# AI Data Centers Are Thirsty and Power-Hungry, Report Warns

> AI data centers are surging into growth markets, but their thirst for power and water strains grids and timelines. This guide hits sites, cooling, utilities, permits, and risk.

- Type: AI news
- Page: https://completeaitraining.com/news/ai-data-centers-are-thirsty-and-power-hungry-report-warns/
- Published: 2026-03-06
- Updated: 2026-03-06
- Author: Jeroen Erne
- Jobs: Real estate and construction
- Reading time: 3 min

### AI data centers are straining energy and water. Here's how to plan, build, and finance accordingly

AI demand is pulling data center projects into every growth market. The catch: outsized electricity and water needs that stress local infrastructure and timelines. If you work in real estate or construction, this isn't a headline-it's a design brief and a risk register.

#### Why this matters for developers, owners, and builders

- Higher load per acre changes site economics, substation needs, and interconnect timing.
- Cooling choices can spike potable water demand unless you plan reclaimed or dry solutions.
- Permitting, community agreements, and supply-chain bottlenecks can stall a deal for quarters, not weeks.
- Tenants want speed, scale, and ESG credibility at the same time-your design and contracts must reconcile all three.

#### Site selection filters that prevent regret

- **Electrical capacity and timing:** Confirm available MVA, upgrade costs, and queue position. Ask for dual feeds, voltage level, and realistic energization dates. Transformer and switchgear lead times can stretch 12-30 months-lock them early.
- **Water security:** Validate rights, seasonal limits, and drought triggers. Prioritize reclaimed sources and confirm treatment, storage, and distribution constraints. Map discharge permits and thermal limits upfront.
- **Cooling fit:** Match climate to approach: air-cooled chillers, dry coolers with adiabatic assist, or liquid options for high-density racks. Model water-use effectiveness (WUE) and set hard caps by month.
- **Fiber and latency:** Proximity to long-haul routes, diverse paths, and carrier-neutral facilities reduces project risk.
- **Zoning and neighbors:** Noise from equipment yards, diesel testing windows, traffic, and visual screening. Bring community benefits and transparency to the first meeting, not the last.

#### Design moves that cut energy and water without killing schedule

- **Thermals:** Hot/cold aisle containment, higher supply-air temps within ASHRAE envelopes, and free-cooling hours. Design for staged upgrades as rack densities rise.
- **Cooling tech:** Choose low- or no-evap strategies where water is tight. Where water is available, set use caps and install metering at make-up and blowdown to manage in real time.
- **Liquid for hotspots:** Start with direct-to-chip in high-density zones before full immersion. Plan service clearances and leak detection into the layout.
- **Electrical architecture:** Right-size UPS topologies and consider battery systems that support grid services to improve economics while meeting reliability targets.
- **Heat reuse:** If you're near district loops or large thermal loads (hospitals, campuses), route waste heat to offtakers. It's capex-heavy, but it scores with cities and ESG screens.
- **Monitoring:** Instrument PUE and WUE at commissioning. Tie DCIM and BMS data to operating playbooks so drift is caught fast.

#### Pro formas and contracts that won't backfire

- **Utilities:** Model demand charges, capacity reservations, and curtailment risk. Negotiate construction service agreements and milestone energization dates.
- **Procurement:** Long-lead items-transformers, switchgear, gensets, chillers-need parallel paths and deposits. Prefabricated electrical rooms can shave months if logistics are nailed down.
- **Energy sourcing:** Explore utility green tariffs, PPAs, or RECs aligned to tenant requirements. Clarify who owns the attributes and reporting rights.
- **Water terms:** Codify drought-stage allocations, alternative sources, and a cooling changeover plan. Price out non-potable tie-ins and on-site storage from day one.
- **Leases and SLAs:** Define density ramps, redundancy tiers, and efficiency targets with incentives and remedies on both sides.

#### Permits, community, and compliance

- **Environmental review:** Groundwater drawdown, thermal discharge, stormwater, wetlands, and air permits for emergency generation. Keep studies concurrent, not sequential.
- **Construction windows:** Noise curfews and traffic management influence phasing and crane days. Share the plan early with local officials.
- **Community benefits:** Local hiring, apprenticeships, and infrastructure upgrades build trust and speed approvals.

#### What to do next

- Run a location scan that scores sites by MVA available, interconnect timing, reclaimed water access, fiber diversity, and permitting path.
- Hold joint pre-application sessions with the utility and water authority; align on dates and deliverables before design development.
- Commission an integrated energy-and-water model; set firm targets for PUE and WUE and tie them to pay apps.
- Pick a cooling strategy that meets annual water caps and can pivot as rack density grows.
- Issue LOIs for long-lead equipment now, with alternates approved.
- Prepare a community brief that explains electrical and water use, mitigations, and benefits in plain language.

Want more industry-specific AI context and case studies for your team? Explore [AI for Real Estate & Construction](https://completeaitraining.com/tag/real-estate-and-construction/).

For background on sector-wide resource use, see the International Energy Agency's analysis of data centers and networks ([IEA](https://www.iea.org/energy-system/end-uses/data-centres-and-data-transmission-networks?ref=completeaitraining.com)) and U.S. guidance on data center efficiency strategies ([DOE Better Buildings](https://betterbuildingssolutioncenter.energy.gov/technology-focus/data-centers?ref=completeaitraining.com)).

---

This is the Markdown form of https://completeaitraining.com/news/ai-data-centers-are-thirsty-and-power-hungry-report-warns/. The whole site is mapped at https://completeaitraining.com/llms.txt.
