News & Insights

The demand curve has broken the old delivery model. AI training clusters now consume hundreds of megawatts per deployment, inference is scaling faster than anyone forecast, and the industry is being asked to build more capacity in the next five years than in its entire prior history. Conventional stick-built construction, with its 24-to-36-month design-permit-build cycle, simply cannot deliver at that pace.

The response is modular, but modular can mean two very different things.

One end of the spectrum has fully modular data centers: complete, self-contained facilities built inside transportable enclosures or containers. Power, cooling, controls, and IT infrastructure are integrated, tested, and shipped as a finished unit. This approach offers the highest degree of factory completion and repeatability, but it also brings limits in scale, transportation, site integration, and customization.

The other end of the spectrum, and more consequential shift for large-scale data centers, is prefabricated modular construction. In this model the building remains a larger integrated system, but major components move off the job site and into factories. Electrical distribution, cooling plants, piping assemblies, busway, and other systems are fabricated, assembled, and tested under controlled conditions before arriving on site. The modules are then integrated into the permanent facility.

The benefits are speed, repeatability, better factory conditions, and relief from a construction labor market that cannot staff enough stick builds. But prefabrication does more than move construction indoors. It changes where engineering happens, when systems are tested, and where responsibility for integration sits.

That distinction matters. A containerized data center can be tested largely as a complete product. A prefabricated system still must become part of a larger building, electrical plant, cooling system, controls network, and utility infrastructure. Its reliability depends not only on the module itself, but on the interfaces between modules and the systems around them.

Modular delivery is therefore not a shortcut that leaves everything else unchanged. It fundamentally alters how systems are designed, coordinated, tested, and commissioned. The schedule advantage is real, but it is only durable if the engineering and commissioning disciplines also move into the factory, into the interfaces between modules, and into a more collaborative delivery model.

The question for owners is not whether to go modular, it is to what degree. Done well, modular saves time. Poor execution and a factory mistake can become an availability problem in the field.

The modular data center market: where growth is headed

The market data reflects how quickly modular has moved from pilot to default. Industry analysts project the global modular data center market growing from roughly $30 billion in 2024 to nearly $80 billion by 2030, a compound annual growth rate (CAGR) approaching 18 percent – and a second forecast tracks a nearly identical trajectory, with North America holding about 41 percent of the market. Independent forecasters converge tightly on the same picture: the market nearly triples by decade’s end at a high-teens growth rate. The driver is structural, not cyclical – AI density, compressed timelines, and stretched equipment lead times have made scalable, factory-built capacity a necessity rather than an option.

Why factory fabrication beats field construction

The case for moving work into the factory rests on process control. A factory has an inventory of parts, a stable skilled workforce, and the ability to schedule, test, and resolve problems under repeatable conditions. These are advantages an isolated, congested, weather-exposed job site can’t match. Prefabrication is, at its core, the industrialization of construction: reducing lead times, meeting high standards with consistency, and lowering capital project risk. The productivity case is well established beyond data centers – industry analysts’ research on modular construction finds that off-site methods can accelerate project timelines by 20 to 50 percent while reducing costs by up to 20 percent, driven largely by standardized design and controlled manufacturing conditions.

That control translates into four concrete merits. First, a controlled environment yields higher, more repeatable build quality than field assembly. Second, work happens in parallel: factory production runs while site civil works proceed, and the two streams converge at delivery instead of stacking end to end. Third, standardized, repeatable blocks let lessons from one unit carry directly into the next, rather than every facility being a bespoke build. Fourth, factory assembly leverages scarce skilled labor far more efficiently, concentrating where it’s most productive.

The trade-off is that risk and cost shift earlier. Owners commit capital sooner by locking in standardized modules, and logistics complexity increases. Most importantly, risk concentrates at the interfaces – the points where independently built modules must integrate. That is precisely where design and commissioning attention must move.

What the industry is doing today – and the payoff

Many top-tier cloud providers now run a modular program in production, and several providers are on second-generation designs. Server-room construction has moved into factories, where pre-assembled skids integrating racks, power distribution, and security systems compress on-site assembly from roughly fifteen weeks to potentially two or three, with internal targets aiming for shell-start to first operational room in under 35 weeks. The approach spans scales, from large hyperscale programs down to factory-built edge pods with integrated liquid cooling that reach operation in months rather than years.

