Portugal blackout spotlights datacenter risk

Datacenters are supposed to be the boring part of the internet. Always on. Redundant. Built for bad days. But the Iberian blackout turned that promise into a much sharper question — redundant against what, exactly? This week, Spain’s competition authority widened its investigation into the 28 April 2025 blackout and named more energy assets, including EDP’s Soto de Ribera thermal plant. EDP pushed back and said the plant was a backup unit that was not scheduled to be operating when the grid failed. That fight matters on its own. But for Portugal’s datacenter ambitions, the bigger point is simpler: if a whole region can go dark at once, “site resilience” stops being a server-room question and becomes a systems question. (euronews.com) ### What actually broke? The blackout hit continental Spain and Portugal at 12:33 CEST on 28 April 2025 and briefly affected part of southwest France. ENTSO-E’s final report, published in March 2026, treated it as a full-system event and laid out recommendations to harden the European grid after one of the region’s biggest recent failures. That matters because this was not a local substation accident. It was a cross-border collapse. (entsoe.eu) ### Why does that matter for datacenters? Because a datacenter does not run on “cloud.” It runs on electricity, diesel logistics, cooling water, telecom backhaul, and people who can still reach the site when transport and comms get messy. Operators love to talk about dual feeds and backup generators. But a r(entsoe.eu) the same long-haul landing corridors. If those are correlated, two “independent” sites can fail together. That is the real lesson here. (entsoe.eu) ### Where does Portugal fit into this? Portugal has been pitching itself as a digital infrastructure hub. Part of that story is geography — Atlantic landings, Europe-facing connectivity, and room to build. Part is timing — hyperscalers and cable owners want more routes that bypass older chokepoints. The appeal is obvious. If you can land traffic in Portugal and move it inland fast, the country becomes more than an edge market. It becomes a bridge. (submarinenetworks.com) ### Why does Nuvem keep coming up? Because Nuvem is exactly the kind of asset that makes Portugal look strategically important. Google’s transatlantic cable is designed with 16 fibre pairs and 384 Tbps of total capacity, and it is meant to connect the US, Bermuda, the Azores, and mainland Portu(submarinenetworks.com)able can route traffic only if the landing stations, terrestrial networks, and datacenters on each end still have power. (submarinenetworks.com) ### So is the risk power or fibre? Both — and more importantly, the overlap between them. A datacenter can survive a utility outage if generators start, fuel keeps arriving, and the network paths stay diverse. It can survive a fibre cut if traffic reroutes and the site stays energized. The hard(submarinenetworks.com) were stored in the same closet. That is what operators now have to audit. (entsoe.eu) ### What should operators be checking now? Not just UPS runtime and generator tests. They need to map where power feeds really converge, where fuel contracts depend on the same depots or roads, and whether “diverse” fibre actually shares ducts, bridges, landing stations, or metro exchanges. They also need to(entsoe.eu)dency map is suddenly the most important document in the building. (entsoe.eu) ### Why is this bigger than Portugal? Because the same sales pitch exists everywhere. Every market wants to be a resilient digital hub. But resilience is not a press release. It is the absence of hidden single points of failure. Iberia just gave Europe a very expensive reminder. The bottom line is that Port(entsoe.eu)more.