Matilda Bailey has spent years at the intersection of networking and large-scale infrastructure, navigating the complex shifts in how the world’s digital backbone is deployed. As traditional hubs like Northern Virginia and Silicon Valley reach a state of saturation, her expertise in evaluating the trade-offs between dense urban interconnection and the untapped potential of rural land has become invaluable to hyperscalers. This conversation explores the evolving criteria for site selection, from the critical role of power grid stability and fiber routes to the nuanced social licenses required to operate in new markets. We delve into how the rise of AI and edge computing is rewriting the geographic playbook for data center real estate.
Northern Virginia and Silicon Valley have dominated development for decades due to established commercial norms. How do you evaluate the tipping point when power constraints or land costs outweigh these legacy benefits, and what specific metrics signal it is time to move into emerging markets like Louisiana or Mississippi?
The decision to pivot away from a primary hub like Northern Virginia isn’t made lightly because these areas offer predictable local governance and a density of interconnection that is hard to replicate. However, we reach a tipping point when the reliability of the grid is compromised or when power constraints begin to stall the time-to-market for new builds. In Pennsylvania, for example, we are seeing a 70 billion dollar race for data center development precisely because operators are looking for areas that can support large, continuous electrical loads over several decades. When land pricing and the complexity of permitting in a suburban hub begin to inflate the total build cost beyond a manageable threshold, emerging markets in Louisiana and Mississippi become attractive. We look for specific metrics such as the availability of long-haul fiber routes and the willingness of local utilities to streamline the delivery of massive power loads before we recommend a geographic shift.
Rural areas like North Dakota or Utah offer cheaper land but often lack dense interconnection. When moving into these less populated regions, how do you mitigate risks surrounding thin fiber routes and the coordination required with local utilities that may be inexperienced with hyperscale electrical loads?
Moving into rural regions requires a very hands-on approach to infrastructure development, as you cannot always rely on the “plug-and-play” environment of an established tech corridor. While it is true that fiber density can be thinner, certain states like North Dakota and Utah have become notable exceptions by proactively extending high-capacity fiber and creating diverse paths to ensure resilience. To mitigate the risk of utility inexperience, we prioritize early coordination with power providers to ensure they can sustain demand for the next 20 to 30 years. We often look at the feasibility of behind-the-meter generation, such as on-site solar or fuel cells, to provide a buffer against grid instability, even if these solutions aren’t yet the primary power source for most hyperscale sites. This coordination is foundational because a data center’s success depends on the grid’s ability to evolve alongside the facility’s growing energy needs.
Permitting timelines and local sentiment regarding water usage can stall a project before it begins. What strategies ensure a smooth “social license” in communities new to data centers, and how do you prioritize jurisdictional tax incentives against long-term operational costs like power demand charges?
Securing a social license in a new community requires a transparent strategy that highlights the shared benefits, such as job creation, infrastructure improvements, and a stable tax revenue stream. Residents in rural or previously non-industrial areas often have valid concerns about noise, environmental impacts, and the massive amount of water required for cooling, so early engagement is essential to address these anxieties before they turn into project-ending opposition. From a financial perspective, a lucrative tax abatement can significantly improve the initial economics of a project, but it should never blind an operator to high ongoing operational costs. We carefully weigh one-time incentives against long-term power rates and demand charges, as a site that is cheap to build but expensive to run will eventually become a liability. Finding a jurisdiction with experienced authorities who offer a predictable and fast-track permitting process is often more valuable than a tax break that comes with bureaucratic delays.
AI training and edge services have different performance profiles than traditional latency-sensitive applications. How do these modern workloads shift the balance between urban hubs like Denver and rural sites, and what steps are involved in aligning a specific geographic profile with a company’s technical needs?
The shift toward AI training and edge computing has fundamentally altered our geographic strategy because not every workload requires the microsecond latency provided by a hub like Northern Virginia. For latency-sensitive services, we still lean toward urban hubs like Denver or Boston, where proximity to high concentrations of end users and enterprises ensures peak performance. However, AI training workloads are often less sensitive to distance and can be placed in rural sites where land is abundant and power is more readily available for massive clusters. Aligning a geographic profile with technical needs involves a deep audit of the specific workload; if the data needs to be processed close to where it is generated, we look for suburban corridors, but if it is a compute-heavy training model, we prioritize the cheapest, most stable power grid regardless of its proximity to a city. This dispersion of workloads is making a single-location strategy obsolete, forcing organizations to build across a much wider set of geographies.
Coastal metros face hurricanes and earthquakes, while inland sites may struggle with extreme heat or grid reliability. When comparing rural Pennsylvania to established suburban corridors, how do you quantify the trade-offs between physical disaster risk and the long-term stability of the local power grid?
Quantifying these trade-offs requires a rigorous risk-and-resilience assessment that looks at both the physical environment and the man-made infrastructure. In coastal metros, the threat of seismic activity or flooding is a constant variable that requires expensive site preparation and hardening, whereas inland sites in Pennsylvania might offer lower exposure to those specific natural disasters. However, an inland site is only as good as its utility footprint; we must evaluate whether the local grid can withstand extreme heat or if the transmission constraints will lead to frequent curtailment. We use historical data to compare the reliability profiles of different utilities, looking for regions that offer competitively priced power with a track record of stability. Ultimately, the goal is to find a site where the long-term operational risks are manageable, ensuring that the facility remains functional even when the surrounding environment or grid is under stress.
What is your forecast for data center geographic diversity?
My forecast is that we will see a permanent move toward a much more fragmented and diverse geographic landscape where the dominance of a few legacy hubs continues to erode. While Northern Virginia and Silicon Valley will persist because of their incredible interconnection density, they will no longer be the default starting point for every hyperscale project. We are entering an era where states like Louisiana, Mississippi, and parts of Pennsylvania will become central to the digital economy, driven by the sheer necessity of power and land. As long-haul fiber routes continue to expand and edge services become more sophisticated, the distinction between “prime” and “secondary” markets will blur, leading to a more resilient and distributed global network infrastructure.