The payoff shows up in both schedule and cost. Highly modularized projects can achieve schedule improvements of 30-to-50 percent, reducing delivery timelines from 24-to-36 months down to roughly 12-to-16 months. Some operators report sub-nine-month timelines for highly standardized, pre-engineered deployments. On cost, industry analyst estimates put an all-in modular solution at roughly $13.5 million per megawatt, compared with about $14.6 million for traditional on-site construction – an 8 percent reduction per-megawatt– while compressing schedules by about 36 percent, or seven to nine months.

That schedule advantage has significant economic value of its own. At an estimated $500,000 in operator value for each month a megawatt comes online earlier, accelerating delivery of a 50 MW data hall can translate into roughly $200 million in undiscounted benefit. Much of that gain comes from shifting work from sequential to parallel: site preparation can advance while equipment and modules are fabricated, factory testing replaces on-site commissioning iterations, and modules arrive validated rather than assembled from scratch in the field. Factory assembly has become a notable construction path across the industry, not an experiment at its edges.

Modular design: challenges and a new model of collaboration

Design challenges

Modular delivery challenges the traditional design model. Instead of designing one integrated facility from the ground up, teams are designing repeatable systems that must work together across physical, electrical, mechanical, controls, and structural interfaces.

A standardized power block, ranging from 500 kW to several megawatts, must protect itself and integrate with adjacent modules and upstream utility infrastructure. As the number of blocks grows, so do the coordination challenges. Electrical items include fault-current contributions, protective-device coordination, grounding, impedance, controls, and the interaction between independently designed systems.

The same is true for mechanical systems. Cooling modules must connect to a larger plant while maintaining flow, pressure, controls, redundancy, and operating sequences. The physical arrangement matters just as much. Modules need to fit through doors and corridors, around structural elements, and into their final positions with the clearances required for installation, operation, maintenance, and replacement.

Prefabrication forces decisions to freeze earlier

Engineers must design for a range of deployment scenarios rather than one known site condition. Standardized modules must still adapt to the realities of each site. There is a real tension between repeatability and site-specific requirements, and the industry is still developing the standards and practices needed to manage that balance.

IEEE’s work on P3710 reflects a gap the industry is now confronting. Modular systems create interfaces that traditional standards do not fully address. This guide is intended to help establish a common framework. But the hard work remains with the project team to make sure the modules connect, communicate, and operate as one system.

Collaboration starts in the model

In a modular project, collaboration cannot wait until construction. The design must prove that the pieces will fit together before they leave the factory.

That starts with a detailed digital model. BIM models developed to LOD 400 can define the actual geometry, connections, supports, access zones, and maintenance clearances of modular systems. DfMA principles take that model a step further, designing assemblies around how they will be fabricated, transported, installed, tested, and serviced. The goal is to find the problem in the model or the factory, not in the field.

VDC extends that discipline to the entire facility. Modules must be checked against the building as it is constructed, not just against a two-dimensional floor plan. Equipment clearances, structural openings, delivery paths, rigging zones, doors, corridors, and installation sequences all need to be understood before the first module arrives. A module that fits on paper but cannot be moved into position is not a successful design.

The same model can bring the electrical, mechanical, structural, architectural, controls, and construction teams together around one coordinated environment. Vendors and fabricators can contribute actual equipment geometry. Contractors can test installation sequences. Owners and operators can review access and maintenance. The model becomes more than a design deliverable. It becomes the common language between the factory and the job site.

Collaboration moves upstream

Because so much of the risk now lives at the interfaces and in the factory, the people who understand fabrication, installation, commissioning, and operation must be involved much earlier.

Fabricators know what can be built and tested efficiently off-site. Contractors know how modules will be transported, rigged, set, and connected. Operators know how the systems must be accessed and maintained for decades. Commissioning teams know where functional boundaries and control sequences can fail.

