Data Center Site Selection: Why Power Defines Where You Can Build

Data center site selection has entered a new era defined by constraint instead of optimization. As power availability tightens and timelines stretch, the ability to secure reliable power has become the factor determining where new capacity gets built.

• Power access is now the primary gating factor in site selection, reshaping how developers evaluate markets.
• Growth is shifting toward power-advantaged regions, while legacy hubs risk losing relative share due to grid constraints.
• Time-to-power gaps between developers and utilities are widening, pushing operators to prioritize speed and certainty.
• Onsite generation is emerging as a core strategy, enabling faster deployment and expanding the pool of viable sites.

How Site Selection Has Changed

Data center site selection used to be a balancing act: Developers weighed land costs against tax incentives and fiber access against labor availability. Power was always part of the equation, but it was rarely the deciding factor.

That has changed.

In 2026, power availability has become the constraint that defines when a project moves forward — or if it moves at all. As AI-driven demand accelerates and data center campuses scale toward gigawatt levels, the grid is struggling to keep pace. Capacity is tightening, interconnection timelines are stretching, and developers are finding that the sites they want to build on can no longer get power in a reasonable amount of time.

According to Bloom’s 2026 Power Report, power access is driving a fundamental shift in where data centers are being built and how developers approach infrastructure planning; many developers now select sites based on power and build the rest of their decision framework around that.

Read on to learn how that shift is reshaping data center site selection, what it means for hyperscalers and colocation developers, and how a power-first strategy is changing where and how new data centers are built.

The New Geography of Data Center Development

Instead of expanding across established hubs, data center development is concentrated in regions where large blocks of power can be quickly secured. Per Bloom’s 2026 Power Report, this is already spurring a change in geographic capacity.

Texas stands out as the most prominent example. By 2028, the state is projected to exceed 40 GW of data center capacity (nearly 30% of U.S. total demand), which represents a dramatic increase in market share as developers prioritize regions with accessible power infrastructure. The Southeast is also gaining momentum — Georgia, in particular, is emerging as a growth market, with its market share projected to rise to 75%.

At the same time, a number of legacy and secondary markets are losing ground. California, Oregon, Iowa, and Nebraska are projected to lose 50% or more in market share in the next three years. Much of this drop can be attributed to the permitting complexity in those states, along with tighter power availability and longer interconnection timelines.

Even mature markets like Northern Virginia continue to grow in absolute terms, though they are declining in relative share as grid constraints make it harder to deliver power at the pace new development requires.

In addition, a broader set of emerging markets is gaining ground. The rest of the U.S. is collectively poised to increase its share by more than 20% as developers extend their siting flexibility in pursuit of power access. These markets may lack the network density and ecosystem maturity of Tier 1 locations, but they offer something extremely valuable: the ability to secure power without multi-year delays.

The data reflects a change in how markets compete. While incentives, land availability, and connectivity still matter, the defining characteristic of a successful data center market is how quickly and reliably it can deliver power at scale.

The Time-to-Power Problem

If power availability is changing where data centers get built, time-to-power is the constraint driving that shift. More than half of developers report that securing power has become more difficult over the past 12 months. At the same time, a gap has emerged between what developers expect and what utilities can deliver, as utilities estimate power will take approximately 1.5 to 2 years longer than hyperscalers and colocation providers expect.

This disconnect is most visible in major data center hubs. In Northern Virginia, the Bay Area, and Atlanta, the divide between expected and actual time-to-power has widened over the last six months, driven by permitting delays and extended interconnection queues — and, in many cases, aging transmission infrastructure. The result is growing uncertainty at the earliest stages of development. Projects that once relied on predictable grid timelines are now facing delays that can stretch on for years, risking both construction schedules and investments.

Power is no longer a downstream dependency that will be secured after a location is chosen. It is now the first design question to be solved, and increasingly, developers are looking beyond the traditional grid to do it.

Key Site Selection Criteria in 2026

Power availability may define whether a site is viable, but it does not operate in isolation. While building a data center location strategy, every major site selection factor is being considered based on how it supports or constrains access to power.

1. Power Availability and Interconnection

Developers are prioritizing markets based on the ability to secure large blocks of power — often 100 MW or more — within a defined and reliable timeline. They look for:

• Total grid capacity.
• Interconnection and queue depth.
• Utility responsiveness.
• Historical performance of large-load connections.

In constrained markets, onsite generation is becoming a valuable alternative, either as a bridge to grid power or as a long-term solution that reduces reliance on interconnection timelines.

2. Land and Physical Infrastructure

Site size and physical characteristics are increasingly tied to power strategy. As campuses continue to grow, developers require ever-increasing acreage for buildings, substations, and transmission connections, as well as — if they opt for it — onsite generation.

Proximity to existing infrastructure like roads, water, and fiber can accelerate timelines, while topography, flood risk, and seismic conditions can complicate development and eventual power delivery.

3. Water Availability

Access to water  is essential as data center requirements scale. Gigawatt-scale campuses require a great deal of cooling, which can limit site viability in arid areas like Phoenix or Las Vegas. On top of that, developers face increasing regulations around water permits.

However, the way power is generated and distributed also plays a role. More efficient electrical architecture and lower heat generation sources, like fuel cell technology, can reduce overall cooling demand, which in turn eases pressure on water resources.

