Summary
- European AI-focused colocation signings reached 420MW in H1 2026, compared with 89MW a year earlier.
- The Nordics account for 66% of the contracted capacity, reflecting lower-cost power and greater renewable availability.
- Deposits and letters of credit are helping operators and funders manage the counterparty risk attached to newer AI infrastructure providers.
European data centre operators signed 420MW of AI-focused colocation capacity during the first half of 2026, almost five times the 89MW recorded in the same period last year, according to new research from CBRE.
The figures point to a rapid shift in the type of customer absorbing new European capacity. Rather than demand being concentrated solely among established hyperscalers and enterprise tenants, emerging AI infrastructure providers — often described as neoclouds — are committing to increasingly large blocks of power-intensive space.
CBRE said 66% of the AI-related capacity signed during the period is expected to be delivered in the Nordic region. Lower electricity costs and comparatively plentiful renewable generation are helping markets including Norway, Sweden, Finland, and Denmark compete for workloads that can tolerate greater geographic flexibility than latency-sensitive cloud services.
The distribution is notable because many of Europe’s traditional data centre hubs are already constrained by some combination of grid availability, planning restrictions, land scarcity, and lengthy connection queues. AI demand may be increasing the absolute amount of capacity required, but it is also changing where operators and customers are prepared to put that capacity.
The second part of CBRE’s findings concerns counterparty risk. Neocloud providers generally have shorter operating histories and smaller balance sheets than the largest global cloud platforms, which has previously made developers, lenders, and infrastructure investors cautious about treating long-term leases with them in the same way as hyperscale contracts.
Operators are increasingly using financial protections such as rental deposits and letters of credit to bridge that gap. Those structures do not eliminate customer concentration or credit risk, but they can give developers and their lenders greater confidence that contracted revenues are sufficiently protected to support project financing.
Power changes the map
The Nordic concentration also shows how AI infrastructure is weakening some of the assumptions that shaped the previous generation of European colocation development. Frankfurt, London, Amsterdam, Paris, and Dublin remain important connectivity and enterprise markets, but access to several hundred megawatts of affordable electricity has become a decisive site-selection factor for large training clusters.
That pushes more development towards markets where land and renewable generation may be easier to secure, while increasing the importance of long-haul fibre, subsea connectivity, and transmission infrastructure. It also gives operators with Nordic land and power portfolios a stronger position when negotiating with customers whose principal requirement is large-scale compute rather than immediate proximity to an established metro hub.
The shift is unlikely to remove pressure from Europe’s established markets. AI inference, cloud interconnection, financial services, enterprise workloads, and sovereign-compute requirements still create demand close to major population and business centres. What changes is the proportion of very large AI deployments that can be separated geographically from those workloads.
For developers, the resulting market is becoming more segmented. Sites with modest but highly connected capacity remain valuable in established metros, while power-rich campuses in secondary locations can compete for much larger AI blocks. Operators able to demonstrate both routes increasingly have more flexibility over which customers they place on which assets.
The financing question will remain central as contracts get larger. A 100MW commitment from a relatively young AI provider creates a different risk profile from a similar lease signed by one of the world’s largest technology companies. Deposits and letters of credit can improve that profile, but the underlying economics still depend on the customer’s ability to maintain compute demand over the length of the contract.
CBRE’s figures therefore show more than a surge in leasing. They point to a European data centre market in which power geography and customer credit are becoming increasingly connected. The Nordics can offer the electricity needed for large AI clusters, while financial structures are being adapted so operators and lenders are willing to build around customers that barely featured in their underwriting models two years ago.
