Viewpoint: AI Is Changing the Engineering Playbook for Data Centers

The following Viewpoint is written by Mark Knipfer, who leads data center services at Integrated Environmental Solutions

The rapid growth of artificial intelligence is triggering one of the fastest infrastructure buildouts the data center industry has ever seen. Hyperscalers, cloud providers and colocation operators are racing to add capacity to support increasingly compute-intensive AI workloads.

For the construction and engineering community, this wave of development brings new challenges. Data centers have always been technically complex facilities, but the scale and speed of AI infrastructure deployment are forcing project teams to rethink some long-standing assumptions about how these buildings are designed and delivered. This applies not only to new builds, but increasingly to existing facilities that must be adapted to support higher-density, AI-driven workloads.

Power Availability is the New Constraint 

Perhaps the most significant shift is that power availability is increasingly becoming the primary constraint on development.

In many markets, utilities are struggling to keep pace with the demand created by large-scale data center projects. These facilities can require hundreds of megawatts of electrical capacity, and interconnection timelines are stretching as grid operators work to expand generation and transmission infrastructure. For developers, this means that securing power—and bringing that power online quickly—has become one of the defining challenges of new projects. For operators of existing facilities, it also raises a parallel question: how to increase usable capacity within current power constraints.

This pressure is changing the economics of data center design. Decisions that affect energy demand, cooling performance and infrastructure efficiency now have direct implications for how quickly facilities can be delivered and how much capacity can be deployed in a given location. In practice, those decisions can also determine whether projects move smoothly from design into construction, or encounter costly redesigns once infrastructure limitations become clear.

Cooling Strategies Are Rapidly Evolving

At the same time, AI workloads are driving significant changes in data center cooling.

High-density computing environments built around GPUs and specialized AI accelerators generate far more heat than traditional enterprise workloads. Rack power densities that were once considered exceptional are becoming increasingly common in AI clusters.

As a result, developers and engineering teams are having to evaluate a wider range of cooling approaches. Conventional air-based systems remain common, but many new facilities are exploring direct-to-chip liquid cooling or hybrid approaches that combine air and liquid infrastructure. For retrofit projects, these choices are often constrained by existing plant, space limitations and operational continuity requirements, making evaluation even more critical.

Each strategy introduces different design considerations. Liquid cooling can improve heat removal for high-density racks, but it also introduces additional infrastructure requirements such as coolant distribution networks, heat exchangers and modified heat rejection systems. Hybrid approaches may offer flexibility but require careful integration with existing facility infrastructure.

These decisions are not purely mechanical design questions, they influence power consumption, water use, operational resilience and the long-term scalability of a facility.

Engineering Decisions Are Increasingly Interconnected

What makes these choices particularly challenging is that they are deeply interconnected.

A change in cooling strategy can affect electrical infrastructure requirements. Power constraints may influence equipment selection. Heat rejection strategies may depend on regional climate conditions or water availability. In retrofit scenarios, these interdependencies are amplified, as new systems must be integrated with existing infrastructure that was not designed for today’s operating conditions.

Historically, engineers have often evaluated these systems using peak-load calculations—designing facilities to perform under the most extreme operating conditions they might encounter. And while these calculations remain essential for ensuring reliability and redundancy, peak scenarios represent only a small fraction of how data centers actually operate. 

Most of the time, facilities run under partial-load conditions, as IT workloads fluctuate and environmental conditions change throughout the year. Infrastructure systems that appear efficient at peak load may behave very differently during everyday operation.

For engineering teams responsible for delivering reliable and efficient facilities, understanding these dynamics is becoming increasingly important.

Evaluating Infrastructure Earlier in the Design Process

Another defining characteristic of the current data center expansion is the pace of development.

The demand for AI computing capacity is compressing project timelines. Developers are under pressure to deliver new facilities as quickly as possible, often while navigating uncertain power availability and evolving technology choices. At the same time, operators are looking to retrofit existing facilities to extend their useful life and support new workloads without waiting for new capacity to come online.

This makes early-stage evaluation of infrastructure strategies particularly valuable.

By examining how different cooling architectures, equipment configurations and infrastructure systems interact, engineering teams can identify potential performance issues before construction begins. Early analysis can also help developers understand the trade-offs between competing design options and make more informed decisions about which strategies best align with project goals.

For complex facilities such as AI data centers, this type of upfront evaluation can reduce risk during later stages of design, construction and commissioning.

A New Phase for Data Center Development

Data center construction has always required close collaboration between developers, engineers and contractors. As AI infrastructure expands, that collaboration is becoming even more important.

Power availability, cooling performance and facility scalability are no longer isolated design considerations, but interconnected challenges that shape how projects are planned and delivered, whether for new developments or the transformation of existing assets.

For the engineering and construction community, the rapid rise of AI infrastructure represents a new phase in data center development. Meeting the demand for next-generation computing capacity will require project teams to evaluate infrastructure decisions more holistically and earlier in the design process.

As these facilities continue to grow in scale and complexity, this more integrated approach to engineering will become essential. The teams that succeed will be those able to evaluate infrastructure decisions earlier, understand how systems interact under real operating conditions, and translate those insights into designs that can move confidently from concept to construction.

Mark Knipfer leads data center services at Integrated Environmental Solutions, where he works with engineering teams and developers to evaluate performance and infrastructure strategies for high-performance facilities.