Summary
- Nebius has reportedly signed an 18MW capacity agreement at Merlin and Edged’s Madrid-Getafe campus.
- The site is part of a wider Iberian data centre platform built around Madrid, Barcelona, Bilbao-Arasur, and Lisbon.
- The reported deal strengthens Madrid’s claim as a southern European AI infrastructure market.
Nebius has reportedly taken 18MW of capacity at the Madrid-Getafe data centre campus, giving the European AI infrastructure provider a first position in Spain.
The capacity sits within the campus being developed by Merlin Properties and Edged, a platform designed to bring large-scale data centre capacity into Iberian markets. Edged lists Madrid-Getafe as a central Spanish fibre and logistics hub, with MG1 at 20MW, MG2 at 48MW, and a stated PUE of 1.15.
The reported agreement connects AI compute demand to one of Spain’s more visible data centre developments. It also sharpens Madrid’s role in the competition for workloads that need dense power, resilient cooling, and a credible path to live capacity.
Madrid moves beyond secondary status
Spain’s data centre market has changed quickly. Madrid is no longer discussed only as a secondary location behind Frankfurt, London, Amsterdam, Paris, and Dublin. The city now sits inside a wider southern European growth story built around enterprise demand, cloud adoption, fibre routes, available development land, and access to renewable energy.
Nebius brings a different kind of demand from conventional enterprise colocation. AI cloud infrastructure is shaped by GPU density, energy cost, cooling performance, and the ability to scale. An 18MW block is large enough to influence both campus utilisation and market perception, particularly in a region trying to attract AI workloads that might otherwise land in northern Europe or the Nordics.
The technical claims around Madrid-Getafe are central to the commercial pitch. Edged markets the campus with waterless cooling, renewable energy, and a low PUE. Those numbers will be tested under real operating conditions as density rises, but they reflect the buying criteria now shaping AI infrastructure decisions. Space alone is not enough. Tenants need electrical density, thermal resilience, efficient heat rejection, and confidence that the site can scale without running into power or cooling limits.
Iberia’s power advantage is not automatic
Iberia is attractive because it offers land, solar and wind resources, improving connectivity, and lower saturation than Europe’s largest hubs. That advantage still depends on firm grid capacity and delivery certainty. Earlier Iberian plans from the Merlin and Edged platform were reportedly adjusted after power supply guarantees could not be secured in Andalusia.
Madrid’s development case therefore rests on more than demand. The market must show that connection capacity, grid reinforcement, renewable procurement, backup systems, and local permitting can keep pace with campus growth. AI tenants will not wait indefinitely for a power path to mature.
The Merlin and Edged model brings real estate scale and specialist data centre development into the same platform. Merlin contributes an established Iberian property base, while Edged brings technical delivery and operating expertise. Nebius adds a visible AI customer signal. Together, those elements strengthen the campus case, but they also place more pressure on execution.
Spain’s data centre boom will be judged by the speed at which contracted demand becomes commissioned capacity. Public announcements can move faster than transformers, cooling plant, and grid studies. Madrid-Getafe will need to convert the promise of efficient, waterless AI-ready capacity into stable operations.
AI load raises the bar
AI infrastructure places less forgiving demands on facilities than traditional enterprise workloads. Dense GPU deployments draw more power per rack, concentrate heat, and increase the cost of downtime. They also shorten the distance between data centre design and customer revenue: if the facility cannot support the hardware, the commercial model breaks.
That raises the importance of cooling architecture. Waterless designs can reduce pressure on local water systems, but they depend on heat rejection equipment, site layout, ambient conditions, and electrical efficiency. High-density tenants will want evidence of performance under load, not only design intent.
For Madrid, the Nebius lease would strengthen the city’s position in Europe’s AI capacity market. It would also pull Iberia’s infrastructure assumptions into a more demanding phase. Land and renewables are useful, but live AI capacity depends on power systems, mechanical design, commissioning discipline, and operational resilience.
If Madrid-Getafe delivers as planned, it will give Spain a stronger reference point for AI-ready colocation. If grid, cooling, or delivery assumptions tighten, the campus will illustrate the same constraint visible across Europe: demand is abundant, but capacity only exists where the physical systems are ready.
