How to Estimate the Power and Water Needs of a Hyperscale Data Center

iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more

Estimate a hyperscale facility in four separate outputs: coincident peak facility demand (MW), annual electricity (MWh or TWh), direct water consumed at the site (liters or cubic meters), and water consumed to produce its electricity. Start with the planned IT load and hourly operating profile, then apply condition-specific PUE and a cooling- and climate-specific site WUE. Keep peak power separate from annual energy, and never describe site WUE as the facility’s complete water footprint.

What the estimate must tell you

A useful estimate answers different planning questions rather than producing one headline number. Utilities need the facility’s maximum coincident demand in megawatts (MW) to size interconnection, substations, feeders and generation. Finance and energy planners need annual electricity in megawatt-hours (MWh) or terawatt-hours (TWh). Water authorities need direct consumption at the campus, its seasonal pattern and its source. Sustainability teams may also need the water consumed by power plants supplying the facility.

#	Preview	Product	Price
1		Spanish -- Thermal Guidelines for Data Processing Environments, 4th ed.		Buy on Amazon
2		Thermal Guidelines for Data Processing Environments by TC9.9 MISSION CRITICAL FACILITIES...	$133.63	Buy on Amazon

Report these quantities in separate lines:

• Peak facility demand (MW): the highest coincident electrical load for the IT plant and supporting infrastructure under the design or operating condition being analyzed.
• Annual facility electricity (MWh/year): the time-integrated consumption of IT equipment, cooling and all other facility systems.
• Direct/site water (liters or m³/year): water consumed at the data-center boundary, principally for cooling in the Lawrence Berkeley National Laboratory (LBNL) accounting boundary.
• Indirect/source water: water consumed in generating the electricity used by the facility. It is determined by the electricity supply and the water intensity of its generation technologies, not by the cooling equipment at the campus.

A peak value cannot be converted to annual energy without a load profile. Conversely, an annual average tells a utility little about the short interval when the campus reaches its maximum demand.

Build the estimate from the IT load

The IT plant is the foundation. Inventory the equipment that will be installed, not a generic “server load.” Include general-purpose servers, high-power AI accelerators, storage, network switches and any specialized appliances. For each class, record nameplate or measured power, expected utilization, idle power, quantity, operating lifetime and the date at which equipment is added or retired.

#1 Best Overall

Spanish -- Thermal Guidelines for Data Processing Environments, 4th ed.

Buy on Amazon

Minimum equipment inputs

Use measured telemetry from an existing fleet when available. If the campus is new, create low, base and high cases for device counts, utilization, idle power and refresh timing. Document whether a “peak” means a manufacturer rating, a design limit, a measured coincident event or a statistically likely operating maximum.

Convert IT electricity to total facility electricity with PUE

Power Usage Effectiveness (PUE) is defined as:

PUE = total data-center electricity demand ÷ IT-equipment electricity demand.

PUE is dimensionless. It captures cooling and electrical overhead such as uninterruptible power supplies, transformers, distribution losses, fans and pumps. LBNL uses these categories in its infrastructure modeling.

Annual-energy calculation

For a reporting period, calculate IT energy first, then apply a PUE representative of that facility and period:

Annual facility energy (kWh) = annual IT energy (kWh) × annual PUE

For example, if an annual IT model produces 300,000 MWh and the modeled annual PUE is 1.25, facility electricity is 375,000 MWh. The multiplication is valid only if the PUE represents the same boundary, weather period and operating mix as the IT value.

Peak-demand calculation

For a design or operating peak, use a coincident peak IT load and a peak-condition PUE:

Peak facility demand (MW) = coincident peak IT demand (MW) × peak-condition PUE

Do not substitute an annual average PUE when sizing a connection. Chillers, pumps, fans and power-conversion equipment may be operating at a different efficiency at the design temperature or during a rapid workload surge. A facility can have an excellent annual PUE and still require a substantially larger peak interconnection.

Model the hourly and seasonal load shape

At minimum, make an hourly model for a full year. For each hour, calculate IT power from the equipment fleet and workload, determine outdoor temperature and humidity, select the available cooling mode, and calculate electrical overhead. Sum the hours for annual MWh and retain the highest coincident facility value for peak MW.

Load-shape checks

• Check whether accelerator jobs are synchronized, creating a short coincident peak.
• Represent overnight, weekend and maintenance-period utilization rather than assuming 100% operation.
• Include seasonal weather, because cooling power and economizer availability vary with temperature and humidity.
• Model staged build-out. A substation may be sized for the ultimate campus while annual energy in the first operating year is much lower.
• Separate critical-load ride-through assumptions from normal operating demand; batteries and generators affect reliability design but do not erase the underlying energy requirement.

