Higher-density AI infrastructure is changing not just power architecture, but everything needed to support it, from testing and commissioning to field services and ongoing operations.

In a recent DCD>Talks episode, Nick Funyak, field services manager at Starline, a brand of Legrand, explains how AI is forcing a redesign of power distribution, making testing, commissioning, and operational services an increasingly important part of data center deployment.

AI is forcing a redesign of power distribution

AI is driving a rapid increase in rack densities, fundamentally changing how power is distributed through the data hall.

"We went from doing 17kW, 20kW, 30kW racks, whereas now we're seeing anywhere upwards of 50kW all the way up to 150kW, which is a humongous jump," says Funyak.

Supporting those higher rack densities require a much higher capacity busway installed directly into the white space above the server racks. That shift creates new engineering challenges around headroom, cooling, and physical integration.

"The busduct was always a gray space type of equipment, whereas now it is in the white space," Funyak explains. "You need a lot more headroom above the racks. You need to be able to cool that busway and cool that power consumption inside that data hall."

Services become part of the infrastructure

Higher-density AI infrastructure is also changing the role suppliers are expected to play. According to Funyak, operators increasingly want partners that can support deployment throughout the project lifecycle rather than simply delivering equipment.

That shift is being driven in part by the industry's ongoing labor shortage, with operators struggling to find enough skilled electricians to support rapidly expanding AI deployments.

"One of the guys there told me that they had 1,000 union electricians on their payroll. He said, 'Nick, I need another 400 to 500,'" says Funyak.

In response, Starline has significantly expanded its service capabilities.

"We have increased our service team over the last four years... just over 900 percent," says Funyak. "That comes from customers asking us to not just supply material but also become a better partner and take further ownership."

For Funyak, services are becoming just as important as the hardware itself, with customers increasingly looking for suppliers that can provide testing, commissioning, and ongoing operational support alongside power infrastructure.

Commissioning becomes an integrated process

As AI data centers adopt higher-capacity busway and liquid cooling, traditional testing and commissioning methods are no longer sufficient.

"In the past, we used to come and do testing with air-cooled load banks inside the data hall," says Funyak. "Now, being that a lot of these things are switching from air-cooled to liquid-cooled, it creates some challenges."

Commissioning increasingly requires either liquid-cooled load banks or relocating conventional air-cooled load banks outside the data hall. That means longer cable runs, additional security considerations, and more detailed planning before testing can begin.

Rather than testing power and cooling systems independently, operators are increasingly commissioning them together to reduce cost and deployment time. While that integrated approach can reduce costs and accelerate deployment, Funyak notes that it also increases complexity. A problem with either the electrical or cooling system can require the entire commissioning process to be repeated, making planning and coordination more important than ever.

More complex testing also means more preparation before commissioning begins. Funyak says operators now need to verify upstream breakers, account for much longer cable runs, and plan how liquid cooling systems will be connected, drained, and refilled during commissioning. While AI deployments have made testing more demanding, he argues those challenges can largely be overcome through careful planning.

Power and cooling converge

Looking ahead, Funyak expects AI to continue driving innovation in both power and cooling infrastructure as rack densities continue to increase.

"It's going to generate a lot of heat," he says. "The challenge moving forward is how do we continue to scale in density and pack as much power as we can into one small data hall and do it efficiently enough that we can cool that without having any potential problems."

As AI deployments continue to scale, testing and commissioning are evolving from standalone electrical activities into integrated services spanning power, cooling, and ongoing operations. For operators, ensuring infrastructure is ready for increasingly demanding AI workloads will require not only new hardware, but also new approaches to delivering and supporting it.

Watch the full DCD>Talks episode with Starline’s Nick Funyak here.