For a data centre, foundation design is not primarily a question of achieving adequate bearing capacity.
The more important question is whether the ground, foundations and slabs can maintain the movement tolerances required by a mission-critical facility throughout construction and operation.
That distinction matters because a foundation can have adequate geotechnical resistance and still perform poorly in service. Differential movement can affect floor tolerances, equipment alignment, drainage, buried utilities, structural interfaces and future expansion areas. The consequences may be disproportionate to the magnitude of the settlement itself.
Modern data centres may include high-density computing equipment, generators, cooling plant, transformers, tanks, extensive utilities and equipment with different static and dynamic loading characteristics. The geotechnical strategy should therefore begin with a performance question:
What movements can the facility tolerate, where can they occur, and over what time period?
Only then should the project team decide whether the appropriate solution is shallow foundations, a raft, ground improvement, deep foundations or some combination of these systems.
1. Start with movement criteria, not foundation type
A conventional foundation design sequence often starts by characterizing the soil, establishing bearing resistance and selecting a foundation. For mission-critical infrastructure, that sequence is incomplete.
The project team should first define what foundation performance means for the facility. Total settlement may matter, but differential settlement between adjacent supports can be more important. Local slab distortion may govern equipment areas. Relative movement between the building and buried utilities may control connection details. Mechanical equipment may introduce separate vibration or alignment requirements.
These criteria should be coordinated between the geotechnical engineer, structural engineer, equipment suppliers, mechanical and electrical designers and the owner. A generic allowable settlement value should not automatically be applied across an entire data-centre campus.
2. Ground variability can be more important than average soil strength
Large data-centre developments often occupy substantial footprints. Even where the geology initially appears straightforward, the site can contain variable fill, buried drainage features, former structures, soft or loose soil zones, changing groundwater, weathered rock and transitions between cut and fill.
The design concern is not simply whether the average soil condition can support the average building load. It is whether changes in ground stiffness across the footprint can produce unacceptable differences in movement.
A foundation crossing from competent native soil into uncontrolled fill may behave differently on either side of that transition. A relatively shallow rock surface in one zone and deep overburden in another can create a significant stiffness contrast. Dewatering or permanent drainage can also change effective stress conditions and therefore settlement behaviour.
Investigation should consequently be targeted toward reducing uncertainty where that uncertainty can change the engineering decision. More boreholes are not automatically better. Better investigation is investigation that resolves the geological transitions and mechanisms controlling performance.
3. Total settlement is only part of the problem
A single settlement number rarely describes the complete serviceability problem. Total settlement affects finished levels and long-term geometry. Differential settlement can distort framing and slabs. Angular distortion may govern sensitive structural or architectural systems. Local slab deformation may affect equipment sitting directly on slab-on-grade. Time-dependent consolidation can continue after commissioning, while excavation rebound or expansive soils can create upward rather than downward movement.
The important limits should therefore be defined according to the component being protected. This follows the same principle discussed in our earlier article How Much Ground Movement Is Acceptable Near Existing Infrastructure?: movement is acceptable only in relation to the performance requirements of the affected system.
4. When are shallow foundations or rafts appropriate?
Shallow foundations can be highly efficient where the near surface ground is competent, reasonably uniform and capable of satisfying the project's settlement criteria.
A raft can also be effective where loads are distributed across a large footprint and the structural system can redistribute localized variations in ground response. But neither solution should be selected simply because bearing resistance is adequate.
A footing can satisfy a bearing-capacity calculation while differential movement still controls the project. Similarly, a raft does not eliminate settlement. It changes the way load and deformation are distributed.
Where foundation stiffness and structural response significantly influence one another, explicit soil–structure interaction analysis may become useful. See our article When Does Soil–Structure Interaction Matter in Geotechnical Design?.
5. When does ground improvement make sense?
Ground improvement becomes attractive where the problematic soils are relatively shallow, where a large building footprint makes deep foundations commercially inefficient, or where the project would benefit from maintaining a shallow-foundation solution.
Depending on the ground conditions, potential approaches can include densification, stone columns, rigid inclusions, soil mixing, injection grouting, preloading or combinations of techniques.
For a data centre, however, specifying the treatment method is not enough. The more important requirement is to define the performance expected from the improved ground.
The design should establish the required settlement reduction, post-treatment stiffness, acceptable variability and verification criteria. Interfaces between treated and untreated ground must also be considered carefully because they can create precisely the differential movement the treatment was intended to control.
Ground improvement should therefore be treated as an engineered foundation system rather than simply as a construction activity.
6. When are deep foundations justified?
Piles or other deep-foundation systems may be appropriate where compressible or highly variable deposits extend to considerable depth, concentrated loads are large, settlement criteria are particularly restrictive, or a deeper competent stratum provides substantially more predictable performance.
But piling should not automatically be interpreted as the conservative option. Deep foundations introduce their own issues: pile installation effects, group settlement, downdrag, rigid rocksockets, testing requirements, pile-cap interaction and construction tolerances.
They can also create an important interface problem. A piled building may settle very little while surrounding slabs, pavements, service yards and buried infrastructure remain supported on soil and continue to move. Differential movement can therefore migrate from the building foundations to the utility connections and surrounding infrastructure.
The foundation system should be selected on the basis of whole-site performance rather than the movement of the primary structure alone.
