Energy and Sustainability

Data center energy audit: a step-by-step guide to cut OPEX 15% to 30% without replacing equipment

6 min read By Noxtel
Ilustración: Auditoría energética de data center: el paso a paso para reducir OPEX entre 15% y 30% sin cambiar equipo

A data center energy audit is the systematic process of measuring, analyzing, and optimizing the electrical consumption of the operation, with the goal of reducing OPEX without sacrificing capacity or availability. Mature operations typically uncover between 15% and 30% of electrical OPEX that can be recovered through operational, configuration, or management changes, without the need to purchase new equipment.

This article describes the step-by-step of an energy audit applied to a data center in Mexico during 2026, which consumption categories are analyzed, which metrics are used (PUE, WUE, CUE, and subsystem-specific metrics), which savings are realistic by operation type, and the common mistakes that lead to over-optimization without return. The goal is for the reader to finish with a framework applicable to their operation, not with a generic “turn off the lights” checklist.

Why energy audits often uncover 20% of undetected OPEX

Three reasons explain the typical gap between what an operation believes it consumes and what it actually consumes.

  • The first is the lack of metering per subsystem: many operations measure total utility feed consumption but not the individual consumption of UPS, cooling, lighting, network, or compute, which prevents identifying where the money goes.
  • The second is the confusion between contracted capacity and consumed capacity: under Mexican industrial tariffs, the maximum contracted demand is paid even when not used, which leads to under-contracting (paying surcharges) or over-contracting (paying for idle capacity).
  • The third is the baseline consumption of redundant systems: UPS, chillers, and generators operating in standby mode consume significant energy (between 3% and 8% of their nominal capacity) just by being on.

    • The typical outcome of a well-executed audit is identifying between 15% and 30% of recoverable electrical OPEX, with savings typically materializing between 3 and 12 months after the plan is implemented. The investment in the audit itself is typically paid back between 3 and 6 months.

    Step 1: Inventory capacity and actual load

    The first step is to build a detailed inventory: contracted capacity with CFE (Comisión Federal de Electricidad), installed capacity of UPS, chillers, CRACs, generators, PDUs, and the actual load (kW) of each subsystem measured with a power analyzer for 7 to 14 consecutive days to capture variation between business days and weekends. The measurement must be three-phase (across all three phases of the electrical supply) and per each critical subsystem.

    Tools: three-phase power analyzer with continuous recording (Fluke 435, Hioki PW3198, or equivalent), analysis software to export trends, and an inventory template that links each measurement to its source (which device it feeds). The cost of the measurement equipment is between 5,000 and 15,000 USD for a small operation; for large operations it is typically rented for 1,000 to 3,000 USD monthly.

    Step 2: Calculate actual PUE and compare it against design

    PUE (Power Usage Effectiveness) is calculated as the total facility energy divided by the IT equipment energy. A PUE of 1.5 means that for every kW of IT load, 1.5 kW total are consumed. Design PUE (documented when the data center was built) is typically between 1.3 and 1.4; actual PUE is typically between 1.5 and 1.8 because operations rarely load the facility to 100% as assumed in design.

    The difference between design PUE and actual PUE is the first source of potential savings. If actual PUE is 1.6 and design is 1.3, there are 0.3 PUE points recoverable, which in a 1 MW operation equals 300 kW of continuous savings, or between 250,000 and 350,000 USD annually in electricity. How is it recovered? Typically with better load management on the UPS (operating closer to 70% load, not 30%), higher room temperature (ASHRAE allows up to 27°C in class A1, many operations run at 19-21°C unnecessarily), and free cooling when available.

    Step 3: Analyze the consumption of each subsystem separately

    The third step is to break down consumption by subsystem: UPS, cooling, lighting, network, IT compute. For each subsystem, compare against benchmark (industry reference) and against design. Inefficiencies typically concentrate in four places: UPS operating below 30% load (efficiency degraded between 5 and 8 points), chillers sized for future capacity now running at low load (efficiency degraded between 10 and 20 points), lighting with old technology (LED reduces between 60% and 80% vs fluorescent), and network (inefficient or oversized switches and routers).

    Table of typical savings by subsystem:

    Step 4: Identify the baseline consumption of redundant systems

    The fourth step is to identify the consumption of systems operating in standby or redundant mode. A UPS in bypass consumes between 1% and 3% of its nominal capacity; an off chiller with active pump consumes 5% to 10% of its nominal; a generator in standby mode with jacket heating consumes between 1 and 3 kW continuously. Multiplied by the number of redundant devices, this baseline consumption can add up to between 5% and 15% of total operation consumption.

    The corrective action is to consciously decide how many redundant systems to keep active and how many to turn off or configure in economy mode. In Tier II operations, turning off redundant systems during low-load periods is viable. In Tier III/IV, the decision requires risk analysis, but there is typically room to configure automatic activation sequences based on demand.

    Step 5: Optimize the contracting with CFE

    The fifth step is to review the contracted tariff with CFE and optimize it. The industrial tariff in Mexico has several components: energy consumed ($/kWh), maximum demand ($/kW), and power factor. The most common mistake is contracting maximum demand far above actual consumption, which pays for idle capacity. The other common mistake is failing to manage power factor, which triggers surcharges when it drops below 0.9. Operating between 0.95 and 1.0 power factor avoids surcharges and, in some cases, generates bonuses.

    Tariff optimization typically requires a study with a certified electrical consultant who analyzes the consumption profile and proposes the best tariff combination (OM, HM, HMC, depending on the case). Typical savings are between 10% and 25% of total electricity bill.

    Common mistakes when implementing the savings plan

    Four mistakes account for most of the optimizations that fail to deliver the expected savings.

  • First, optimizing without measuring a baseline: without prior measurement, there is no way to know if the estimated saving materialized.
  • Second, focusing only on technology without changing operations: buying a more efficient UPS but continuing to operate at 25% load does not deliver the savings.
  • Third, failing to document the changes: without documentation, optimizations are lost when staff rotates.
  • Fourth, implementing everything at once without prioritizing: the savings plan should prioritize low-cost, high-impact actions (setpoint adjustment, UPS rebalancing) before larger investments (LED migration, chiller replacement).

    • How to start an energy audit in your data center

    Three steps to begin.

  • First, commit to a 14-day measurement with a power analyzer at the main utility feed and at least three critical subsystems.
  • Second, build an inventory of capacity vs. actual load using the measured data.
  • Third, prioritize corrective actions by cost and estimated return, and implement low-cost high-impact ones first.

    • A complete energy audit typically takes between 4 and 8 weeks and pays for itself between 3 and 12 months after the plan is implemented.

    Sources

    [1] Uptime Institute — Data center industry resources — https://uptimeinstitute.com/

    [2] The Green Grid — Data center efficiency industry resources — https://www.thegreengrid.org/

    [3] ASHRAE — Technical Resources (thermal management guidance) — https://www.ashrae.org/technical-resources

    [4] IEEE — Institute of Electrical and Electronics Engineers — https://www.ieee.org/

    Ready to operate like Noxtel?

    Take our online diagnostic — 8 questions, 3 minutes. Get your tier and an actionable report in your inbox.

    Take the free diagnostic

    Also in Energy and Sustainability

    ← Back to categories