Summary

• Madrid, Barcelona and Lisbon are linked by network latency of around seven to eight milliseconds, according to a DE-CIX commissioned study.
• Iberia has more than 100 data centres, 35 submarine cables and a planned data centre pipeline approaching 9GW.
• The distributed model could spread AI infrastructure beyond one constrained metropolitan hub, although much of the proposed capacity remains unbuilt.

Iberia’s attempt to become a major European AI infrastructure market is increasingly being built across several cities rather than concentrated in one data centre hub, with Madrid, Barcelona and Lisbon connected closely enough to support workloads distributed around the peninsula.

A study commissioned by DE-CIX argues that the three cities now form what it calls a “Distributed Digital Capital”, combining data centres, internet exchanges, terrestrial networks and subsea cable landings into one regional infrastructure market. The label is DE-CIX’s own framing rather than an independent industry classification, but the underlying infrastructure figures help explain why Iberia is attracting greater attention from cloud and AI operators.

The study counts more than 100 data centres, 35 submarine cables landing at 20 coastal locations, 12 major terrestrial backbone networks, 13 internet exchanges and more than 800 connected networks across the peninsula. Around seven to eight milliseconds of latency separates Madrid from Barcelona and Lisbon, allowing data and computing workloads to move between the three centres without treating them as entirely independent markets.

A planned data centre pipeline approaching 9GW provides the most striking number, although pipeline capacity should not be confused with operating infrastructure. Projects can be delayed, resized or abandoned as developers encounter power, planning, financing and customer demand constraints.

AI changes the economics of concentration

Europe’s established digital infrastructure hubs grew around dense concentrations of fibre, network operators, cloud providers, customers and data centres, with each new participant benefiting from proximity to an existing ecosystem.

That concentration also creates pressure because large clusters compete for grid connections, land and planning capacity while high density AI computing adds substantially to electricity requirements. A city can offer excellent connectivity and still struggle to accommodate successive generations of large computing facilities if power infrastructure cannot expand quickly enough.

Iberia offers a different geographic proposition. Madrid provides a large inland business and interconnection market, while Barcelona and Lisbon add major subsea connectivity. Supporting locations including Sines, Bilbao and Valencia can extend that footprint further.

AI operators can potentially use that distribution to separate workloads according to their requirements. Large training clusters need substantial power and high capacity networking, but not every task needs to sit in the same building or city. Inference workloads can be placed closer to users, while storage, model training and recovery capacity can occupy different locations if network performance is sufficient.

The model only works if connectivity remains consistent across those sites. Moving large volumes of data between facilities without creating unacceptable latency, cost or operational complexity becomes as important as the capacity of the individual data centres themselves.

Subsea cables give Iberia a wider role

The peninsula sits between European networks and cable routes serving the Americas, Africa and the wider Mediterranean, giving Spanish and Portuguese facilities potential relevance beyond their domestic markets.

That position becomes more valuable as AI infrastructure spreads geographically because training and inference depend on moving large datasets and model outputs between computing regions. Network routes consequently become part of the economics of where servers are installed.

The region can also make a European sovereignty argument because data and computing can remain within EU jurisdictions while retaining efficient international connectivity. Sovereignty alone will not overcome poor economics, but regulatory control adds another location factor alongside electricity cost, fibre availability and proximity to customers.

DE-CIX has an obvious commercial interest in that argument. The company operates exchanges in Madrid, Lisbon and Barcelona and describes Madrid as the peninsula’s largest internet exchange, connecting more than 220 networks with over 11Tbps of connected capacity.

Its involvement does not invalidate the infrastructure data, although it makes the distinction between measurable assets and promotional interpretation important. The existence of cables and proposed data centres is easier to establish than the claim that Iberia has become the world’s first example of a new category of digital capital.

The 9GW pipeline faces the harder test

Planned capacity ultimately needs electricity, planning approval, equipment and customers. AI data centre announcements across Europe have grown quickly, while grid connection queues and transmission constraints have become major barriers to construction.

Iberia has potential advantages through renewable generation and available development locations, but large computing facilities still require dependable power around the clock. New projects can also create local pressure on transmission networks, land and water depending on their cooling architecture.

Spreading development between several nodes may ease some of those constraints because demand is not forced into one metropolitan area. The trade-off is greater coordination across networks, power markets, facilities and jurisdictions.

Madrid, Barcelona and Lisbon therefore offer a stronger proposition than the branding around them suggests: existing network density, international cable access and proposed computing capacity could allow AI infrastructure to expand across a wider geography than Europe’s traditional single city clusters.

Whether Iberia becomes a major AI hub will depend on how much of the announced capacity reaches operation and whether customers use it. Nine gigawatts on a development pipeline makes an impressive map; powered racks carrying production workloads provide the more meaningful measure.