Recommended Free Tools
Estimate an AI data centre in stages: define what the estimate covers, total the expected electrical load of the IT equipment, apply an explicit facility-overhead assumption for a preliminary power figure, then size the heat-removal path around the actual equipment and site. Rack count and floor area alone cannot establish the load, and a desk estimate is a planning range—not a buildable design.
What are you estimating: IT power, facility power, or cooling load?
These are related quantities, but they are not interchangeable:
As an Amazon Associate I earn from qualifying purchases.
- IT power is the electrical demand of the computing equipment in scope: accelerators, servers, storage, networking and other IT loads.
- Facility power includes IT power plus the electricity used by supporting systems, such as cooling and electrical distribution. A preliminary estimate can use IT power multiplied by an assumed Power Usage Effectiveness (PUE).
- Thermal load is the heat that must be captured, transported and rejected. IT electrical consumption is the first-order source of this heat, but the cooling plant’s electrical draw is a separate facility load—not an amount to add to the thermal load as though both were electrical megawatts.
State the boundary before doing arithmetic: IT rooms, the whole data-centre facility, or a campus. Also specify the geography and climate, planning phase, redundancy target, and whether the figure means connected nameplate capacity, expected coincident demand, or annual energy. Peak demand is measured in MW; energy over time is measured in MWh. For an annual estimate, use the assumed average demand over the period, not peak MW as if it ran continuously.
How do you estimate the IT load?
Build an equipment inventory
List quantities and vendor-specified power ranges for accelerators, host CPUs, memory, storage, network fabric and other in-scope equipment. Record the intended rack layout and expected workload or utilization scenario. Use the selected hardware’s specifications and a stated operating case; neither a marketing maximum nor a generic rack-density figure substitutes for project inputs.
Calculate low, central and high cases by varying documented assumptions—for example, equipment configuration or expected coincident operation—and show which assumptions change between cases. ASHRAE describes rack-level and floor-area measures as ways to characterize maximum load, while emphasizing that cooling capacity should match actual heat load. These measures help describe a layout; they do not independently determine the compute load. ASHRAE Handbook guidance.
Keep design capacity distinct from expected demand
A facility may need electrical and cooling capacity for a design case that differs from its expected average workload. Report expected coincident demand separately from connected nameplate load and from the capacity needed to meet the specified redundancy target. For final electrical design, engineers also need to assess redundancy and transient or step behavior; a simple sum of device ratings does not resolve those questions.
Rank #2
How do you estimate whole-facility electrical demand?
For a first screening estimate, use:
Estimated facility power = estimated IT power × assumed PUE
PUE is a facility-level measure: total facility energy divided by IT energy over the same boundary and period. It is not a universal constant, nor is it a formula for thermal load. State the PUE assumption and its basis; where the design is not yet established, calculate scenarios rather than presenting one unverified figure as typical. Include distribution losses and cooling-system energy either through the facility-level assumption or a more detailed subsystem model, but not both.
Rank #3
ASHRAE’s integrated-design page illustrates how architecture can affect the multiplier. In its 50 MW IT-load example, it gives PUE around 1.40–1.60 for a traditional chilled-water architecture and around 1.05–1.15 for a dry-cooled architecture. These are page-specific illustrative comparisons, not independent industry-wide measurements or guaranteed outcomes for a proposed site. ASHRAE integrated-design guidance.
| Illustrative architecture in ASHRAE’s 50 MW IT-load example | Illustrative PUE range | How to use the figure |
|---|---|---|
| Traditional chilled-water example | Around 1.40–1.60 | Scenario from the cited design page; not a site-specific prediction. |
| Dry-cooled architecture example | Around 1.05–1.15 | Scenario from the same page; not a guaranteed design result. |
For any project, the estimate still depends on the chosen equipment, operating conditions, climate, redundancy and system design. A PUE multiplier estimates total facility electrical demand; it does not say how much heat is captured in liquid versus room air, or what cooling capacity the heat-rejection system needs.
