Change is nothing new in the data center industry. AI is pushing power density and cooling demand beyond the limits facilities were built to handle, and the technology inside is evolving faster than the buildings themselves.

At the same time, developers are navigating site-specific constraints, environmental limits, and communities that have more say in what gets built than ever before. Designing for flexibility is no longer a preference; it’s a requirement.

PowerHouse is one of the companies challenging conventional thinking. By laying everything they know about data centers on the table and rearranging the pieces for the realities of today, the company seeks a balance that works for developers, operators, and crucially, the communities that host each individual project. In doing so, it aims to demonstrate that responsible environmental choices can be made without compromising performance.

PowerHouse’s senior director of data center development and construction, Ginger Phelps, has spent more than 25 years in industrial construction and heavy power utilities. Her unique perspective on the challenge is direct: the most successful projects aren't the ones with the most sophisticated technology; they’re the ones built to adapt.

Drawing on her experience, Phelps helps guide PowerHouse’s approach to water- and air-cooled data center systems. While many of the materials and technologies remain familiar, their application in modern data centers presents an entirely different challenge.

“I’ve been doing this for a long time, and in my experience, the most successful projects are the ones that are flexible and dynamic, with rigorous processes and a thoughtful approach to overall project execution. They can adapt to the changes that arise during a project and even after it is complete. That discipline ultimately resonates through the design that gets built,” says Phelps.

More power more problems

Over time, AI workload densities are getting higher and higher. With this newfound density, power, and revenue comes unprecedented levels of heat, reshaping data center cooling requirements. It’s a familiar trade-off: the more capability operators unlock, the more challenges they must solve.

As an industry, we are far more aware today of the impact water consumption can have on local communities, making thoughtful resource management just as important as cooling performance itself. The single-solution approaches that worked for traditional workloads are no longer enough.

While a one-size-fits-all approach may be quick and convenient on day one, the operational and financial risks can quickly rack up as workloads become more dynamic, bringing new cooling strategies to the fore. Phelps explains:

“What may have taken a pretty large footprint to have consumed that much energy is now being done in such a small space. Energy-efficient solutions that would have worked actually just don’t now. So we’re pivoting toward concepts that you wouldn’t have had to consider before.”

As data centers expand beyond traditional, oversaturated hubs into more rural or less-developed regions, constraints around water, staffing, and supply chains become more pronounced. Water availability itself may be limited. Regulatory pressures may, and rightly, prioritize residential needs ahead of commercial demand. And in more remote locations, simply finding the personnel required to operate and maintain complex water-cooled systems can be a challenge.

The water vs power trade‑off

The broader challenge is that every cooling strategy involves trade-offs. When was the last time you encountered a single, perfect solution with no caveats – one that was truly rinse and repeat? If such a solution exists, it is certainly rare.

“Building in Texas versus building in Virginia, there’s going to be dramatically different environmental challenges. It’s going to be drier and warmer in Texas than it is in Virginia, but for these systems, humidity and wet-bulb temperature are real challenges, and so we have to be very, very thoughtful of the spaces,” adds Phelps.

If you use less water, for example through air‑cooled, dry, or hybrid systems, you reduce water demand, but this often comes at the cost of lower chiller efficiency, higher power consumption, and larger physical footprints. On the flip side, water‑intensive systems that use evaporative cooling can deliver highly efficient heat rejection and lower energy use, but consume more water, raising overwhelming environmental and community concern. As Phelps emphasizes:

“If you’re only looking at footprint efficiency and power consumption, you’re not going to realize the full range of solutions available and your only options will be water-heavy systems, so it’s about accounting for a wider range of factors, including water consumption and community needs.”

The key, therefore, is involving experienced specialists from day zero, identifying difficult or disruptive infrastructure decisions from day one, and evaluating cooling strategies through a broader lens of lifecycle environmental impact, power availability, site conditions, and community expectations.