4. Permitting and Regulatory Environment

In many regions, the complexity and duration of permitting processes are directly tied to the power infrastructure being deployed. Projects that depend on large-scale upgrades or traditional backup systems may face longer review cycles, while projects deploying lower-emission, modular technologies may streamline approvals.

Almost equally importantly, due to their eco-friendliness, lower-emission technologies may improve community acceptance.

5. Incentives and Total Cost of Power

Rather than focusing on just the electricity price, developers are considering the total cost of power over time. They include in their deliberations the potential financial impact of delays, as well as infrastructure buildout. In some cases, a lower-cost market may be bypassed if it can’t deliver power within the required timeframe.

State and local tax incentives for data center development still matter, but they are increasingly not enough to outweigh power availability or time-to-power. Here, onsite generation can offer a predictable, multi-year pricing structure while reducing exposure to grid volatility, making it very attractive to developers.

6. Connectivity and Network Infrastructure

Historically, proximity to fiber and end users dictated site selection, particularly for latency-sensitive workloads. This is still important, but the rise of AI workloads (especially inference) has led to greater geographic flexibility.

Developers are increasingly willing to trade some degree of network proximity for access to power, particularly for large-scale campuses where energy demand outweighs latency sensitivity. Connectivity remains a key factor, but it’s now regarded within a power-first framework instead of as the primary driver of location decisions.

Building a Power-First Site Selection Framework

In 2026, leading developers are adopting a power-first framework, one that treats power availability and delivery timelines as the starting point for every site evaluation.

Step 1: Define Power Requirements and Timeline

Developers are defining not just total load requirements, but what that power needs to be available (and at what level of reliability). As campuses scale toward hundreds of megawatts — and increasingly towards gigawatt levels — these requirements are becoming more demanding.

This step is the baseline for every decision that follows: How much power is needed, how quickly it must be delivered, and what level of load-following capability the site can support.

Step 2: Screen Markets by Power Availability

Developers must target markets that can realistically support them. That means evaluating grid capacity, interconnection timelines, and utility track records. As the 2026 Power Report indicates, developers are increasingly prioritizing markets based on speed and certainty of power delivery, rather than traditional resources.

Regions where large blocks of power can be secured quickly have the advantage here. Markets that cannot meet that criteria are often eliminated early in the process.

Step 3: Evaluate Onsite Power Options

In many cases, grid access alone is no longer sufficient to meet development timelines. Developers are now increasingly evaluating onsite generation as part of the initial site selection process — not as a fallback, but as a core component of the power strategy. A growing share of operators expect to deploy fully onsite-powered campuses by 2030, highlighting a broader shift toward long-term integration of onsite power.

This step allows developers to expand the pool of viable sites, including locations where grid access is limited but onsite generation can provide faster or more reliable time-to-power.

Step 4: Layer in Secondary Criteria

Once a group of markets has been established, developers can look at more traditional site selection factors. Land availability, water access, permitting complexity, and connectivity are all evaluated, but they are within a set of markets that have already passed the “power screening,” so to speak.

Step 5: Model Total Cost and Delivery Scenarios

The final step compares how different strategies perform over time. Developers model multiple scenarios — grid-only, hybrid, fully onsite configurations — to understand the trade-offs between cost and timeline. This includes examining not just energy pricing, but the financial impact of delays, along with infrastructure requirements and long-term scalability.

In this framework, the “best” site is not necessarily the one with the lowest upfront cost. It’s the one that can deliver power at the required scale within the required timeframe.

How Onsite Power Generation Changes the Site Selection Equation

Onsite power generation is shifting from a contingency plan to a core component of data center development. The share of developers expecting to operate fully onsite-powered campuses by 2030 has risen sharply in the last 12 months, with roughly one-third of data centers now expected to rely on onsite power as a primary energy source.

Onsite generation removes one of the biggest constraints in the development process: the interconnection queue. By generating power at the site, developers can bypass multi-year delays associated with grid updates and transmission expansion, compressing time-to-power from years to months and enabling faster project timelines.

It also expands the universe of viable locations. Sites that previously would have been ruled out due to limited grid access can now support large-scale developments, allowing operators to take advantage of favorable land and permitting conditions.

In addition, onsite power integrates more directly with next-gen data center architecture. As facilities move toward higher-density, more modular designs, locally generated power can reduce transmission losses and simplify distribution, improving overall system efficiency.

Within this landscape, fuel cell technology is emerging as a leading option for onsite generation. Developers are increasingly considering solutions that offer short deployment time lines, lower local emissions, modular scalability, and alignment with long-term sustainability goals — all of which are becoming critical factors in both site selection and project approval.

The Future of Data Center Site Selection

The next generation of data center development will be defined by how effectively operators solve for power. As campuses grow toward gigawatt sizes, and AI workloads continue to drive demand, power can no longer be treated as something to be secured after a site is selected. Instead, it must be designed into the project from day one.

Geographic diversification is accelerating as developers expand beyond traditional hubs in search of better markets. Onsite generation will become a standard component of data center architecture, not just a supplementary solution, and power strategy itself will emerge as a competitive differentiator, shaping which operators can build and scale further.

The definition of a “good” site will continue to evolve. It will not be the location with the lowest cost or strongest incentives; it will be the one that can deliver reliable power at the required scale and within the required timeframe.

The implication is straightforward. In the AI era, site selection is no longer about where to build. It’s about when you can get power — and how you plan to deliver it.

To learn more about Bloom Energy’s data center power solutions, contact us today.

View citations