Data Centers: Separating Fact from Fiction

Everyone. Simply put, data centers are the physical infrastructure that makes nearly every modern internet service possible. Every time someone:

• Streams a movie on Netflix,
• Searches on Google,
• Shops on Amazon,
• Uses social media such as Meta platforms,
• Stores photos in the cloud,
• Participates in a video call,
• Uses online banking,
• Accesses electronic medical records,
• Interacts with artificial intelligence (AI) tools,
• Reads a FAQ,

their request is ultimately processed and stored in a data center somewhere. Virtually everyone who uses the internet benefits from data centers because they provide the storage, computing power, and connectivity that make online services possible. Without data centers, the modern internet would not function.

Data center use of water supplies is a small drop in the bucket compared to most other large-scale activities. As AI researcher and policy analyst Andy Masley notes, “On the national, local, and personal level, AI is barely using any water, and unless it grows 50 times faster than forecasts predict, this won’t change. … All U.S. data centers (which mostly support the internet, not AI) used 200–250 million gallons of freshwater daily in 2023. The U.S. consumes approximately 132 billion gallons of AsAfreshwater daily.” That translates to about 0.2 percent of freshwater supplies, a figure consistent with the Taxpayers Protection Alliance’s (TPA’s) current estimates for states such as Utah, California, Texas, and Pennsylvania. As noted by Reason’s Robby Soave, “California’s almond farms consume 4.2 billion gallons of waters per day, according to Reason’s Christian Britschgi. Data centers consume just 46 million gallons per day. … compared to all the other things that use water—golf courses account for 1.4 billion gallons per day—it’s just a drop in the bucket.”

A related concern is that extensive use of AI “wastes” vast amounts of water. This too is unfounded. For example, to match the water footprint of manufacturing one pair of jeans (3,000 to 4,000 liters of water), a user would need to interact extensively with an AI model, generating roughly 150,000 to 200,000 conversational prompts—even assuming high alarmist estimates of AI water consumption.

There is simply no evidence that data centers are increasing electricity prices. According to a comprehensive 2026 analysis by the Institute for Energy Research, the “number of data centers does not correlate with current electricity prices.” There is “a slightly positive but still statistically insignificant relationship between data centers and fast price increases,” meaning that any correlation is likely due to chance.

The reasoning behind this finding is fairly simple. Electric power systems require substantial upfront investments in generation facilities, transmission networks, and other infrastructure, along with long-term financial commitments that must be paid regardless of overall electricity consumption. As demand increases, these fixed costs can be distributed across a larger volume of electricity sales, helping to—if anything—keep rates lower on a per-unit basis. However, as TPA’s state-level electricity consumption estimates show, data centers simply are not accounting for a large enough percentage of overall consumption to move the needle.

As the below chart shows, related concerns about AI using significant amounts of electricity are also unfounded—using a conservative figure of approximately 0.3 watt-hours (Wh) per typical AI text prompt.

Source: TPA analysis of U.S. Department of Energy and the Energy Information Administration figures. 

While many policymakers are concerned about growing data center resource demands, consumption will likely level off or even decline over the next ten to fifteen years because of a number of factors, including more efficient water and electricity usage. For example, at the silicon level, the integration of next-generation optical computing and neuromorphic chips will likely reduce the electrical power required per computation compared to traditional copper-based processors. This can result in an “order of magnitude” less energy being consumed.

Simultaneously, data centers are moving away from evaporative water cooling in favor of closed-loop liquid cooling, direct-to-chip microfluidics, and ambient air systems. By sealing water inside a continuous, recirculating loop or utilizing non-potable dielectric fluids, facilities can drop their operational water consumption significantly while handling the extreme heat rejection required by high-density artificial intelligence clusters.

States and localities that embrace data center projects benefit from increased economic activity and, very often, surging tax revenue. According to a recent analysis by the National Conference of State Legislatures, “In Loudoun County, the center of Virginia’s so-called data center alley, data centers now pay enough in property taxes to fund the county’s entire general operations budget. A comprehensive 2024 data center study in Virginia found that data centers could bring significant benefits to localities in economically distressed areas of the state from increases in local tax revenue.”

Beyond direct taxes, developers frequently inject millions of dollars into communities upfront by completely financing major public utility upgrades, including rebuilding aging municipal water and sewer lines, upgrading local electrical grids, and expanding fiber-optic networks at no or very low cost to taxpayers.