March 22, 2026
Green Data Centers: Metrics, Design, and Procurement Guide
Evaluate green data centers using PUE, WUE, carbon, renewable energy, heat reuse, refrigerants, embodied impacts, and current EU reporting context.
Updated September 19, 2026 to replace a single-threshold definition of “green” with measurable energy, water, carbon, heat-reuse, and lifecycle criteria.
A green data center is not defined by one PUE number, a renewable-electricity contract, or a certification logo. It is a facility whose environmental performance is measured across clear boundaries and improved without hiding trade-offs elsewhere in the system.
For a buyer, the practical task is to set performance objectives for energy, electricity sourcing, water, refrigerants, carbon, materials, heat reuse, and IT utilization; define how each will be measured; and compare proposals under the same workload, climate, availability, and lifecycle assumptions.
This guide explains the metrics, current EU reporting context, design decisions, and procurement evidence needed to turn “green” from a marketing adjective into an operating requirement.
There is no universal green data center threshold
A PUE below 1.4 does not by itself make a data center green, and PUE above 1.4 does not by itself prove a poor design. Climate, load, redundancy, cooling architecture, measurement boundary, and operating state affect the result. The ISO/IEC 30134 series defines individual resource-performance indicators; it does not set one global pass/fail limit or combine them into a universal green score. The current ISO/IEC 30134-2:2026 explicitly states that the series does not specify limits or targets for its KPIs.
Define “green” for the project by naming the outcomes, metrics, boundaries, baselines, targets, and verification method. A credible brief should also state which trade-offs are acceptable. A lower-energy evaporative design may consume more water. Greater electrical redundancy can increase embodied and operating impacts. Warmer liquid cooling may improve heat-reuse potential but only if a suitable heat user exists.
Measure more than PUE
| Measure | What it answers | What it does not answer |
|---|---|---|
| PUE | How total data-center energy compares with IT energy over the same period and boundary | Useful computing, carbon intensity, water, peak power, or embodied impacts |
| WUE | Direct water consumption relative to IT energy for a defined operating boundary | Water scarcity, source-water impact from electricity, or embodied water unless separately included |
| REF | The share of data-center energy accounted for by renewable energy under the selected method | Hourly physical supply, avoided emissions, or local grid impact without more detail |
| ERF | The share of total data-center energy reused outside the data center | The usefulness, timing, temperature, or efficiency of the receiving system by itself |
| Operational GHG emissions | Emissions associated with electricity, fuel, refrigerants, and other operating sources | Construction and equipment emissions unless the inventory includes them |
| Embodied impacts | Impacts from materials, manufacture, transport, construction, replacement, and end of life | Operational performance unless combined in a whole-life assessment |
| IT utilization or useful work | How effectively the IT capacity serves the workload | Facility overhead or environmental impact without energy and resource data |
ISO/IEC 30134-9 defines WUE for use-phase water consumption. ISO/IEC 30134-6 defines Energy Reuse Factor. Report the indicator and the underlying absolute quantities so a ratio does not conceal growth in total energy or water use.
Energy: efficiency and absolute consumption
PUE is total data-center energy divided by IT energy for a defined period. It is useful when the numerator, denominator, measurement points, and reporting period are consistent. It should be accompanied by total facility energy, IT energy, load, climate, and major operating conditions.
A low-load facility can have a worse PUE while consuming less total energy. More efficient IT can reduce the denominator and change the ratio even when the facility's absolute energy falls. Conversely, an excellent PUE at a large or poorly utilized computing load can still coincide with high total consumption.
Design measures can include efficient power conversion, right-sized transformers and UPS systems, high-efficiency fans and pumps, containment, higher allowable temperatures, liquid cooling where the equipment requires it, economizer operation, and controls that respond to actual load. Model annual performance with the site weather, load profile, equipment curves, and intended redundancy state. Do not use full-load component efficiency as an annual result.
The data center PUE guide explains measurement categories, annual calculations, and why PUE cannot size the incoming electrical system.
Electricity sourcing and carbon accounting
Renewable electricity can come from on-site generation, a direct power purchase, a supplier product, or energy-attribute certificates. These arrangements have different relationships to time, location, additional generation, and physical delivery.
For greenhouse-gas reporting, the GHG Protocol Scope 2 Guidance distinguishes location-based emissions, which reflect the grids where electricity is consumed, from market-based emissions, which reflect qualifying contractual instruments. Where both apply, report both rather than calling contractual matching “zero-carbon operation” without the method.
Ask for:
- annual electricity consumption and peak demand;
- location-based and market-based emission factors and results where applicable;
- the contractual instrument, generation geography, technology, vintage, and matching period;
- on-site fuel and refrigerant emissions;
- how backup-generation tests and outages are treated;
- the baseline and method for any avoided-emissions claim.
Water: separate inventory, consumption, and source impact
A closed coolant loop can retain its working fluid while the outdoor cooling system consumes water. Conversely, a dry-cooler design can reduce on-site consumption while the electricity supply has its own water footprint. State the boundary.
Track initial fill, stored water, make-up, blowdown, evaporation, treatment losses, discharge, annual direct consumption, and water source. Evaluate the result against local scarcity and competing uses, not only a global litres-per-kWh target. The AI water-use guide separates direct, indirect, and embodied water and explains why tank volume is not annual consumption.
Dry heat rejection can reduce direct water use but may require more fan energy or mechanical cooling at high ambient temperatures. Adiabatic assistance can improve hot-weather performance while using water during selected hours. Compare the annual site-specific energy-water trade-off.
