A colocation operator in Johor announces that its newest facility has a power usage effectiveness of 1.3. The press release calls it one of the most efficient data centres in the region. The figure may be accurate. It may also be a design value calculated before the first server was installed, a reading taken on a cool night at full load, or an annual average that leaves out the office block next door. Each is a different claim, and only one of them tells a tenant, an investor or a regulator how the building actually performs.
That distinction now carries real commercial weight. Data centre electricity use is rising fast across Asia, governments are tying incentives and capacity approvals to efficiency metrics, and companies that buy cloud and colocation services are asking for data they can put in their own emissions inventories. A headline number is no longer enough. What matters is whether the number can be traced, repeated and compared.
Why Data Centre Measurement Has Become a Board-Level Issue
The International Energy Agency estimates that data centres used around 415 terawatt-hours of electricity in 2024, about 1.5% of global consumption, and projects that figure will more than double to around 945 TWh by 2030. The IEA also expects data centre electricity use in ASEAN to nearly double over the same period.
Malaysia shows the local impact. In September 2026, the Energy Commission said data centres accounted for 9.28% of the country’s total electricity consumption. The Knight Frank Data Centre Atlas 2026 put Johor’s live data centre capacity at 1,110 megawatts, with a further 8,542MW in the pipeline. The government has told Parliament it expects data centres to use around one-third of Malaysia’s electricity supply by 2035.
For operators, this means efficiency metrics now affect access to incentives, power and land. For tenants, a data centre’s performance flows directly into their own reported emissions. For lenders and investors, unreliable metrics are a credit and reputational risk.
What Singapore and Malaysia Are Actually Asking For
Regional requirements are often summarised loosely. The table below sets out what each instrument covers and what kind of obligation it is.
| Instrument | Key metrics | Who it applies to | Nature of requirement |
|---|---|---|---|
| Singapore Green Data Centre Roadmap (IMDA, May 2024) | Aim for all data centres to reach PUE of 1.3 or lower at 100% IT load over roughly ten years, alongside energy-efficient IT equipment | The Singapore data centre sector as a whole | A policy direction set with industry, not a statute with penalties |
| Singapore DC-CFA2 (EDB and IMDA, launched December 2025) | PUE of 1.25 or better at 100% IT load; BCA-IMDA Green Mark for Data Centres 2024 Platinum; IT equipment meeting SS 715:2025; at least 50% of new capacity powered by eligible green energy pathways | Applicants selected for new capacity allocation | Evaluation and allocation conditions |
| Malaysia Guidelines for Sustainable Development of Data Centre (MITI) | Design PUE of 1.4 or below for high-voltage hyperscale facilities above 21.25MW, with higher limits (1.6 or 1.7) for other categories; design WUE of 2.2 m³/MWh or lower; CUE declared under ISO/IEC 30134-8 | Applications to MIDA for tax incentives under the Digital Ecosystem Acceleration (DESAC) scheme received until 31 December 2027 | Conditions for incentive eligibility, including annual disclosure of actual WUE in the sustainability report |
Two points stand out. First, both countries rely heavily on design values at the approval stage, so operational evidence has to come later. Second, Malaysia’s guidelines expect organisations to keep monitoring and disclosing actual water performance and to improve towards 2.0 m³/MWh over ten years. Reporting is therefore expected to continue well after the ribbon-cutting.
The Metrics, and What Each One Cannot Tell You
The ISO/IEC 30134 series defines the main data centre key performance indicators. Each one measures something specific, and none of them, used alone, says a facility is sustainable.
| Metric | What it measures | What it does not show |
|---|---|---|
| PUE (Power Usage Effectiveness) | Total facility energy divided by IT equipment energy | Whether the servers themselves are efficient, how clean the electricity is, or how much water was used |
| WUE (Water Usage Effectiveness) | Water consumed per unit of IT energy | Local water stress, or whether the water was potable, reclaimed or recycled |
| CUE (Carbon Usage Effectiveness) | CO₂ emissions per unit of IT energy | Embodied carbon in construction, or refrigerant leakage, unless separately accounted for |
| REF (Renewable Energy Factor) | Share of data centre energy from renewable sources | Whether the renewable supply is matched in time and location to actual consumption |
| ERF (Energy Reuse Factor) | Share of energy reused outside the data centre, such as waste heat | Whether a real user for the heat exists in practice |
A facility can post an excellent PUE by using evaporative cooling, which cuts electricity but raises water consumption. Another can lower its PUE as tenants fill the halls, because fixed overheads are spread across more IT load, without any change to its cooling plant. That is why credible reporting presents metrics together rather than choosing the most flattering one.
The Data Centre Evidence Ladder
A practical way to judge any data centre sustainability claim is to ask which rung of evidence it reaches.
