By Dipl.-Ing. Daniel Bacon
A few years ago, if you had asked which slab system would dominate data centre floors, the safe answer was the pre-stressed hollow core slab. It is one of the most efficient structural products ever industrialised: enormous bearing capacity for the amount of concrete and steel it uses, a low carbon footprint per square metre, and a flat soffit that MEP engineers love. For the data centres we were designing then, spans around 8 m, sustained live loads around 12 kN/m², roughly 1.2 t/m², it was close to the perfect answer.
It is being designed in less and less often now. Not because anything is wrong with the slab, but because the building changed around it. The available cross-sections simply cannot reach the spans and loads the sector now asks for. Thicker, stronger hollow core sections may well be developed in time; they are not the standard product today.
The rack got heavier and the grid got wider. And the floor system that has emerged to carry that combination is one borrowed from heavy industry and parking structures: the double-T slab. This article is about why.
What changed on the floor
The thing that sets the geometry of a data centre floor is not the structure. It is the server rack.
A standard rack occupies roughly 600 × 1200 mm in plan, and racks are deployed in tight rows to maximise computing capacity per square metre of floor. Every column that lands inside a rack hall interrupts a row, forces a gap, and removes deployable positions. In a building where floor area is effectively sold as capacity, a column is not a neutral structural element, it is lost revenue.
That economic pressure pushes the column grid outward. A few years ago we were designing grids around 8 m. Today the expectation is more typically 10 to 12 m, and usually asymmetric: one longer span in one direction, a shorter span in the other.
At the same time, the loads have climbed. The 12 kN/m² we used to design for has risen to roughly 17 to 20 kN/m² of sustained load as rack densities increase, power, cooling, and the equipment itself all getting heavier per square metre.
Wider grid, much higher load. That is the combination that quietly redrew the structural problem, and it is what moved the floor beyond the reach of the hollow core slab.
Why the beams span short and the slabs span long
On a grid with one long and one short direction, the efficient load path is to run the primary beams across the shorter span and let the slab span the longer one.
The beams collect most of the load and deliver it to the columns. Keeping them short makes two of the hardest problems easier. First, deflection: long, heavily loaded beams are very difficult to hold within the tight limits a data centre needs. Second, the connections: transferring these large loads into the columns is far more manageable over a shorter, stiffer member. In the large majority of cases, spanning the beams the long way is simply the less economic choice.
That leaves one clear question. With the beams handling the short direction, what slab system can bridge the long 10–12 m span under 17–20 kN/m², and still hold deflections tight enough that the server racks do not tilt?
Read more: Data Center Structural Design: Understanding Live Load Requirements
Why the obvious systems fall away
The two systems you would reach for first drop out for opposite reasons.
The hollow core slab, our usual starting point, cannot carry this combination of span and load with the cross-sections available today. It is a beautifully efficient product within its envelope, but push it to 10–12 m under 17–20 kN/m² and you run out of section. The very thing that makes it efficient, the voids, is what limits it here. Thicker sections may emerge in future; for now, the sector has moved past what the standard product can do.
The cast in-situ flat slab goes the other way. It can be designed for almost anything, but at these spans and loads it becomes very thick, very heavy, and very expensive to build: full formwork and propping across the whole floor, a sequential pour-cure-strip cycle that slows the programme, and a large volume of concrete that drives up both the embodied carbon of the structure and the load on the foundations. On a multi-storey data centre you end up paying for that floor twice, once in the slab, again in the foundations carrying its weight. It is rarely justifiable.
Between an efficient system that cannot reach the span and a flexible one that is uneconomical at it, the double-T sits in the middle and does the job.
What a double-T actually is, and why it fits
Seen end-on, a double-T is exactly what the name says: a flat top deck with two downstand webs, so the cross-section reads as a double T. The webs give the element real structural depth without filling the section with solid concrete, which is precisely how it spans 12 to 18 m while keeping self-weight moderate.