Advanced analysis also becomes part of that collaboration. CFD modeling can validate airflow and thermal performance before modules are installed, helping teams understand how standardized cooling systems will behave within the actual building and IT environment. Electrical studies, controls simulations, and other digital analyses can provide the same confidence for systems that must operate together but may be designed and fabricated by different parties.

The design team’s role therefore changes. It is no longer simply the author of a fixed set of drawings. It becomes the steward of a coordinated system. Bring the owner, vendors, fabricators, contractors, commissioning team, and operators together around a common model and a common set of performance requirements.

Modular commissioning: challenges and a new model of collaboration

Commissioning challenges

Modular delivery moves the center of gravity of commissioning upstream. Modules undergo factory acceptance testing before they ship, so on-site work shifts from ground-up validation toward placement, final connections, and energization. That is a real gain, but it introduces the single most important caveat in modular commissioning: factory-tested does not mean system-coordinated. Modules are acceptance-tested in isolation, but system-level behavior only appears once they interconnect on site. A protection relay setting that coordinates within one power block may fail when that block runs in parallel with three others behind a shared utility transformer.

The practical consequence is that integrated systems testing is no longer a single event at the end of construction. It is distributed – partly executed at the factory, where module-level integration can be proven under controlled conditions, and partly executed on site, where cross-module behavior, campus-level cooling distribution, and utility interaction can only be validated once everything is connected. Planning must account for both halves and, critically, for the seam between them.

That seam creates a second reality unique to modular delivery where some commissioning must be repeated. A module can leave the factory fully validated and still arrive compromised because shipping and handling expose it to vibration, shock, thermal cycling, and moisture that can loosen connections, shift terminations, or damage components. So point-to-point checks, connection torque verification, insulation resistance, and functional tests that were passed at the factory often must be re-run after delivery to confirm the module survived transit intact. This is not redundant effort; it is a critical quality assurance measure that helps maintain product integrity and performance as assembly shifts off-site, and it should be planned and budgeted upfront rather than addressed through possible field failures.

The remaining on-site scope, then, is fundamentally about integration and verification: fault-current contributions from parallel sources, protective-device coordination across module boundaries, grounding continuity across physically separated positions, post-shipment re-testing, and the behavior of campus-level cooling feeding factory-plumbed liquid-cooled racks. Repeated, identical modules help but each expansion phase can change the system’s fault and coordination picture, so integration testing must be treated as a recurring activity, not a one-time closeout.

How collaboration changes

This reshapes who commissions what, and when. The commissioning agent’s involvement now reaches back into the factory—witnessing factory acceptance testing, validating that what leaves the plant matches design intent, and documenting a baseline that on-site teams can test against after delivery. Because integrated systems testing is split across factory and field, the commissioning provider becomes the thread of continuity between the two, ensuring nothing falls into the gap between “passed at the factory” and “verified on site.” Module manufacturers effectively own the first level of validation; the independent commissioning provider owns the system-level truth that only emerges once modules are connected and confirms that truth still holds after shipment. Operations teams benefit from being present during on-site integration testing, seeing how the assembled facility behaves and how it signals problems, so they inherit a facility they understand rather than a stack of documentation. Commissioning becomes continuous and distributed across factory and field, which only works when those parties are coordinated deliberately from the start.

How Salas O’Brien can help

Modular delivery rewards owners who treat design and commissioning as continuous, integrated disciplines rather than sequential handoffs. Salas O’Brien brings that perspective across the full lifecycle of data centers, from early design through factory testing to on-site integration and turnover.

Working with Salas O’Brien provides:

Design for integration – We design the module and the building together. LOD 400 BIM, DfMA, VDC, and CFD help resolve clearances, transport paths, structural interfaces, connections, and thermal performance before fabrication. The module should fit, connect, and perform before it ever reaches the site.

Factory-to-field commissioning – Our teams witness factory acceptance testing and carry that validation through to on-site integration testing, so factory-tested becomes genuinely system-coordinated.

Early, collaborative delivery – We bring fabricators, contractors, and operators into design when their input creates the most value, and give operations teams a facility they helped shape.

To discuss how modular design and commissioning can compress your delivery timeline without compromising reliability, contact one of our data center experts below, or email us at [email protected].

For media inquiries on this article, reach out to [email protected].