Data centers: separating the claims from the local reality

A data center can make useful AI possible and put pressure on a community's resources. Assessing one means asking what it does, what it consumes, and who receives the benefits. These are questions we can investigate without deciding in advance that every project is a disaster or a success.

“All that electricity is for AI”

Data centers also run websites, cloud storage and other digital services. The IEA's estimate of 415 terawatt-hours in 2024—about 1.5% of global electricity use—covered data centers as a whole. Presenting that entire amount as AI consumption overstates what the figure measures.

The IEA's 2025 base case projected about 945 terawatt-hours in 2030. That is a forecast with assumptions, not a future reading from a meter. It helps illustrate the possible scale of growth; it cannot establish the demand of a particular proposed facility. Read the IEA analysis.

“The global share is small, so local concerns are exaggerated”

The denominator changes the question. A worldwide percentage says little about whether one area's grid has spare capacity when a large new customer arrives.

The IEA's 2026 analysis explains that price effects depend on supply, demand and infrastructure. More customers can spread existing costs when capacity is available. Rapid growth can instead require expensive additions and put pressure on supply. Neither an automatic price increase nor an automatic saving follows from the words “data center.” Read the 2026 analysis.

Our practical question: what capacity is available here, when will the facility need it, and who pays if the forecast changes?

“Better chips will solve the energy problem”

Efficiency improvements are encouraging. They mean a given task can require less energy. Total demand also depends on how many tasks people run and how demanding those tasks become.

The IEA reported that data-center electricity use grew 17% in 2025 even as efficiency per AI task improved. Its April 2026 update describes both clean-energy procurement and projects pursuing on-site gas. Efficiency and cleaner supply deserve attention, but a promise of improvement needs a timetable and measurable results. Read the update.

“Every data center uses water in the same way”

Cooling design matters. So does the source of electricity. Berkeley Lab's 2024 US report distinguishes water consumed directly by data centers from water associated with producing their power. It estimates direct US data-center water consumption at roughly 66 billion liters in 2023; this includes facilities serving uses beyond AI.

Consumption and withdrawal are different measurements: water taken from a source is not necessarily all lost from that source's immediate water cycle. A national total also does not tell you whether a proposed site's water demand is manageable during a local drought. Read Berkeley Lab's report.

For a specific proposal, ask for annual and peak demand, cooling design, the water source, and drought arrangements. Ask which figures cover the site and which cover electricity generation. Comparing unlike totals can produce a dramatic headline without a useful answer.

“There are no benefits”—or “the jobs settle the argument”

Virginia's legislative research commission found economic benefits in its 2024 review, particularly from construction and local tax revenue. It also found fewer ongoing workers than some other industries. A construction job count should not be presented as permanent staffing.

The same review documented potential electricity-cost increases and complaints about persistent noise near some homes. These are historical findings from one major market, not a verdict on every facility. They show why a local assessment should include residents' experience alongside investment totals. Read the Virginia review.

What would make a proposal convincing?

These are our editorial questions, rather than claims that any particular developer has met the standard:

• Is demand backed by a credible operating plan, with clear milestones?
• Are infrastructure costs and risks assigned transparently?
• Do water and noise assessments reflect the actual location and nearby residents?
• Are promised jobs separated into construction and ongoing roles?
• Can the community check performance after the project opens?

The hopeful part is that design, siting, measurement and public decisions can influence the outcome. Optimism is most useful when it makes those choices more concrete.