Complexity is becoming one of the defining characteristics of the AI data center era.

As operators add capacity, they face constrained grid connections, more dynamic workloads, and stricter uptime requirements. Reliable power is no longer just a matter of adding generation. It requires a coordinated approach across the power stack, from design and redundancy to monitoring and operational support.

In this DCD>Talks episode, David Burke, group service support director at Clarke Energy, a Rehlko company, explains why an integrated approach to power infrastructure is becoming essential to delivering the availability and resilience demanded by next-generation AI facilities.

AI changes data center power expectations

AI workloads place different demands on power infrastructure than traditional enterprise computing. The challenge is not simply how much power is needed, but how reliably it can be delivered in environments where grid access may be limited. Planned AI campuses are pushing development into locations where securing grid connections can be difficult, while the workloads themselves create sharp load changes that stress power systems.

“We're seeing the changes or the challenges in two areas,” says Burke. “One is this clear drive for more data center capacity... The secondary element comes in with the AI loads, which are creating significant load step changes.”

That combination is driving greater interest in on-site generation. Rather than simply supplementing the grid, these systems need to provide continuous power while coping with fast-changing demand. Burke points to the combination of gas-powered generation and battery energy storage as one way of balancing sustained power delivery with rapid response to changing loads.

“That's where some of the benefits of integrating battery energy storage with engines come in, in terms of that immediacy of response from a battery system that helps kind of fill those reaction gaps that any kind of mechanical generation equipment... would inherently [have],” he says.

As power systems become more complex, operators must think in terms of a wider energy ecosystem in which generation, storage, monitoring, and service support work together as a coordinated system.

Designing for availability

For Burke, five-nines availability, or 99.999 percent uptime, starts long before a site is operational. Decisions about generation capacity, redundancy, and system architecture shape how resilient a facility will be when equipment fails or demand shifts unexpectedly.

“With the best service delivery in the world, it isn't possible to compensate for a scheme that has been poorly defined,” he says. “The determination of the particular assets needed in terms of power generation, right from the outset, is really, really important.”

What Clarke Energy is now doing is applying decades of service experience to an environment where availability expectations have moved to a different level entirely. Rather than the standards common in more conventional industries, data center operators increasingly expect five nines across critical infrastructure. According to Burke, meeting those expectations requires a corresponding increase in investment across the support ecosystem.

“We're building on the existing infrastructure in terms of personnel, in terms of training, in terms of inventory, and localized inventory,” says Burke. “Ultimately, it is the same delivery, considering the same principles, just with a ramp-up of that investment to suit the expectations that we know come from the data center industry.”

Predicting problems before they happen

Maintaining availability also requires continuous visibility into asset performance. Burke says data-driven monitoring has become central to modern service delivery, allowing teams to detect faults, deploy engineers, and sometimes intervene remotely before issues escalate.

“This continuous connection to the assets, this ability for the assets to report fault conditions in real time, for us to be able to respond very quickly in terms of deploying people, or in terms of remote intervention to return a unit to service, has become an absolutely integral part of us delivering our service contracts,” says Burke.

The company is also using operational data to identify degrading performance before it leads to downtime.

“We are absolutely delivering on certain key focus areas within our engines or within battery energy storage to have a level of prediction of potential issues, allowing us to get in there and solve issues before they lead to downtime,” says Burke.

Burke acknowledges that continuous connectivity can raise cybersecurity concerns but says modern monitoring platforms are designed with those requirements in mind.

Global expertise, local delivery

Delivering high availability depends on more than technology. It also requires the people, processes, and infrastructure needed to support power systems over time. Burke argues that this requires a balance between global expertise and local execution, sharing lessons learned across regions while maintaining the local presence needed for service delivery.

“I think the global element really comes in within the kind of focus of a role like mine, which is about implementing best practice in terms of sharing lessons learned, in terms of capturing those core principles that lead to good service delivery,” says Burke.

However, he emphasizes that meeting availability targets requires more than centralized expertise. Any effective support model also depends on strong local investment.

“The local element is then that investment in the personnel, in the training of those people, and in the local inventory and tooling to allow that concept to be delivered at the more local level,” he explains.

The broader message is consistent throughout the discussion: high availability is not simply about choosing the right power generation, storage, or monitoring technologies. It depends on integrating design, operational visibility, and service support from the outset so that generation, storage, monitoring, and maintenance work as one coordinated system.

Watch the full DCD>Talks episode with Clarke Energy here.