- Design value. Calculated from engineering models before operation. Useful for approvals, but a forecast rather than a result.
- Commissioning test. Measured under controlled conditions, often at or near full load. Shows the plant can perform, not that it does so day to day.
- Annualised operational value. Measured over twelve months of real operation, including hot seasons and partial occupancy. This is the first rung that describes actual performance.
- Trended and assured value. Several years of operational data, consistent boundaries, and review by an independent assurance provider.
- Exceptional and comparable result. Assured performance that clearly exceeds a recognised benchmark for a comparable facility type, climate and load.
Most marketing claims sit on rungs one or two. Most stakeholder decisions require at least rung three.
Five Details That Change the Number
1. IT load
Singapore’s benchmarks are stated at 100% IT load. A new facility running at 30% occupancy will usually report a worse PUE than its design value, and an honest report says so rather than quoting only the full-load figure.
2. Measurement period
Malaysia’s guidelines specify an annualised average. A single month in the cooler monsoon season is not equivalent to a full year in a tropical climate.
3. Boundary and metering points
Whether offices, chillers shared with other buildings, or transformer losses sit inside the boundary affects the result. The boundary should be stated and kept constant from year to year. Our guide to proving real energy savings covers baselines and boundary choices in more depth.
4. Emission factor
CUE depends heavily on the grid. Malaysia’s guidelines use 2021 grid emission factors of 0.758 tCO₂/MWh for Peninsular Malaysia and 0.198 tCO₂/MWh for Sarawak. Applying the guidelines’ formula (grid factor multiplied by PUE), an identical facility with a PUE of 1.4 would have a CUE of about 1.06 kgCO₂/kWh in Peninsular Malaysia and about 0.28 in Sarawak. The building is the same; the grid is not. Where renewable electricity is purchased, the report should state the basis used and the instruments behind it, as explained in our article on substantiating renewable energy claims.
5. Water source and stress
Malaysia’s guidelines encourage new sites in Peninsular Malaysia to avoid areas with a water stress index of 0.8 or above, and allow reclaimed or reused water. A WUE figure means more when it comes with the water source and local stress context, a point developed in our guide to corporate water stewardship.
An Evidence Checklist for Operators and Tenants
- State whether each figure is a design, commissioning or operational value.
- Report the measurement period, average IT load and facility boundary.
- Cite the standard used, such as the relevant part of ISO/IEC 30134.
- Publish PUE, WUE and CUE together, and explain any trade-offs.
- Disclose the emission factors and any renewable energy instruments used for CUE.
- Track refrigerant top-ups separately, since leakage emissions do not appear in PUE. See our guide to measuring refrigerant leakage.
- Keep metering calibration records and raw interval data available for audit.
- For tenants: request tenant-specific energy allocation data, not only the facility average.
Communicating Results Without Overstating Them
The simplest safeguard is to keep three kinds of statement separate.
- Commitment: “The facility is designed to achieve a PUE of 1.3 at full IT load.”
- Performance: “Over the 2026 financial year, at an average IT load of 62%, the facility recorded an annualised PUE of 1.42 and a WUE of 1.9 m³/MWh.”
- Exceptional achievement: “Independently assured results show the facility sustained a PUE below a stated regional benchmark for three consecutive years at comparable load.”
These example statements are illustrative, not data from any real facility. Terms such as “green data centre”, “carbon neutral” or “net zero” should only be used with the supporting evidence in the same place, not in a separate report that readers are unlikely to find.
When a Data Centre Milestone Becomes Record-Worthy
Some infrastructure achievements are exceptional enough to warrant independent recognition, and the evidence logic above applies there too. Scale and sustainability are different achievements. Amazon Web Services, for example, is an Asia Record holder for the largest cloud infrastructure footprint by a single provider in 2025. That is a recognition of regional reach, not of energy or water efficiency, and a sustainability claim would require its own measurement.
For operators, this distinction is useful. An efficiency or water result that sits at rung five of the evidence ladder (measured over a full year, assured, and clearly above a stated benchmark) is the kind of corporate record Asia’s recognition bodies can assess on its merits. Organisations wondering how to get an Asia Record for such a result should start with the evidence file rather than the announcement. Asia Record’s published verification process asks applicants to show how an achievement was measured and under what conditions. Record recognition in Asia complements, but never replaces, Green Mark certification, incentive approvals or regulatory compliance.
What to Ask Before the Next Announcement
Before a data centre efficiency figure appears in a press release, a tenant report or a financing document, five questions are worth asking. Is it a design or operational value? Over what period and at what load? Within which boundary? Using which emission factor? And has anyone independent checked it? A facility that can answer all five has a performance record. One that cannot has a target, which is worth stating honestly as a target.
