The Five-Minute Problem Reshaping Data Centre Power Design

As AI workloads continue to reshape data centre infrastructure, much of the attention has focused on compute capacity, GPUs and cooling systems. Alongside those developments, changes are also taking place within the power layer that supports them.

The move from traditional enterprise workloads to high-density AI environments is placing new demands on power systems, particularly uninterruptible power supply (UPS) infrastructure. Components that have historically operated with relatively stable requirements are now being assessed against different performance expectations as rack densities increase and power demand rises.

In this interview, Dr Zijie Tang, General Manager at Huaxin, a Gerchamp Group company, discussed how AI is influencing battery requirements, why established technologies are facing new pressures, and how operators are rethinking the role of energy storage within modern data centre environments.

Why AI Is Changing Data Centre Power Requirements

The shift is already visible in infrastructure planning.

“Ten years ago, a typical rack might have been around 5 kilowatts,” Tang said. “Now, with AI and large model training, that’s increased to 100 kilowatts and beyond.”

UPS modules have evolved alongside these changes. Battery systems, however, are operating under a different set of constraints.

Data centre UPS deployments are typically designed for short-duration, high-power output, often within a five-to-ten-minute window. As power requirements increase, battery performance characteristics become more visible within the overall infrastructure design.

Revisiting Established Battery Technologies

Lead-acid batteries remain widely deployed across the sector, reflecting their long operational history and established safety profile.

Within AI-driven environments, however, discharge characteristics are receiving greater scrutiny.

“The capacity is there,” Tang explained, “but it cannot be delivered fast enough.”

In high-rate discharge scenarios, battery performance can differ significantly from nominal capacity ratings. Operators often compensate through oversizing, increasing the number of batteries deployed to achieve the required output.

That approach can influence multiple aspects of facility design, including floor space utilisation, installation requirements, and long-term operational planning. Repeated high-rate discharge cycles may also affect service life, increasing replacement frequency over time.

The Significance of the Five-Minute Window

At the same time, backup requirements have continued to evolve.

“In the past, systems were designed for 20 or 30 minutes of backup,” Tang noted. “But today, around five minutes is usually enough.”

For many facilities, the primary objective is to provide sufficient time for generator startup, load transfer or controlled shutdown procedures. Within that operating model, the ability to deliver power quickly becomes increasingly important.

The discussion moves from total stored energy toward power density, response characteristics, and efficiency during short-duration discharge events.

Safety Considerations in High-Density Environments

As operators evaluate different battery technologies, safety remains a central consideration.

Lithium-based systems offer advantages in energy density and performance, while also introducing additional discussions around thermal management and failure containment.

“Safety is always a key concern in data centres,” Tang added. “If something goes wrong, it’s not just a battery issue — it can affect the entire facility, business continuity, and even data security.”

Tang pointed to intrinsic safety as an increasingly important design consideration, where safety characteristics originate within the chemistry itself rather than relying solely on external mitigation systems.

Within that context, nickel-zinc batteries are attracting attention for their use of non-flammable aqueous electrolytes alongside high-power discharge capabilities.

The focus, Tang argued, is increasingly moving toward technologies that align safety, performance, and operational simplicity within the same system.

Infrastructure Design in Transition

Many of the technical challenges associated with AI infrastructure are already well understood across the industry.

“It’s not that operators don’t understand the technology,” he said. “But many decisions are still based on old habits.”

Practices such as oversizing and large capacity buffers were developed around the characteristics of previous battery technologies. As workload requirements change, those assumptions are being reassessed alongside wider infrastructure design decisions.

The conversation increasingly extends beyond batteries themselves and into the relationship between compute density, power architecture and facility design.

Defining the Next Generation of Battery Systems

Looking ahead, Tang identified several attributes that are shaping future battery development:

• Intrinsic safety
• High power density and fast response
• Optimised lifecycle cost
• Strong environmental performance

“There is no perfect battery,” he explained. “It’s always a trade-off. The goal is not to maximise one parameter, but to find the right balance for the application.”

Across the discussion, the battery appears less as a supporting component and more as part of the wider architecture challenge created by AI infrastructure.

Rack densities continue to rise, backup windows continue to shrink, and the assumptions that shaped UPS design for much of the last two decades are being revisited as operators look for greater power density, faster response, and more predictable performance.

In that environment, the conversation around batteries increasingly sits alongside discussions of cooling, compute, and power distribution as part of the same infrastructure system.