Data centre automation is becoming a critical construction discipline

Share this article

As AI data centre campuses grow, controls engineering is becoming central to construction delivery and operational resilience. Three systems integrators explain why developers can no longer leave automation decisions until commissioning.

A controls engineer arrives at a data centre construction site to commission an air handler already installed on the roof. The equipment cannot do what the specification requires, but changing it now would mean undoing work that has already been completed. The problem could have been identified before procurement, had the people responsible for making the system operate been involved earlier.

Jason King, Vice President of Data Centers at RoviSys, encountered exactly this situation. “I was on-site commissioning an air handler on the roof, and it wasn’t going to do what the specs needed,” he recalls. “I realised that immediately, but I also realised immediately the amount of people involved in the decision to get that thing bolted on the roof who did not catch that problem that I caught was a lot.”

At Rockwell Automation Fair, King joined Dan Ryan, Partner at BW Design Group, and Chad Deviller, Data Center Division Manager at Prime Controls, to discuss why controls specialists are moving towards the centre of data centre construction. As projects become larger and delivery schedules tighten, decisions about automation can no longer be treated as finishing work.

Building controls become industrial infrastructure

The systems keeping today’s data centres operating bear little resemblance to the relatively simple environmental controls familiar from older commercial buildings. Cooling is more demanding, while the coordination of equipment and electrical infrastructure imposes greater requirements on the automation architecture. A failure in one part of the system can have consequences well beyond that piece of equipment.

“Legacy data centres used to be commercial outfits. They weren’t that advanced. They had very rudimentary controls and just flooded a space with a lot of cool air,” King explains. “That model of the data centre does not exist anymore. It’s been gone for quite some time.”

Ryan sees the same transformation in the level of engineering expertise now required. “What used to be traditional, almost hotel room type HVAC controls has now evolved into very complex building automation systems, and layered on top of that is very complex electrical power monitoring systems,” he says. “You have massive amounts of data that you’re now collecting.”

These requirements have raised the profile of Division 25, the construction discipline covering integrated automation. Integrators must make equipment from multiple suppliers function as a coordinated system, often across different communications platforms. They also have to question whether a proposed control architecture is suitable for a mission-critical facility.

King says RoviSys has become willing to walk away when a client insists on a conventional commercial controls approach that cannot meet those requirements. “We’ve gotten pretty aggressive and told them, ‘Hey, you want to put a DDC style commercial system in this data centre? It’s not going to work. I’m not going to do that project.'”

The problem with arriving last

On many construction projects, the systems integrator still arrives after equipment has been selected and major design decisions have become difficult to reverse. That sequence leaves specialists responsible for making the facility work without giving them enough influence over the choices that determine whether it can.

“In the construction cycle, the SIs are typically the closers,” Deviller says. “We come in last with this scale and this speed to market that everybody’s pushing for. They’re asking a lot more of us than they would have traditionally four or five years ago.”

The panel suggested that experienced developers are beginning to change this practice. King says controls contractors are increasingly brought into the project alongside the major electrical and mechanical trades, rather than being asked to solve integration problems after installation. “The industry has recognised that the control system is as important as the other tier one trades,” he says. “We’re starting to be brought in earlier and earlier now.”

Early engagement can expose mismatches between specifications and equipment capabilities while alternatives are still available. Deviller argues that integrators can also bring knowledge accumulated on previous projects into engineering reviews, reducing the likelihood that problems will be rediscovered during commissioning. This is especially valuable when ambitious schedules leave little contingency for replacing equipment or redesigning connections.

Standardising the hyperscale campus

The pressure to repeat proven designs is growing as developers plan campuses with multiple buildings and unprecedented power requirements. Ryan referred to projects approaching 900MW, where staffing every task entirely on site becomes increasingly impractical. Offsite assembly and modular packages can ease construction pressures, provided the equipment arrives ready to integrate into a consistent controls strategy.

“The idea that we’re going to have 10,000 contractors on site at a particular project is just not feasible,” Ryan says. “It’s all about the skidded systems, the modular systems, and really getting as much manufacturing offsite as possible.”

Consistency is harder to achieve when different contractors deliver successive buildings. Owners may accept variation in equipment where necessary, but they cannot afford incompatible approaches to monitoring and control across a campus they must operate as one infrastructure estate.

“If you have a gigawatt campus that’s going to consist of 20 buildings, it’s a very real scenario that you’re going to have Prime, Design Group and RoviSys building one, two, three,” Deviller says. “Well, the owner doesn’t want three different solutions on their campus. They want a consistent approach.”

The practical answer is to establish common engineering principles early enough to influence equipment procurement and modular construction. Standardisation should make installations repeatable without preventing engineers from accommodating genuine differences between buildings. The panel also highlighted the value of forward visibility for suppliers, allowing production capacity and critical components to be planned rather than ordered against an already compressed deadline.

Resilience is a design decision

An automation system can satisfy a commissioning checklist yet still be poorly prepared for disruptions once the data centre becomes operational. As cooling and power systems become more interconnected, fault tolerance depends on how equipment responds to changing conditions, not simply whether individual devices have backup capacity.

“We have to have a control system that can take a punch and survive,” Deviller says. “In the prior days, as Jason mentioned earlier, the commercial approach was, if a chiller goes down, we’ve got minutes to resolve that. We can send an operator to the floor and get that chiller back online.”

Deviller believes that approach is increasingly inadequate for AI infrastructure, where a delayed response may jeopardise service continuity. “The control system has to be able to not be reactive, to be predictive,” he says.

That expectation changes the questions developers should ask before construction begins. The control architecture must be assessed for redundancy and its ability to manage failures without depending on improvised operator action. Fixing weaknesses once a facility is live can be much more disruptive than confronting them during design.

Delivery speed meets engineering limits

The rush to bring capacity online has not eliminated the time required to engineer a reliable facility. King warned that schedules need to reflect equipment availability and the work required to integrate it, rather than starting with a commercial deadline and assuming contractors will find a way to meet it.

“Understand your business case first, because business cases for the data centre owners should drive a lot more decisions,” he says. “Make sure you have a schedule that’s reality based. Make sure you have equipment and components purchased that meets that schedule.”

Scarcity of experienced integrators gives those specialists less incentive to accept unrealistic programmes. “There are just not enough system integrators to go around, and we’re not going to play games,” Ryan says. “You’ve got to do it right, or we’re not going to participate.”

Data centre developers have concentrated enormous attention on securing power and accelerating construction. Neither achievement guarantees that completed infrastructure will perform as intended. The panel’s message was that engineering decisions about controls need to be made while they can still shape the facility, not after equipment has been installed and deadlines are closing in. In increasingly large AI campuses, a reliable automation architecture is part of what makes capacity deliverable in the first place.

Related Posts
Others have also viewed

Data centre automation is becoming a critical construction discipline

As AI data centre campuses grow, controls engineering is becoming central to construction delivery and ...

Britain is building a testbed for the next generation of AI computing infrastructure

The development of artificial intelligence infrastructure is increasingly dominated by established computing architectures, making it ...

AI data centres face a new race to turn infrastructure into revenue

The economics of artificial intelligence are changing how data centre developers measure construction performance. Completing ...

The biggest risk to enterprise AI may be infrastructure that is not ready

The transition from experimental artificial intelligence to production systems is exposing a weakness in enterprise ...