For large spans the elements are usually pre-stressed. High-strength strands are tensioned before casting, then released once the concrete has cured, compressing the section and bending it into a slight upward camber. In service, that built-in compression works against the bending the load creates, so the slab carries more, spans further, and, critically here, deflects far less over time.
That camber has to be controlled, not just admired. If the upward deflection is not predicted and managed properly, the topping cast over the element can end up too thin at mid-span, where the camber is highest, and the mesh reinforcement in that topping can lose its required concrete cover. Getting this right is a detailing exercise that rewards experience, and one of the quieter reasons to keep the design with a firm that has built these elements before.
Deflection is also where the double-T earns its place. A floor that sags under sustained load does not just look wrong: it tilts server racks, distorts raised floors, and disrupts cooling distribution. The stiffness of a well-designed pre-stressed double-T is one of its strongest arguments for this application, provided that stiffness is verified against the loads and the layout the floor will actually see.
It is also a well-understood product. The double-T has been a workhorse industrial cross-section for decades, cast in standardised formwork, with or without pre-stressing. There is nothing experimental about it, only a good match between an established industrial element and a new structural demand.
Read more: Precast Hollow Core Slabs for Data Centers
Why contractors like it on site
The structural fit is only half the reason the double-T is winning. The other half is the programme.
- Few elements, large coverage. Each unit already covers a large area of floor, so a deck goes down in relatively few lifts.
- Factory reinforcement. Most of the reinforcement is already in the element. On site you add only a moderate mesh in the structural topping.
- Self-formwork. The elements act as permanent formwork for the topping, no temporary formwork, no working platform.
- No propping. Designed correctly, they need no props. That single fact changes the site: once the topping is cast, fit-out works underneath can begin immediately, because there is no forest of props in the way.
- Parallel fabrication. Elements are manufactured off-site while groundwork and foundations proceed, then assembled quickly on delivery.
- Standard cranes. Weight is moderate, so no exotic lifting plant is required.
- Embedded rails. Rails can be cast into the webs, so MEP equipment hangs directly off the structure without drilling.
In a sector where the opening date is a contractual obligation rather than an aspiration, capacity is usually sold before the structure tops out, that combination of parallel fabrication, fast erection, and immediate fit-out access is worth real money.
Where the double-T asks something back
No system is free of constraints, and a good engineer will tell you the double-T’s before you commit, not after.
The defining one is the soffit. Unlike a hollow core slab, a double-T has no flat underside, the two webs hang down into the space below. That has two consequences worth understanding early.
Cooling and airflow. Depending on the cooling strategy, the webs can interrupt air movement below the slab. This has to be checked against the mechanical design, not assumed away.
MEP routing. All services run beneath the webs, not through them. It is tempting to ask for openings in the webs to claw back some height, resist it. Web penetrations complicate fabrication, reduce the flexibility of the design, and add cost, for very little gain. Keep the webs solid and route below them. Where web depth eats into the clear zone, one practical answer is to design the sprinklers to discharge upward into the cavity between the webs, rather than fighting for space underneath. (The flat slab portion of the element, by contrast, can usually take openings without risk to its bearing capacity, it is the webs you protect.)
Embedded rail detailing. Casting support rails, HTA-type channels, into the underside of the webs lets MEP equipment hang off the structure without drilling, which is a real benefit. But where the web is pre-stressed, the detailing matters: a clear distance must be kept between the pre-stress strands and the rail’s anchor studs. The rails are normally zinc-coated, and without that separation the galvanic pairing can drive electrochemical corrosion that attacks the pre-stress wires and shortens the life of the element. It is a small detail with a large consequence, and another point that belongs in the design office, not on site.
Procurement. Double-Ts are normally supported on steel built-in parts cast into the beams, and only a limited number of manufacturers produce these. Both the cost and the lead time of those connections need to be in the programme from the start, they are not something you source at short notice.
None of these is a reason to avoid the system. They are reasons to coordinate the mechanical design and the procurement early, which on a data centre you want to be doing anyway.
