The insatiable appetite for data centers raised concerns about an unsustainable future, and as demand continued to soar, the struggle to meet this rapid growth had become more pronounced than ever. However, the skyrocketing demand for generative AI integration now seems to be finding a local equilibrium, which can be seen from the postponements of numerous projects by leading tech giants.
It is still too early to say that an equilibrium between supply and demand has been reached, but two of the primary metrics that are indicative of market activity are rising vacancy rates and slowing pre-leasing activity. Although data center demand is by no means weak, compared to the pace of the last few years, frenetic demand has slightly withdrawn. Occupancy rates remain strong as companies look to expand their generative AI capabilities and move or expand their operations to hyperscale facilities. Northern Virginia, aka “Data Center Alley,” has stood as the unswerving global leader in data center power capacity. The current available power capacity in the state of Virginia totals approximately 6 gigawatts. At sub-1 percent to end 2024, and despite the nearly 2 gigawatts of power capacity under construction, the market’s vacancy rate is considered frictional and is expected to remain so without a significant uplift.
The footprint needed for AI-intensive computing is shrinking, increasing the risk of overbuilding.
In Northern Virginia, the pre-lease rate was 98.1 percent for year-end 2023 but has decreased to 93.5 percent in 2025 Q1. Across the country, a more uniform distribution of pre-leasing rates is emerging, ranging from 0 to 100 percent; previously, pre-lease rates were clustered more toward the higher end of the curve, with fewer large markets either fully or nearly fully pre-leased. Among the largest data center markets, only Northern California and Reno are fully pre-leased in 2025 Q1, while supply-constrained markets like Phoenix and Atlanta now only have about half of their space leased before completion as compared to a year ago. This is likely because tenants leased any available space due to supply shortages, prioritizing power availability over quality. As demand increased, new data centers were constructed with policy support and allowed new entrants into the top 10 markets, like Reno.
Selecting a location for data centers is crucial due to their need for cooling water and adequate power infrastructure. The current administration supports expanding data centers, creating new opportunities for site selection. AI growth, which depends on these centers, is a shared interest across political lines. To reduce the typically high construction costs, the Department of Energy is offering federal land already equipped with necessary amenities like power and water. Additionally, there are initiatives to fast-track permits in certain regions to build new power sources, as part of executive orders to enhance domestic energy production.
Recently, nuclear power has come into focus as a source of sufficient and reliable energy that does not rely on the main grid or draw energy from existing household and business needs. Due to recent advances in liquid cooling methods, it also means that data centers do not have to collocate with vast sources of water for cooling.
6
This expands location choice beyond the traditional set of land used for data centers, which includes privately owned and available tracts, but there are other obstacles. Building small modular reactors and microreactors will still take several years. Therefore, the transition to this independent energy source not reliant on the main power grid will require time.
Another risk that developers and investors must keep in mind is that technology is outpacing the useful life of the built structures that house data center operations. This is to say, the physical footprint needed for data centers is shrinking, increasing the risk of overbuilding. However, in the context of this property type, there is no risk at the moment of overbuilding power capacity. Robust demand endures given that data centers are still in a growth part of their business cycle and are expected to remain so in the medium term over the next five years, at minimum. The outlook for the sector is positive, and growth is likely to persist for the next decade, even if it begins to decelerate by the early 2030s.
Liquid cooling innovations reduce the need to co-locate near abundant water supplies.
At the same time, this raises concerns about obsolescence—not of data centers themselves but of the physical square footage that will be needed to store the equipment in the future. There has been a transition toward reducing the physical square footage required for facilities. Rather than demolishing existing structures and constructing smaller ones, which would be both financially prohibitive and wasteful, these properties have been upgraded with new, compact equipment that requires less space. These buildouts incorporate advanced technology to optimize their functionality with a smaller footprint. These advancements bring both advantages and disadvantages. From an operator perspective, less land and fewer construction materials mean more manageable development costs, expanded margins, and the potential to invest in more centers. However, for municipalities that have come to rely on this property type to reinforce their tax base that was decimated by obsolescence in retail—particularly regional malls—and certain office properties, this has the potential to once again cause a recalibration of tax revenue for counties and cities across the country that are dependent on data center property taxes, based on their current sizes, to support their commercial real estate tax income.
For municipalities that depend on data center property taxes to maintain their tax base—previously weakened by the decline of retail properties, particularly regional malls, and certain office spaces—the potential loss of this property type could lead to a recalibration of tax revenue. This adjustment would affect counties and cities nationwide that rely on these taxes to support their commercial real estate income. Local governments may need to explore alternative sources of revenue to compensate for these deficits.