Day 1: Tomorrow’s technology hasn’t been developed yet

The speed at which AI hardware evolution is outpacing infrastructure capacity raises fundamental questions about data center design, particularly how to account for the inevitability that rising power densities will impact cooling strategies even before facilities are fully deployed.

At the same time, AI is very much still an emerging part of the data center business, and no one can say with absolute clarity how demand will progress, whether it will continue to grow as rapidly, stabilize over time, or eventually drop off.

“PowerHouse has done a great job of trying to stay cutting edge, working on projects and building for customers who are already designing and planning to implement technology that hasn’t even been developed yet. Projects start before technology is even conceived or thought of, and definitely not built or used.”

In today’s reality, design and construction lag behind compute, so it is no longer realistic to design infrastructure around fixed assumptions about future workloads or technologies.

Instead, the only viable approach is to design for flexibility and context. Cooling solutions, as part of this, must be adaptable enough to accommodate denser, hotter, and more demanding AI systems as they emerge over a facility’s lifecycle.

This is why flexibility is becoming the defining principle of modern data centers, making “designing for change” essential.

Creativity under constraint

Planning for flexibility in data center cooling increasingly requires a creative, context-driven approach. PowerHouse advocates designing around worst-case scenarios from day one. For example, what happens if the facility is running at maximum IT load on the hottest day of the year while simultaneously experiencing a power failure that takes primary pumps offline?

In practice, that level of resilience requires creative combinations of redundancy and infrastructure design, such as redundant power for pumps, chillers, equipment supported by additional UPS capacity, as well as on‑site water storage sized to sustain operations through extreme conditions.

“Reliability also depends on securing sufficient water availability from the local authorities, particularly when facilities rely on makeup water or are trying to reduce its use. We have a certain amount of water that’s available from the county at any given time, and we have to think about how long we could run. And so our tank is sized for that condition.”

Importantly, this does not mean operating backup systems continuously “just in case.” Rather, it means ensuring resilience is built into infrastructure when it’s needed most.

This is where creativity in planning becomes critical, particularly as operators navigate regulatory and environmental constraints alongside technical ones. Phelps points to closed-loop liquid cooling as one example of how new approaches can address these overlapping challenges. While still requiring careful consideration to balance energy and water use responsibly, closed-loop systems can offer significant efficiency gains when running at full load.

“It’s really impressive for us that the amount of energy, water, and power consumed is much lower than what you would have seen traditionally.”

In some cases, these systems can deliver substantially more computing power within comparable footprints while maintaining efficiency and still meeting the high cooling-performance standards customers expect under their SLA agreements.

Creativity also extends to water sourcing strategies. Such systems can become even more efficient when non-potable or greywater sources are prioritized over potable supplies.

While unsuitable for drinking, non-potable water is often perfectly suited for data center cooling. This provides a more environmentally responsible solution through water reuse, while also opening a potentially less contentious deployment pathway, as it does not compete directly with residential consumption or other higher-priority uses.

“Municipalities are looking for places where, if you have a different water source – whatever that source may be – as long as it’s not potable water, there may be opportunities in the future to help offset further. Then you don’t need as much power, and you don’t need as much potable water, which has generally been the public concern.”

Toward smarter stewardship

Flexibility in data center cooling comes with nuance. Ultimately, it is not about reinventing systems entirely, but about reworking and reintegrating existing approaches to suit different contexts and environments. Alongside this, success increasingly depends on understanding the place as much as the technology, and recognizing that the industry itself is still in flux.

“We’re definitely in a revolution of sorts when it comes to data center evolution, and as we continue to see that settle out, we’ll find a better balance of the right mix of water consumption, power consumption, density in the space,” summarizes Phelps.

The point is not that one technology is inherently better than another. Rather, data center operators should explore every viable option for improving efficiency and sustainability. Instead of relying on a single solution for every deployment, the buildings that stand the test of time require a flexible, context-based approach. Getting that right means staying curious, staying flexible, and resisting the temptation to build the same facility twice.

To learn more about PowerHouse’s commitment to responsible development, visit powerhousedata.com/sustainability.