Summary
- The University of Hamburg’s Regional Computing Centre requires repairs after structural damage and pollutant contamination were identified.
- Parts of the facility remain in use because essential technical infrastructure is still installed, including high-performance computing systems.
- The project exposes the refurbishment risk facing older research computing estates, where building fabric, cooling systems, and live services are tightly entangled.
The University of Hamburg is preparing repairs to its Regional Computing Centre after structural damage and pollutant contamination were identified in the building.
The facility, located around Rothenbaumchaussee and Schlüterstraße, remains partly operational because essential technical infrastructure is still installed. The work is expected to involve more than routine refurbishment, with reported repairs including work to the load-bearing façade and temporary steel support to stabilise parts of the structure.
The computing centre is not a commercial colocation facility, but its challenge is familiar across the wider digital infrastructure estate. Technical systems have to remain available while the building around them needs intervention, and the repair programme has to protect continuity, access, equipment, and services at the same time.
Research compute inside ageing buildings
University and research computing often sits in a different part of the market from hyperscale data centre development, yet the operational pressures are increasingly similar. Compute, storage, networking, and security systems are expected to support high-availability services, while many institutions still run them from buildings designed before today’s power densities, cooling methods, and service expectations became standard.
The University of Hamburg’s Regional Computing Centre also supports high-performance computing. In 2024, the university brought its Hummel-2 cluster into operation at the facility, replacing an earlier system and adding new research computing capacity for astrophysics, chemistry, physics, computer science, machine learning, and other AI-related workloads.
Hummel-2 includes 178 compute nodes with 192 compute cores each, 32 GPGPU cores for specialist applications, 5.2PB of storage, and a further 500TB of SSD/NVMe capacity. The university has said the CPUs, GPUs, and memory in the compute nodes are directly water-cooled, with a longer-term goal of reusing about two-thirds of the waste heat by extracting it through the cooling water.
That technical profile raises the stakes for refurbishment. A water-cooled research cluster is not just a stack of IT equipment. It depends on pipework, heat exchangers, pumps, controls, water quality, service access, and building-level resilience. Repairs to the fabric of the building therefore touch the operating environment around the compute system, even when the work itself begins as an estates problem.
Refurbishment becomes continuity planning
Structural damage and pollutant contamination add complexity because they can limit access, constrain working methods, and alter sequencing. If the building has heritage or architectural constraints, the repair route may be slower and more careful than a standard commercial refurbishment. Technical teams then face the practical question of how to keep infrastructure running, protected, and recoverable while contractors work around it.
Older technical estates across Europe are now facing similar pressures. Research computing, public-sector platforms, telecoms rooms, and enterprise data rooms often carry important workloads, but their buildings may have accumulated power, cooling, cabling, and security changes over many years. Asset condition can become a resilience risk long before a server fails.
The move towards liquid cooling strengthens the link between building and compute. Direct water cooling can improve thermal performance and create heat-reuse opportunities, but it also places more mechanical infrastructure inside the operational chain. Pipe routing, isolation, leak protection, pumping capacity, maintenance access, and heat export all have to coexist with building fabric constraints.
For universities, the difficulty is compounded by research demand. HPC systems are shared platforms, and downtime can affect grants, simulations, doctoral work, and collaborations. Moving workloads away from a damaged building may not be quick if software environments, storage, data movement, and specialist hardware are tightly coupled to the local system.
The Hamburg repair programme will be judged less by the visibility of the works than by its ability to protect essential services during the intervention. As European institutions invest in AI and high-performance computing, more of these projects will sit in the gap between data centre engineering and traditional estates management. Buildings built for earlier computing generations will have to carry denser, hotter, and more operationally critical loads, while remaining safe and maintainable.
