AI workloads and digital sovereignty have moved from side topics to fixed assumptions in every investment model. No new data-centre project in Europe can realistically move forward without addressing them — they now shape every aspect of design, financing, and operation.
AI computing has completely changed the equation. Rack densities that used to hover around 5 to 10 kW are now pushing 30, sometimes 70 kW. That kind of load forces a rethink of everything: electrical design, cooling strategy, and even the site’s energy source. Liquid-cooling corridors and on-site generation are becoming standard. The upfront capital spend is higher, but once a site can handle that density, tenants almost never leave. AI clusters don’t migrate easily, so once they’re in, you’ve essentially secured long-term yield.
Sovereignty concerns are having a similar effect on geography. More and more tenants insist on keeping capacity inside trusted jurisdictions where the regulatory environment is clear and data rules are predictable. That’s why capital is flowing toward markets that can combine solid grid access with compliant data regimes and realistic build timelines.
The underwriting process has widened too. Committees are no longer just looking at returns — they want to know about grid-queue exposure, the realism of connection dates, local permitting conditions, community sentiment, renewable integration, and ESG resilience. None of these are “nice-to-haves” anymore; they’re part of the core due-diligence checklist.