds, affecting tens of millions of people. Restoration took most of a day in many areas. Spain's government and grid operator conducted formal investigations; ENTSO-E convened an expert panel.
Grids were designed around large spinning machines whose physical mass resists sudden frequency change. As that generation is replaced by inverter-connected sources, the system's behaviour in the first seconds of a disturbance changes — and those first seconds are precisely where a cascade is decided. The planning models in wide use were built for the old physics. This is an inference on our part from the public technical discussion, and the formal attribution is a matter for the investigations.
- Immediate: transport, payments, water pumping, mobile networks and hospitals affected across two countries.
- Medium: accelerated European scrutiny of inertia, grid-forming inverter requirements and system strength.
- Long: every operator running a high-renewables mix now needs models that represent sub-second behaviour, not just steady-state adequacy.
- Industrial: large consumers discovered which of their processes cannot ride through a grid event, and what a cold restart actually costs them.
The class is a model that was accurate when it was built and quietly stopped matching the thing it describes. It is the most under-discussed risk in industry, because nothing announces it. Process plants, building portfolios, logistics networks and financial risk models all drift the same way — the world changes underneath a model that keeps returning confident answers.
One decision, made continuously: is this dispatch and configuration stable against a credible disturbance right now, given what is actually connected today? That requires a model kept current from the system's own data, not one refreshed on a planning cycle.
- Detect model drift as a measurable quantity rather than discovering it during an event.
- Rehearse disturbances against today's configuration, not last year's.
- For large industrial consumers: know which processes survive a dip and what a restart costs, before you find out.
- Give a regulator a model whose assumptions are explicit and auditable.
Level 5 is a model that revalidates itself against reality continuously. It is the only level that catches drift, because drift is invisible to any model that is only checked when it is rebuilt.
Score your own position in five minutes — same model, no email wall. Or read what the levels mean.
Same seven steps, your asset, your data, your decision. That is the first half of a scoping workshop.