Estimate direct water with site WUE

Water Usage Effectiveness (WUE) for the site is expressed as:

Site WUE = water consumed at the facility (liters) ÷ IT electricity (kWh).

For a selected reporting period:

Direct site water (liters) = site WUE (L/kWh) × IT energy (kWh)

Use a WUE that matches the cooling architecture, local climate, setpoints and operating pattern. The denominator is IT energy, not total facility energy, so state the definition beside every WUE value.

Cooling-tower water

Evaporative cooling towers reject heat by evaporating water and also require blowdown to control concentrated minerals. Makeup water therefore includes evaporation, blowdown and other operational losses. The U.S. Department of Energy Federal Energy Management Program (DOE FEMP) explains: “A cooling tower system by necessity uses an extensive amount of water because the warm water from the chiller’s condenser water loop is cooled by evaporating water into the surrounding atmosphere.” Actual consumption depends on heat load, cycles of concentration, weather, tower efficiency, maintenance and the amount of time the tower operates.

Rank #2

Thermal Guidelines for Data Processing Environments by TC9.9 MISSION CRITICAL FACILITIES (2004-06-30)

$133.63

Buy on Amazon

Air-cooled and hybrid systems

Air-cooled systems can reduce direct site water where water availability is constrained, but they may use more electricity during hot conditions and can impose a larger peak. Hybrid designs switch between air and evaporative modes according to weather, load and water constraints. Model each mode and its transition rules rather than assigning one annual WUE to the entire year.

Keep direct water and source water separate

Site water is not the complete water footprint. Electricity generation can consume water upstream, and the amount varies with the grid mix and the water intensity of each generation technology. LBNL models site WUE separately from source water associated with electricity generation.

Do not describe source water as water drawn at the campus. Conversely, do not present a site WUE as the full water footprint. If the electricity contract changes seasonally or hourly, use the corresponding supply mix instead of a single unsupported national average.

Inputs that most change the answer

• AI and accelerator density: high-power chips increase IT energy and the heat that cooling systems must reject.
• Idle power and utilization: two fleets with identical nameplate capacity can have very different annual energy.
• Cooling architecture: chillers, cooling towers, direct liquid systems, air cooling, economizers and hybrids have different PUE and WUE behavior.
• Setpoints and airflow: temperature, humidity, fan speeds and containment determine how much mechanical cooling is needed.
• Climate and water availability: outdoor conditions determine economizer hours, peak cooling power and evaporative demand.
• Electrical infrastructure: UPS, transformers, distribution, fans and pumps add overhead that is easy to omit from an IT-only estimate.
• Grid supply: the generation mix and its water intensity determine source-water results.

A repeatable estimation workflow

1. Define the boundary and outputs. Specify whether the study covers one building or a campus, whether water means withdrawal or consumption, and the reporting year. Commit to separate peak MW, annual MWh, direct water and source water.
2. Create the equipment schedule. List quantities and power curves for servers, accelerators, storage and networking. Add installation, retirement and growth dates.
3. Build the workload profile. Assign utilization and idle power by hour or representative operating period. Produce low, base and high cases.
4. Calculate hourly IT power. Sum device power for each time step, preserving both average and coincident peaks.
5. Select cooling modes. For each hour, determine whether air-side or water-side economizing, mechanical cooling, evaporative cooling or a hybrid mode is available under the local temperature and humidity limits.
6. Apply condition-specific PUE. Add UPS, distribution, fans, pumps and cooling loads. Use an annual PUE for annual energy and a design-condition PUE for peak demand.
7. Apply site WUE. Multiply IT energy by the WUE for the selected cooling mode and weather period. Include evaporation and blowdown where applicable.
8. Calculate source water separately. Multiply electricity by the water-intensity factors for the actual or modeled supply mix. Report the boundary and mix year.
9. Stress-test assumptions. Re-run equipment count, utilization, idle power, weather, setpoints, cooling design and water-availability cases. Show the range and identify which assumption drives it.
10. Review engineering and accounting checks. Confirm units, coincident timing, water definitions, equipment operating limits and the treatment of backup systems before using the result for a utility or permit decision.

How national figures provide context—but not a site design

LBNL’s 2025 Update, published in 2026, models U.S. data-center electricity using equipment-shipment information, per-device annual electricity models, cooling-system performance simulations, facility types and locations. Its reference case estimates 649 TWh in 2030, equal to 11.8% of total U.S. electricity use in that year. The compounded uncertainty range is 521–843 TWh, or 9.5–15.3% of total U.S. electricity use. These are national modeled outcomes, not a forecast for an individual campus.