How to challenge the design
If you are a developer or contractor, you do not need to run the calculations yourself. You need to know whether the slab system was chosen by analysis or by habit. A few direct questions will tell you which.
- Justify the slab system, and show the comparison. Ask the engineer to justify the chosen system and to show how it compares to the alternatives, on material use, structural depth, deflection, cost, and programme. You are not looking for an assertion that one system is best; you are looking for the comparison that led there.
- Why are the beams spanning the way they are? As a default, the beams should span the short direction and the slab the long one, the more economic arrangement in the large majority of cases. If the engineer has chosen to span the beams the long way, there may be a sound reason, but it needs a solid justification. Ask for it.
- How does the system perform under live load, against the real rack layout? The deflection question is not abstract. It should be answered against the server rack layout as actually planned, using the allowable tilt that the rack supplier can provide. Performance under live loading, measured against that layout, is what protects the equipment.
- Where does the MEP go relative to the webs? This is the question that protects your programme. The answer should be “below the webs, coordinated,” with web penetrations explicitly avoided.
- Are the built-in connection parts in the procurement programme? If the lead time on the steel supports has not been checked, the speed advantage you are paying for may not be real.
Vague answers, “this is what we always use,” or “the contractor recommended it”, are a flag. The right system depends on the project. It should be reasoned, not inherited.
Read more: Soil Improvement vs. Piling: The Data Center Foundation Question That Swings Your CapEx
Conclusion
The double-T slab is winning the race in data centre floors for a simple reason: it is the system that actually fits where the sector has moved. As spans widened toward 10–12 m and sustained loads climbed past 17 kN/m², the hollow core slab ran out of section and the in-situ slab ran out of economy. The double-T spans the distance, carries the load, controls the deflection that keeps racks upright, and does it on a programme that lets the rest of the building proceed in parallel.
It is not the right answer everywhere, and it asks for early coordination of cooling, MEP routing, embedded-rail detailing, and connection procurement in return. But for the data centre floor as it is built today, wide grid, heavy racks, schedule tied to revenue, it has earned its place at the front of the field.
gbc engineers works with data centre developers and contractors on structural design, slab system selection, load and deflection analysis, and MEP coordination from concept through construction. If you are choosing a floor system for a data centre, that is the stage to get these questions answered.
Frequently asked questions
Why are hollow core slabs used less often in modern data centres?
Not because of any flaw, they are extremely material-efficient and low-carbon, and remain an excellent choice within their range. The available cross-sections simply cannot carry the current combination of span and load: as grids widened to 10–12 m and sustained loads rose to 17–20 kN/m², the standard hollow core section runs out of capacity. Thicker sections may be developed in future, but they are not today’s standard product.
Why do the beams usually span the short direction and the slab the long direction?
The beams collect most of the load and carry it to the columns, so keeping them short keeps the two hardest problems manageable: holding deflection within limits, and designing connections that transfer very large loads into the columns. The slab then takes the longer span, exactly what a pre-stressed double-T is built for. This is the more economic arrangement in the large majority of cases; spanning the beams the long way can be done, but it should be specifically justified.
What is the main drawback of a double-T slab?
It has no flat soffit. The two downstand webs reduce the clear zone below the slab, can affect airflow depending on the cooling strategy, and require all MEP services to be routed beneath the webs rather than through them. Where the webs are pre-stressed and carry embedded rails, the rail detailing also needs care to avoid corrosion of the strands. These are coordination and detailing tasks, not deal-breakers, but they have to be resolved early, alongside the lead time on the steel built-in connection parts.
|
About us
gbc engineers
is an international engineering consultancy with offices in Germany, Poland, and Vietnam, having delivered 10,000+ projects worldwide. We provide services in structural engineering, data center design, infrastructure and bridge engineering, BIM & Scan-to-BIM, and construction management. Combining German engineering quality with international expertise, we achieve sustainable, safe, and efficient solutions for our clients.
|