keepcalculator

06 Numbers and Powers

AI Data Center Construction: Power & Water Impact

Card No. 06·6769 · Updated · Standard formula, shown below

This guide explains how building an AI data center can change local electricity bills and strain water supplies, and what you can do to prepare. You’ll learn the key factors that drive demand, how to estimate the impact on your community, and practical steps to mitigate problems before construction begins.

What You’ll Need

  • Planned data center specifications (IT load in megawatts, cooling type, location)
  • Local utility rate structure and water pricing information
  • Average household electricity and water usage data for your area
  • Access to public records or utility reports on grid capacity and water stress
  • Basic spreadsheet or calculator for projections

Step-by-Step

  1. Estimate the IT load. AI data centers typically run at higher power densities than traditional ones. Find the planned IT capacity in megawatts (MW). If not published, assume 10–50 MW for a large facility, but check local filings.
  2. Add cooling and overhead. Total facility power is usually 1.3–1.5 times the IT load. Multiply the IT MW by 1.4 for a rough estimate. For example, a 20 MW IT load becomes about 28 MW total.
  3. Convert to annual electricity use. Multiply total MW by 8,760 hours per year and then by a utilization factor (often 0.7–0.9 for AI workloads). A 28 MW facility at 80% utilization uses about 196,000 MWh per year.
  4. Compare to local residential consumption. Look up the average annual household electricity use in your region (in the U.S., roughly 10,500 kWh/year). Divide the data center’s annual MWh by that number. The 196,000 MWh example equals about 18,700 households’ worth of electricity.
  5. Check grid capacity and rates. Contact the local utility or read their integrated resource plan. Ask if the grid can handle the new load without building new generation or transmission. If new infrastructure is needed, costs may be passed to all ratepayers through higher tariffs or demand charges.
  6. Assess water usage for cooling. Determine the cooling technology. Evaporative cooling towers use 2–5 gallons per kWh of IT load (roughly 8–20 liters per kWh). Direct-to-chip liquid cooling with closed loops can cut that by 50–80%. For a 20 MW IT load with evaporative cooling, estimate 40–100 million gallons per year (150–380 million liters).
  7. Compare to local water supply. Find your municipality’s annual water production or the watershed’s sustainable yield. A data center using 100 million gallons a year is significant for a small town but negligible for a large city. Check if the water source is already over-allocated.
  8. Review local incentives and regulations. Some jurisdictions offer tax breaks or expedited permits for data centers, which can affect how quickly they get built. Look for water use permits, environmental impact reviews, and any required mitigation measures.
  9. Estimate your own bill impact. If the utility must build new capacity, rates may rise. Ask the utility for a rate case forecast. In some areas, data centers pay for their own substations, reducing the impact on existing customers. In others, costs are socialized.
  10. Prepare a community response. If you’re a resident or local official, request a public meeting with the developer and utility. Ask for a detailed water and energy impact statement. Consider negotiating a community benefits agreement that funds efficiency programs or water conservation.

Common Mistakes

  • Using IT load instead of total facility power – underestimates by 30–50%.
  • Assuming all data centers use the same cooling. Location and design matter greatly.
  • Ignoring the time of day. AI workloads often run 24/7, unlike residential peaks, which can change grid dynamics.
  • Forgetting that water used for cooling is often returned (though warmer) or evaporated. Net consumption matters, not just withdrawal.
  • Overlooking the supply chain: manufacturing GPUs and servers also consumes water and energy, but that’s outside the local footprint.

Tips

  • Ask for the PUE (Power Usage Effectiveness) – a lower PUE (e.g., 1.1) means less overhead.
  • Look for water reuse or air-cooled designs in arid regions.
  • Check if the data center can use treated wastewater (non-potable) for cooling – many new facilities do.
  • Monitor utility rate cases – they are public and show projected demand.
  • Engage early. Once a permit is issued, it’s harder to change the design.

Troubleshooting

  • No public data on the facility: File a public records request with the local planning department. Many jurisdictions require an environmental review that includes energy and water estimates.
  • Utility won’t share grid capacity: Look for the utility’s Integrated Resource Plan (IRP) or system reliability report, often posted online. It will show projected load growth and planned new generation.
  • Water usage numbers seem too high or low: Verify the cooling type. If the facility uses closed-loop cooling with chillers, water use is minimal. If it uses evaporative towers, it’s high. Ask for the design’s WUE (Water Usage Effectiveness).
  • Your electricity bill hasn’t changed yet: Construction takes years. The impact may come when the facility goes live. Keep monitoring utility rate filings.
  • Local officials are dismissive: Bring your own estimates to a public meeting. Show the household-equivalent numbers. Often, a visual comparison is more persuasive than raw MW.

Frequently Asked Questions

Q: How much electricity does a typical AI data center use?
A: A large facility can use 50–100 MW total, equivalent to 40,000–80,000 homes. Smaller ones use 5–20 MW.

Q: Will my electricity bill definitely go up?
A: Not always. If the utility has spare capacity, rates may stay flat. But if new power plants or transmission lines are needed, costs are often spread across all customers.

Q: Do data centers use water even if they don’t have cooling towers?
A: Yes, but much less. They still need water for on-site cooling loops, humidification, and sanitation. Air-cooled designs can cut water use by 90% compared to evaporative towers.

Q: Can data centers use seawater or wastewater for cooling?
A: Some coastal facilities use seawater, but it requires corrosion-resistant materials. More common is using treated municipal wastewater (non-potable) for cooling towers, which reduces strain on drinking water supplies.

Q: How long does it take for a data center to start affecting my bills?
A: From groundbreaking to operation is usually 2–4 years. The impact on rates may come earlier if the utility starts building new infrastructure in anticipation.

Q: What can I do if I’m worried about water shortages?
A: Attend public hearings, submit comments on environmental reviews, and ask local officials to require water recycling or air cooling. You can also support water conservation programs that offset the new demand.

Related

Method: the standard formula for this calculation, rounded as each result states. See how the calculators work or report an error.

More cards in this drawer

The whole Numbers and Powers drawer →