The pace of AI adoption is forcing cooling conversations to include power and lifecycle performance across the deployment and delivery of data center infrastructure.

For Becky Wacker, VP of data center solutions at Trane, this evolution is reshaping the relationship between power, cooling, and compute performance.

Wacker is responsible for sales and fulfilment for Trane’s data center business, working closely with customers to understand their thermal management business needs. With her main scope being HVAC equipment, she understands just how the landscape has changed.

“It used to be easier when compute was steady, but now AI workloads are operating hotter and ‘spikier’ – it takes more planning because both cooling and compute are going to use power,” she said. “As we think about different workload densities at the chip, we have to think about how we can react faster. It now needs to be more dynamic.”

During the DCD>Talks episode, Wacker emphasized that the best AI facilities have been co-designed, rather than assembled. In the past, data centers were built in certain regions with favorable characteristics, but now the landscape is evolving to see data centers being built in many different locations with more variable characteristics.

“It’s a lot harder to ‘cookie-cutter’ design – it’s now about looking at power usage effectiveness (PUE), water usage effectiveness (WUE) and the type of power used on site,” she said. “We’re also seeing more on-site generation, either supporting the grid or being fully islanded on its own from a microgrid perspective.”

At Trane, she explained the company is improving its systems to consider how its customers are designing the data center as a whole and how each data hall ties together.

“We’re also thinking about efficiency because we can no longer just apply simple systems,” she added. “We now need digital tools to shift load and pull those different pieces together. A less optimized data center will impact the compute power at the chip.”

In a fast-changing industry, Trane has witnessed plenty of customers not yet ready to take the leap into using the newest technologies, Wacker noted. Clients are asking how they can be ready for changing demands and how the thermal loop can be tied into that.

“It’s becoming more of a strategic conversation about what the plan is and how it’s going to shift over time,” she added. “We’re trying to understand and pinpoint where different heat loads are going to land and how we’re going to extract them – largely moving toward liquid cooling.”

As priorities shift, however, there will be plenty of innovation moving beyond general liquid cooling, like direct-to-chip, into full immersion cooling strategies and beyond.

“That’s the fun part for our engineers – these strategic conversations with customers and clients are more future-oriented than before,” Wacker said. “We’re building strategies together to predict what’s coming and be ready to modify as that future comes to fruition.”

To achieve this successfully, the importance lies in keeping AI infrastructure efficient, resilient, and future-ready to confront changes. As the industry considers the data center lifecycle, Wacker emphasized the importance of having a team to co-innovate with.

“Where are we, and how do we design to make sure that retrofits are possible in the future as things change?” she questioned. “We also have to make sure that, with the spiky loads we see today, the way we’re running our data centers is as efficient as we designed it to be.

“We need to stay ahead of it and find issues faster than just waiting for something to fail.”

Click here to listen to the full DCD>Talks interview with Becky Wacker.