Power, Water and Fiber: What Really Decides Where Next-Gen Data Centers Go
- Natalie Stipan
- 337 views
Cheap land no longer settles the question. Grid capacity, PUE, water access, fiber diversity and permitting timelines now decide where new data centers get built.

If you've ever wondered why a data center lands in a rural county with cheap land and no obvious tech base, the answer is rarely just the price of dirt. Land and energy costs still matter, but they're now the opening line of the site-selection conversation rather than the closing argument. Developers have to line up dependable power, enough water, diverse fiber, workable weather and a labor pool, and they have to do it at the same time.

AI, cloud and digital services have changed how those assessments get run. A site that looked like a bargain on paper can become expensive to operate once electricity prices climb or expansion gets blocked by the grid. My read is that the next generation of facilities will be chosen on a mix of technical, environmental, economic and permitting factors, and the cheapest acre often loses that contest.
Power Capacity Matters More Than the Rate Card
Electricity is the factor that shapes everything else, especially for AI-heavy workloads. Dataintelo estimates that data centers worldwide consumed about 415 TWh in 2024, roughly 1.5% of total electricity demand, and projects that figure will reach about 945 TWh by 2030 as AI and digital services expand.
Finding a utility that will sell you power isn't the hard part. A large site may need many hundreds of megawatts of stable capacity, plus substations, transmission lines and backup supply. To put the scale in perspective, a 200 MW data center running at an average utilization of 85% would consume about 1.49 million MWh a year.
A location with spare capacity can beat a cheaper-rate location, but the timeline is the catch. Transformer availability, queues for new power lines and the terms of long-term supply contracts can each push a project back by years. Teams in this position typically treat the grid connection schedule as a construction-critical path item, not a procurement detail.
Climate Drives Cooling Costs for Years
Temperature determines how much energy you spend pulling heat out of servers. Conventional cooling loses efficiency as ambient temperatures rise, while liquid cooling is better suited to the concentrated heat that modern AI accelerators produce. That shift changes which climates look attractive.
The efficiency yardstick is Power Usage Effectiveness, or PUE: total facility energy divided by the energy used by IT equipment. A PUE of 1.2 means the facility draws 1.2 units of power for every unit that reaches the computing hardware. Put a 100 MW IT load at a PUE of 1.2 and the site needs 120 MW in total. The same load at a PUE of 1.5 needs 150 MW. That 30 MW gap is the difference between two budgets, and it compounds every year the facility runs.
Climate is only one input. Humidity, cooling technology, server density and day-to-day operating practices all move the PUE number. So the best sites tend to combine a favorable climate with reliable infrastructure and acceptable environmental risk, rather than winning on temperature alone.
Water Is the Factor Teams Often Underweight
Water rarely makes the early shortlist, and that's a mistake. Many data centers rely on water-based cooling, and consumption climbs during extreme heat and heavy compute periods. Regions dealing with drought or competition between farms, homes and industry can impose limits that shape operations for years.
How much water a site uses depends heavily on its cooling design. A facility with a continuous 100 MW IT load consumes about 2,400 MWh of IT electricity per day, and the cooling demand attached to that load can translate into significant water needs. The exact figure depends on climate, equipment and design, so there's no single number to plan around.
Most companies now assess water stress at the watershed level instead of relying on rainfall averages alone. Several options can cut freshwater use, including recycled wastewater, closed-loop cooling and air cooling. Each has a cost. Those systems tend to draw more power, require more capital and add maintenance work, so water and energy decisions have to be made together.
Fiber, Latency and Regional Clustering
Data centers move large volumes of traffic to customers, businesses and other facilities, so connectivity sits near the top of the list. Latency, measured in milliseconds, is especially important for finance, industrial automation and transportation. For some time-sensitive services, a difference of just 10 milliseconds can matter.
Proximity to internet exchanges, cable landings and cloud hubs helps, but distance alone doesn't guarantee good performance. Fiber diversity, traffic patterns, routing and the availability of alternative paths all affect how resilient a site really is. A location with two genuinely separate fiber routes is worth more than one that looks close on a map but depends on a single trench.
Dataintelo puts the global next-generation data center market at $163.2 billion in 2025, projecting growth to $510.6 billion by 2034, a compound annual growth rate of 13.5% over the forecast period. Those are the firm's figures, and they point to sustained demand for compute, network capacity and sites that can host AI and cloud workloads. As investment grows, the sites that win will be the ones whose grid, water and connectivity can keep up.
Clustering has its own logic. An established hub often offers several independent fiber routes, which can reduce risk for new projects. Heavy concentration, however, puts pressure on local grids and land values, which tends to push some growth toward secondary markets over time.
Permits, Workforce and Climate Hazards
Even a strong site can stall. Permits, environmental reviews and limits on energy and water supply can delay construction, and the grid upgrades a project depends on can take longer to finish than the facility itself. Timelines usually hinge on local permitting procedures and the state of nearby grid infrastructure.
Staffing is the other constraint. Operators need electrical engineers, cooling specialists, security staff and technicians who can keep the facility running around the clock. Areas with a strong IT presence offer a deeper talent pool. Remote sites may have cheaper land, but they usually require more spending on recruitment, training and retention.
Climate hazards need the same scrutiny. Flooding, wildfire, heat waves and storms can all damage facilities and the roads and transmission lines that serve them. That means weighing historical hazard records, climate forecasts, insurance costs and the resilience of surrounding infrastructure.
Four Questions to Put to Any Shortlist
Dataintelo's framework reduces the decision to four areas. Grid readiness asks whether the required capacity can be delivered on schedule. Resource resilience asks whether power and water hold up through seasonal shortfalls and severe weather. Expansion potential asks whether the site can accommodate construction in later years. Operational readiness asks whether skilled staff, support services and network links are in place.
These factors interact. A cheap parcel may have a grid connection that takes too long to energize. A region rich in renewable generation may still struggle with transmission capacity. Looking at each factor alone misses those trade-offs, which is why the assessment has to run in parallel.
Here's the thing: the cheapest land is no longer a reliable predictor of where a data center will succeed. The sites best positioned for the next wave will be the ones where power, water, connectivity, workforce and permitting line up. Watch grid interconnection queues and water planning in candidate regions, because those two timelines will probably decide which locations get built first.