7. Hybrid systems require careful transition design
A large data-centre campus may use several foundation strategies simultaneously. The main data hall may be piled. A support building may use shallow foundations. Generator pads may have another solution. Ground improvement may be used below slabs or equipment areas. External yards and buried utilities may remain conventionally supported.
Such combinations can be entirely appropriate. The risk lies at their interfaces.
A stiff piled building beside a settling slab-on-grade, for example, can create relative movement at utility penetrations. Improved ground ending abruptly beside untreated soil can produce a local distortion zone. Future construction beside an operating building can alter stresses, groundwater or access conditions.
These interfaces should be designed deliberately rather than discovered during commissioning.
8. Slabs, equipment loads and vibration form part of the same system
Data-centre equipment can impose substantial floor demands. From a geotechnical perspective, this means slab behaviour cannot be separated from subgrade behaviour.
The structural slab may be adequate but still experience serviceability problems where its support stiffness varies. Heavy equipment movement routes should be identified. Generator and rotating-equipment foundations may require dynamic assessment. Equipment anchors and slab joints may also interact with long-term deformation.
Where vibration-sensitive equipment is involved, the ground, foundation and superstructure should be considered as one dynamic system rather than as independent design packages.
9. Construction can determine whether the design assumptions remain valid
Data-centre construction is often highly schedule-driven. Earthworks, utilities, foundations, building construction and equipment installation may proceed in overlapping phases. Under these conditions, a sound geotechnical design can still be compromised if the prepared ground no longer represents the assumptions used in design.
Subgrade exposed to precipitation can soften. Construction traffic can disturb sensitive soils. Fill sources can change. Utility trenches can locally reduce support stiffness. Dewatering can affect groundwater. Ground improvement can vary spatially if verification is inadequate.
Construction-stage geotechnical control should therefore focus on preserving and verifying the assumptions that matter to long-term performance.
Does the ground actually constructed in the field behave like the ground assumed in the design?
10. Seismic performance involves more than foundation resistance
Where seismic hazard is significant, the geotechnical assessment may need to address liquefaction, cyclic softening, seismic settlement, lateral ground deformation and soil–structure interaction.
Operational continuity also depends on the equipment and nonstructural systems supported by those foundations. Foundation movement, structural response, equipment anchorage, piping and electrical systems should not be evaluated independently when seismic deformation could affect several simultaneously.
11. Verification and monitoring should answer engineering questions
Not every data-centre project requires extensive instrumentation. Where movement tolerances are tight, ground improvement is extensive or the geology is particularly variable, however, monitoring can provide valuable evidence that expected behaviour is being achieved.
Settlement surveys during fill placement can identify developing movement. Ground-improvement verification can test whether the required performance has been achieved. Foundation and slab level surveys can establish baselines before sensitive equipment is installed. Piezometers may be useful where consolidation or groundwater response matters.
The important principle is that monitoring should be linked to decisions. Collecting data without predetermined interpretation criteria, thresholds and responsibilities creates information, but not necessarily risk reduction.
12. A practical foundation-selection framework
| Ground / project condition | Principal question | Potential strategy |
|---|---|---|
| Competent, relatively uniform shallow soils | Can total and differential movement meet facility tolerances? | Shallow foundations or raft |
| Shallow weak or variable deposits | Can improvement produce sufficiently uniform stiffness? | Ground improvement + shallow foundations |
| Deep compressible soils or very tight movement criteria | Does a competent deeper stratum materially improve predictability? | Deep foundations |
| Different structures and loading zones | Can relative movement between systems be accommodated? | Carefully detailed hybrid systems |
| Heavy or vibration-sensitive equipment | Are slab, subgrade and dynamic characteristics compatible? | Integrated slab / foundation / dynamic assessment |
| Significant seismic hazard | Could ground deformation threaten operational continuity? | Seismic ground-response and deformation assessment |
The decision is therefore not “Which foundation is strongest?” It is:
Which foundation strategy provides the required level of movement control and predictability for this particular ground profile, facility layout, construction sequence and operational requirement?
When is specialist geotechnical review valuable?
The case for specialist review becomes stronger when the site combines a large footprint with significant ground variability, restrictive settlement criteria, ground improvement, mixed foundation systems, soft soils, uncontrolled fill, variable rockhead, groundwater sensitivity, seismic concerns or future expansion beside an operating facility.
The value of that review is not another bearing-capacity calculation. It is determining whether the investigation, performance criteria, foundation concept, construction controls and verification program form one coherent system.
The engineering decision
There is no universally superior foundation type for data centres. Shallow foundations, rafts, improved ground and deep foundations can all be appropriate.
The better solution is the one that provides sufficiently predictable movement performance for the actual ground conditions, operational tolerances, construction sequence and long-term requirements.
For mission-critical facilities, that generally means treating ground, foundations, slabs, utilities, equipment and construction controls as one performance system.
References
- ANSI/TIA-942-C, Telecommunications Infrastructure Standard for Data Centers, 2024.
- ASHRAE Handbook, Data Centers and Telecommunication Facilities.
- ASHRAE, Structural and Vibration Guidelines for Datacom Equipment Centers.
- Canadian Geotechnical Society, Canadian Foundation Engineering Manual.
Cover photograph: Geoffrey Moffett / Unsplash.