Heat reuse: capacity is not delivered benefit
Data-center heat can serve buildings, district networks, industrial processes, agriculture, or other loads when temperature, distance, timing, and commercial arrangements align. A warm-water liquid-cooling system can improve the quality of the available heat, but it does not create a customer or guarantee useful delivery.
Define:
- temperature and flow at the handoff point;
- hourly or seasonal heat availability;
- the receiver's demand profile and required temperature;
- heat exchangers, heat pumps, pipes, pumps, backup, and ownership;
- energy used to upgrade and deliver the heat;
- meters, contracts, downtime, and the method for calculating exported energy and ERF.
Thermal MW is capacity. Delivered MWh is energy actually reused over time. Keep them separate.
Refrigerants and cooling-system impact
Cooling choices affect refrigerant type, charge, leakage risk, maintenance, energy, and end-of-life recovery. Record the refrigerant, global warming potential under the applicable regime, total charge, leak-detection method, expected service practice, and recovery plan. A system with low operational energy but a high leakage rate may not be the better carbon option.
Liquid cooling does not remove refrigeration automatically. The outdoor plant may use dry cooling, adiabatic support, or mechanical chillers depending on equipment temperatures and climate.
Embodied carbon, materials, and circularity
Operational energy often dominates a data center's impact, but lower-carbon electricity and shorter hardware cycles increase the importance of materials and equipment. Whole-life assessment can include the structure, concrete, steel, electrical equipment, cooling plant, batteries, IT hardware, transport, maintenance replacements, and end of life.
ITU guidance describes embodied impacts as a lifecycle question and cautions that studies are not comparable unless their conditions and datasets align. Use project-specific quantities and transparent product data where available. Compare service life, capacity actually used, replaceable modules, repairability, reuse, and end-of-life routes rather than relying on one generic carbon factor.
Phasing can reduce stranded materials when capacity is added only as demand becomes credible. It can also duplicate enclosures or shared systems if the campus is poorly planned. Model the actual phases and common infrastructure.
EU reporting is not a universal EU PUE limit
Commission Delegated Regulation (EU) 2024/1364 establishes information and KPI reporting for operators of data centers with installed IT power demand of at least 500 kW. It covers energy, water, renewable energy, and heat-reuse information and defines PUE, WUE, REF, and ERF calculations for that framework.
The regulation does not set one EU-wide PUE ceiling that every new data center must meet. The European Commission is continuing work on rating and minimum-performance policy; its current data-center energy-performance page distinguishes the existing reporting scheme from further rating and performance work.
National law can add requirements. Germany's Energy Efficiency Act, Section 11, for example, applies PUE, renewable-electricity, and planned energy-reuse requirements according to commissioning date and scope, with specified conditions and exceptions. Check the actual country, site, operator, load, commissioning date, and current law. A product name or modular format does not determine legal scope.
Certifications and voluntary schemes
Building, facility, management-system, and product credentials assess different objects. LEED or BREEAM can assess a building project under a named scheme and version. ISO 50001 concerns an energy-management system. EN 50600 and ISO/IEC 22237 address data-center facility infrastructure. The EU Code of Conduct for Data Centres is a voluntary initiative. None automatically proves a specific annual PUE, zero water consumption, renewable supply, or whole-life carbon result.
For every credential, ask:
- what object is assessed;
- who holds the certificate or registration;
- which site, design stage, scope, scheme, version, and level apply;
- whether the status is planned, submitted, awarded, current, or expired;
- which measured performance claims are supported separately.
Are modular data centers inherently greener?
No. Factory-built modular delivery can improve repeatability, quality control, phasing, testing, and material take-off. It can help match installed capacity to demand and allow site and factory work to proceed in parallel. Those are mechanisms worth evaluating, not proof of a better environmental result.
The comparison must also include enclosure materials, transport, foundations, external power and cooling plant, duplicated systems between phases, site work, expected utilization, service life, maintenance, and end of life. Compare a modular and conventional option with the same IT workload, availability objective, site boundary, and reporting period.
How ModulEdge approaches environmental performance
ModulEdge offers factory-built modular data-center platforms from 20 kW to 2.3+ MW per deployment unit, scalable to multi-megawatt capacity. Platform selection can support staged deployment, air or liquid-cooling architectures, free-cooling opportunities, dry or adiabatically assisted heat rejection, and heat-reuse interfaces where the equipment and site make them appropriate.
These architectural options do not establish a universal ModulEdge PUE, WUE, carbon saving, or heat-reuse percentage. Project performance depends on the IT equipment, utilization, climate, electricity supply, water strategy, controls, redundancy, and measurement boundary. A proposal should include the assumptions, simulation or calculation method, meters, and acceptance plan for any stated target.
Set measurable environmental requirements before comparing data-center proposals
Share the IT load and phasing, site climate, electricity and water conditions, operating model, applicable regulation, and performance priorities. ModulEdge can map them to a platform, plant concept, metering plan, and project-specific evidence.
Green data center procurement checklist
- IT workload, utilization, installed and peak load, phasing, and service life
- Annual facility and IT energy, PUE boundary, load profile, climate, and model method
- Electricity source and both location-based and market-based carbon where applicable
- Direct water inventory, consumption, WUE boundary, source, and scarcity context
- Refrigerant type, charge, leakage controls, maintenance, and recovery
- Heat-reuse temperature, demand, delivered energy, auxiliary energy, and ERF method
- Embodied-carbon and material boundary, quantities, product data, replacements, and end of life
- Applicable EU, national, and local reporting or performance requirements
- Meter locations, data ownership, retention, verification, and annual reporting process
- FAT, SAT, commissioning, and post-occupancy performance review
A green data center brief should make the environmental claims testable. If a supplier cannot state the boundary, method, conditions, and evidence behind a number, the number should not decide the project.