The range changes with assumptions about equipment installations, specialized graphics-chip shipments, AI-chip operating life and AI-server idle power and utilization. For a project, replace those national assumptions with the equipment schedule and operating data for the proposed site.

LBNL’s 2024 United States Data Center Energy Usage Report projects 60–124 billion liters of direct/site water use by U.S. hyperscale data centers in 2028. That is a national projection and should not be divided by an assumed number of campuses to create a “typical” facility. It is also from an earlier report than the 2025 electricity update, so the two figures are not a synchronized forecast.

Compare candidate designs or sites on the same axes

When evaluating alternatives, hold the reporting boundary and weather year constant. A compact comparison should include:

No cooling option is universally best. A water-constrained site may accept higher electrical demand from air cooling; a hot, water-abundant site may favor evaporative assistance; a mild climate may support long economizer operation. The correct choice follows from the modeled climate, water boundary, reliability requirement and workload.

Operational practices that change energy and water

Contain airflow

Separating hot and cold aisles limits mixing between server exhaust and supply air. DOE FEMP says the described airflow practices can enable higher chilled-water temperatures and reduced airflow, with 20% less chiller energy according to its referenced Best Practices Guide for Energy-Efficient Data Center Design. Treat that figure as the guide’s cited result for the described practices, not as a guaranteed whole-facility reduction. Actual savings depend on climate, containment quality, controls and the rest of the cooling plant.

Use economizers when conditions allow

Air-side economizing uses suitable outside air instead of mechanical cooling; water-side economizing bypasses or reduces chiller operation. Both depend on outdoor temperature, humidity, air quality, filtration and equipment operating limits. Count the hours and load served by each mode in the hourly model rather than assuming an annual percentage.

Set temperature and humidity deliberately

Setpoints affect chiller lift, fan power, economizer availability and tower operation. Keep every chosen setpoint within the equipment operating envelope. A change that lowers mechanical cooling can increase fan energy or affect water use, so evaluate total facility electricity and direct water together.

Do not confuse filtration with demand reduction

DOE FEMP cautions that side-stream filtration by itself does not reduce facility power or water use. It must be paired with technologies or operating changes that reduce cooling demand or otherwise change evaporation, blowdown or heat rejection. When assessing an intervention, identify the physical mechanism and put that mechanism into the PUE or WUE model.

Common estimation errors and fixes

• Using nameplate IT power as annual energy. Replace it with an hourly utilization and idle-power profile.
• Multiplying peak MW by 8,760 hours. This treats a peak as continuous operation; integrate the actual load shape instead.
• Applying annual PUE to a design peak. Use a PUE measured or modeled for the coincident peak condition.
• Calling site WUE the total water footprint. Add a separately calculated source-water result and label both boundaries.
• Ignoring blowdown. Tower makeup water includes more than evaporation; include cycles of concentration and operating practice.
• Copying a national average to a campus. National scenarios are context. Use project equipment, cooling, weather and grid data.
• Hiding seasonality. Report monthly or seasonal water and peak results where permits or supply constraints are seasonal.
• Reporting an unsupported single number. Show low, base and high cases and list the assumptions that create the spread.

Technical reference for operating limits

DOE FEMP points readers to ASHRAE Thermal Guidelines for Data Processing Environments, 4th Edition for equipment operating-condition guidance. Use the current edition and verify its applicability to the equipment and control strategy being designed. The guideline informs allowable conditions; it does not replace a project-specific thermal, electrical and water model.

What a decision-ready result looks like

A final estimate should include a one-page assumptions register and a results table. The table should show low, base and high cases for coincident peak facility MW, annual facility MWh, direct/site water by month and year, and source water under the stated electricity mix. Beneath it, identify the equipment schedule, utilization and idle-power assumptions, PUE and WUE by operating mode, weather data, setpoints, water boundary, grid boundary and uncertainty drivers.

This format lets a utility test interconnection capacity, a water planner test seasonal demand, and an engineering team see which design change actually moves the result. It also prevents a national scenario, a one-time peak and a site-water metric from being mistaken for interchangeable measures.

Quick Recap

Bestseller No. 1

Spanish -- Thermal Guidelines for Data Processing Environments, 4th ed.

Bestseller No. 2

Thermal Guidelines for Data Processing Environments by TC9.9 MISSION CRITICAL FACILITIES (2004-06-30)

$133.63

Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.