The US has a data center problem.

Half of planned US data center builds are currently being delayed or outright canceled. However, it’s not land or even capital constraints that are behind this. It’s power availability.

In these constrained markets, developers are increasingly facing power shortfalls that render many projects non-viable. However, the gap between power scarcity and project success is becoming increasingly bridgeable.

Securing sufficient electrical grid capacity is arguably the top priority for planned data center projects. Across the US right now, data centers represent a combined capacity of about 51GW. By 2028, new data centers will require an additional 44GW of capacity. The issue? The grid will only be able to supply 25GW by then. That is a significant gap between supply and demand.

By reducing a data center’s reliance on the grid, developers can convert stranded or stalled sites into viable deployments without waiting years for grid expansion. One approach is shifting 20 to 30 percent of cooling-related electrical loads onto alternative fuel sources.

The real constraint on data center growth is grid power

In many key markets, including Northern Virginia, Phoenix, Chicago, and Silicon Valley, interconnection timelines for large loads often extend three to five years or longer, creating a structural bottleneck in the power supply.

The reason is not a lack of funding or site availability, but often the grid interconnection capacity. Specifically, the utilities are increasingly unable to deliver sufficient power within the developers’ required timelines.

This unprecedented demand growth is being driven by AI training and inference workloads, hyperscale cloud expansion, enterprise digital transformation, and Edge or latency-sensitive applications. Despite billions in capital deployment and aggressive land acquisition, a growing number of projects are stalling before construction even begins.

The result is a growing class of development assets that are acquired, zoned, and even fully permitted, but cannot proceed due to insufficient available grid capacity.

While it is easy to think of these sites as failed developments, they are simply power-constrained assets. In such constrained markets, the first instinct is often to add on-site power, but in many cases, there is a simpler bridge to power when the shortfall is 20 to 30 percent.

Data center cooling is now a grid strategy problem

Data centers are not partially viable when underpowered. From IT load design constraints and power density targets for AI workloads to redundancy requirements and cooling system sizing, every kW and MW is needed for a specific purpose.

If a facility’s total power demand can be reduced by approximately 20 to 30 percent, the same site that was previously unviable due to existing grid capacity can suddenly be successfully developed. Accomplishing this, however, requires a significant paradigm shift. Instead of pursuing incremental efficiency gains on typical data center designs, developers should rethink where and how electricity is consumed within the facility.

Traditionally, cooling has been treated as an operational efficiency metric, measured by PUE (Power Usage Effectiveness). But when power is constrained, it becomes a direct determinant of project viability.

In a typical data center, cooling accounts for around 30 percent of the facility's total energy consumption. Reducing cooling load directly reduces:

• Total MW demand from the grid
• Required substation capacity
• Interconnection queue position risk

In other words, cooling is no longer a given overhead that shows up solely in efficiency metrics, but a main driver of grid capacity consumption that can be directly influenced, making previously unviable sites viable.

Reducing the demand for cooling on the grid improves site viability

If cooling is a primary driver of power consumption, it is also one of the most powerful levers available to developers.

Conventionally, the data center industry has chosen to optimize electrically driven cooling systems by incrementally improving efficiency through better airflow design, heat exchange, and compressors. Of course, these advancements matter. But they do little to address the real core constraint cooling poses: the total electrical demand.

A better alternative approach is beginning to take root. Instead of treating cooling as a load to be minimized, it can be redesigned to reduce dependence on grid electricity altogether. By shifting how cooling is powered, from electricity toward alternative fuel sources (like natural gas), developers can meaningfully reduce total grid demand without compromising performance or density targets.

This is far better than a marginal efficiency gain. Rather, it’s a structural change in how energy is supplied and consumed within the data center.

Consider this: For a 1MW IT load with a peak PUE of 1.4, the total facility load is 1.4MW, of which cooling and auxiliaries account for 0.4MW. By shifting cooling off-grid and onto alternate energy sources, it becomes possible to reclaim approximately 0.3MW (based on industry-average cooling loads). That represents an approximately 30 percent increase in new power available for reallocation to IT capacity.

As a result, cooling has now become more than an operational consideration. Instead, it is a strategic tool for expanding the range of viable sites in power-constrained markets, improving deployment timelines and unlocking capacity that can be better used for revenue-generating compute.

Importantly, a growing number of data centers are pursuing on-site generation to reduce grid dependency. In fact, 38 percent of data centers are expected to incorporate on-site generation by 2030, with 27 percent being fully powered by onsite generation by then. But on-site generation alone does not solve the challenge.

Leveraging microgrids and other distributed energy systems can help address power scarcity, but it does not eliminate the underlying issue.

The new competitive advantage: power efficiency at the design level

The data center industry is entering a phase in which success is no longer defined solely by access to capital or land aggregation. Instead, the critical differentiator is the ability to design facilities that fit within constrained grid realities without sacrificing compute scale.

The companies that solve the gap by reducing demand intelligently rather than waiting for supply expansion will unlock a generation of stalled infrastructure and gain access to the most constrained markets first. In today’s environment, grid-aware design is quickly becoming the most important site selection strategy.