How Location, Waste Heat, and Modularity Shape the Next Generation of Data Centers

The demand for computing power is growing faster than ever, driven by the advancing digitization of industrial processes. This is causing data centers to become larger, more technologically advanced, and more complex in terms of regulations. With capacities of 40, 70, or even over 100 megawatts, data centers today are high-performance systems that require a close integration of all disciplines from the first feasibility study to commissioning. At the same time, time pressure is increasing: Technological innovation cycles now run faster than many traditional construction processes. This is one of the biggest challenges of the coming years.

At ATP architects engineers, we draw on over two decades of integral expertise in data center planning – from colocation centers with one megawatt to campus projects in the triple-digit megawatt range. Several years ago, we established specialized DC Expert Hubs located in Vienna, Zurich, Frankfurt, and Krakow, among others. In these hubs, we work across countries on data centers and projects using a common BIM and Revit platform, which has been an ATP standard since 2010. This platform lays the foundation for complex projects with international teams to function smoothly across all disciplines and phases of work.

Why Location Matters Today
Suitable locations for data centers are becoming scarcer, inner-city sites are increasingly unsuitable for large data centers because the power capacities are simply no longer sufficient. Major cities like Vienna have limited network capacities available; developers are therefore already looking in the surrounding areas for suitable land.

The result: Today, almost anyone planning a large data center almost inevitably ends up in the countryside – also because the conditions for brownfield projects are rarely met. Data centers have extreme requirements: support and suspension loads of the building services engineering are around ten kilonewtons – loads that existing buildings are usually not designed for. In addition, there are requirements for power connection capacity and planning permission, which brownfield developments often fail to meet in practice. The detour via the demolition of existing structures ultimately leads back to new construction.

This does not mean that brownfield projects have no future. For smaller data centers with a need for rapid implementation, an appropriate existing building can indeed offer advantages – especially since approval procedures for existing structures can sometimes be shorter. Special location qualities, such as an existing water right for cooling, can make brownfield developments additionally attractive. However, the decision must be made on a solid foundation – one of the central tasks in the early work phases.

If the data center heats the swimming pool
Besides the location choice, one factor is becoming increasingly important: the use of waste heat. Data centers produce enormous amounts of heat. The regulatory framework is increasingly making the use of this waste heat a requirement: For example, the Energy Efficiency Law (EnEfG) in Germany prescribes staggered waste heat utilization rates for new data centers from one megawatt - while it is currently 10 %, it should be 20 % by 2028. In parallel, the Heat Planning Law (WPG) stipulates that heat networks in municipalities will be fed from renewable energies or unavoidable waste heat by 65 % in the future. Data centers can thus become a possible component of municipal heat supply – if the location is right.

The proximity to hospitals, swimming pools, industrial plants, or existing district heating networks becomes a tangible planning criterion. Those who identify potential heat consumers early on in the site selection process not only create regulatory security but also genuine added value for the project and the region.

Added Value in Modularity
In addition to site selection, planning and construction time are essential factors for success. Technological developments, especially in chips and AI hardware, are so dynamic that delays in the construction process lead to real competitive disadvantages. The solution lies in modularity. Prefabricated structural systems from the factory not only shorten the shell construction phase but also allow pre-assembly of cooling lines, cable trays, and building services components, which then only need to be assembled on-site. Projects that would hardly be feasible with traditional in-situ concrete construction can thus be implemented in significantly less than 24 months of construction time.

Modularity is not only a question of new construction. Refits – i.e., the replacement of outdated server infrastructure in existing data centers – are becoming an equally important issue. Server generation cycles are shortening, AI-heavy workloads require increasingly dense and powerful rack configurations. Those who do not plan adaptability from the beginning quickly face a structural problem. This applies to both new construction and existing buildings.

The direction is clear: When server generation cycles become shorter than planning and construction times, the data center itself must become a modular, interchangeable component. Those who consider this in their designs today are not just planning for the next project, but for the generation after next.