How do you estimate the heat-removal requirement?
Start with the selected IT load as the first-order heat source to remove. Trace where that heat goes: some may be captured directly in liquid loops, while the remainder enters room air. Include non-IT heat sources when they fall within the design boundary. The cooling system must capture and transport the heat, then reject it under the site’s design conditions.
Separate that thermal duty from the electricity consumed by pumps, fans, chillers or other cooling equipment. Those devices add to facility electrical overhead while moving and rejecting heat; thermal MW and electrical MW are not interchangeable. The final heat-rejection capacity depends on the actual equipment and site conditions and needs engineering review.
Best Value
Which cooling concept should you compare?
Compare only concepts compatible with the selected hardware and facility plan. AI racks can exceed conventional density assumptions, so use the actual equipment requirements rather than a sector-wide density guess. The U.S. Department of Energy’s 2024 design guide records historical HPC examples of 60 kW rack density observed in 2013 and deployments above 125 kW in later periods. Those examples show why older assumptions may not fit, but they are not current universal AI design values. DOE data-centre design guide.
ASHRAE’s AI data-centre framework discusses high-density thermal approaches, including liquid cooling, and thermal classes and economization. For each candidate, check allowable inlet conditions, rack density, facility-water conditions, heat-exchanger or coolant distribution unit arrangement, economizer potential, local climate and water availability, operations, and resilience. ASHRAE energy and thermal guidance.
| Concept | What to evaluate in the project |
|---|---|
| Air cooling | Whether the selected racks’ heat can be handled within allowable inlet conditions; airflow and the site’s heat-rejection conditions. |
| Direct-to-chip liquid | Which heat is captured in liquid, facility-water conditions, coolant distribution unit and heat-exchanger arrangement, and what residual heat remains in room air. |
| Rear-door heat exchangers | Compatibility with rack layout and heat capture, water-loop design, maintainability and the remaining room-air load. |
| Hybrid arrangement | How liquid and air loads divide, how each path is controlled and maintained, and whether the combined system meets the operating and resilience requirements. |
Do not pick a concept from density alone. Compare deliverable IT capacity and rack density under stated conditions alongside power overhead, thermal performance, water boundary and availability, resilience, climate and utility fit, and adaptability to future equipment or workloads.
PC Slower Than It Used to Be?
A free scan shows the junk files, broken settings and background clutter dragging Windows down - then fixes them in one click.Free scan · Windows 10 & 11Outdated Drivers Are Slowing You Down
One free scan finds every outdated or missing driver and matches the right update for your exact hardware.Free scan · exact hardware matchHow should water, carbon and other trade-offs be reported?
PUE alone does not describe every resource trade-off. ASHRAE’s framework references PUE, Water Usage Effectiveness (WUE), Water Usage Impact (WUI), Carbon Usage Effectiveness (CUE) and other Green Grid measures. Define the metric and its boundary whenever you report a value; water availability and climate suitability also matter when comparing cooling approaches. See the ASHRAE framework overview and its tools and standards listing.
What should the estimate include before it informs a design?
- Document scope and cases. State facility boundary, site and climate, planning phase, redundancy target, load definition, and low, central and high assumptions.
- Show the inventory and calculations. Identify equipment quantities and vendor power inputs; show the IT total, PUE assumption or subsystem model, and resulting facility-power range.
- Describe the thermal path. Identify heat captured by liquid, residual room-air heat, other in-scope heat sources and the intended site heat-rejection path.
- Separate planning outputs. Do not treat a preliminary demand estimate as final utility service, generator or UPS sizing, cooling capacity, or annual energy. Each is a distinct design output.
- Validate with current project evidence. Check current vendor data, measured or modeled workload profiles, local weather, applicable codes and standards, and integrated electrical and mechanical engineering.
The PNNL/ASHRAE/NEMA framework is intended to guide planning through operations and retrofit; it does not establish mandatory requirements or replace applicable codes and standards. ASHRAE framework overview.
Quick Recap
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.

