For commercial, industrial, and institutional data center operators, energy costs are one of the largest and most controllable line items in the budget. Yet many facility teams still evaluate their operations without a clear, standardized way to measure how efficiently that energy is actually being used. That’s where data center PUE comes in.
Power usage effectiveness (PUE) is the industry’s most widely adopted benchmark for data center efficiency, and it has become the default way operators, engineers, and sustainability teams talk about performance. PUE is more than a reporting number. When you understand what actually drives it, PUE becomes a direct lens into your cooling system and a practical guide for where to invest in upgrades.
This guide breaks down:
- What PUE is and how it’s calculated
- What a good, average, and poor score means for your facility
- How cooling infrastructure decisions determine where you land on that scale
- Practical steps to improve your PUE value over time
What Is Data Center PUE?
Data center PUE is the ratio of the total amount of energy a facility consumes to the energy delivered specifically to IT equipment. In simpler terms, it’s the ratio of total facility energy to computing power, expressed as:
PUE = Total Facility Energy ÷ IT Equipment Energy
A PUE of 1.0 would mean every watt entering the facility goes directly to servers, storage, and networking equipment, with zero energy lost to cooling, lighting, or power delivery components. That’s a theoretical ideal. In practice, no data center achieves a perfect PUE ratio, because some energy is always required to support IT equipment, most notably for cooling.
The power usage effectiveness metric was developed by The Green Grid in 2007 and has since been adopted globally as the standard for measuring data center efficiency. According to Uptime Institute, average PUE across the industry has hovered in the 1.5 to 1.6 range for years, though top-performing facilities push well below that.
What makes PUE so useful is its simplicity. Facility managers don’t need complex modeling to understand it: the lower the number, the more efficiently a facility is running. Calculating it accurately, and using it to drive real improvements, requires understanding exactly what goes into both sides of that ratio.
How Data Center Energy Consumption Is Calculated
Calculating data center energy consumption sounds simple on paper, but getting an accurate PUE number requires knowing exactly where to measure and what to include. The formula only works if both sides of the ratio, total facility energy and IT equipment energy, are measured consistently and at the right points in the system.
Total facility energy should be measured at or near the facility’s utility meter. For a dedicated data center, this is typically the meter that powers the entire site, capturing every load, including IT equipment, cooling systems, lighting systems, and power delivery components. For a mixed-use facility where the data center shares a meter with other operations, teams often need to estimate and subtract the non-data center portion to isolate an accurate reading.
IT equipment energy should be measured after all power conversions, switching, and conditioning have occurred, typically at the rack level using power distribution units (PDUs). This captures the actual power delivered to servers, storage, and networking equipment, rather than power that gets lost or converted along the way.
A few practical considerations complicate measuring PUE in the real world:
- Subsystems present in one data center facility may not exist in another, making direct comparisons difficult
- Some loads, like cooling towers or chillers, support both data center and non-data center functions and can’t always be cleanly separated
- Instrumenting every PDU and control equipment component individually is often impractical or costly
- Utility meter readings may capture loads unrelated to the data center itself
Because of these variables, most operators track PUE as a trend for a single data center over time rather than as a fixed number to compare across facilities. A PUE that drops from 1.55 to 1.40 over two years tells you meaningfully more than comparing your facility’s PUE to a competitor’s, since climate, building materials, and IT load type all affect the ratio independently.
Equipment Energy vs. Facility Energy, Explained
The most common point of confusion in measuring power usage effectiveness comes down to what counts on each side of the equation. Here’s a simple breakdown:
| Category | Included in Calculation | Examples |
| IT Equipment Energy | Numerator’s counterpart; the “useful” energy | Servers, storage, networking equipment, control equipment like KVM switches |
| Total Facility Energy | Everything the facility draws | IT equipment energy plus cooling, lighting systems, power delivery components, air conditioning |
Getting this distinction right matters because misclassifying even one subsystem, like counting a shared chiller plant entirely as IT load, can throw off your PUE number and mask where energy is actually being spent. The equipment energy usage side should represent only what’s directly computing, storing, or moving data. Everything else, no matter how essential, falls into the “overhead” side of the ratio, and that overhead is exactly where cooling system upgrades have the most influence.
What Counts as a Good, Average, or Poor Data Center Efficiency Score?
Once you know how to calculate your PUE ratio, the next question is what that number actually means. A PUE value on its own doesn’t tell you much until it’s placed against industry benchmarks and your own facility’s history.
According to Uptime Institute’s ongoing global surveys, the average PUE across the industry has remained fairly flat for years, generally landing between 1.5 and 1.6. That average includes a wide mix of facility types, ages, and climates, so it’s a useful reference point, but not a target. Leading operators, especially those with modern cooling systems and tight controls, routinely achieve much lower numbers.
Here’s a general breakdown of what different PUE ranges signal operationally:
| PUE Range | Rating | What It Typically Means |
| 1.0 to 1.2 | Excellent | Highly optimized cooling, often using economization, containment, and precision controls. Common in newer, purpose-built facilities. |
| 1.2 to 1.4 | Good | Efficient practices in place, room for incremental gains. Often achieved with hot/cold aisle containment and right-sized chiller plants. |
| 1.4 to 1.6 | Average | In line with industry norms, but likely leaving cooling-related savings on the table. |
| 1.6 to 2.0 | Below Average | Significant energy is going to non-IT overhead. Aging equipment or inefficient airflow management are common causes. |
| 2.0+ | Poor | A large share of total power is being lost to cooling, power delivery, or other overhead. Usually signals outdated infrastructure or a facility never designed with efficiency in mind. |
A few things worth keeping in mind when interpreting your own data center efficiency metrics:
- Lower is better, but context matters. A commercial facility in a hot, humid climate will typically carry a higher PUE than one in a temperate region, even with comparable equipment and practices.
- Facility age and design intent play a large role. Older buildings retrofitted for computing equipment often carry structural limitations, like ceiling height or duct layout, that make achieving a low PUE more difficult without significant renovation.
- Workload density affects the ratio too. Facilities running high-density computing, such as AI or HPC workloads, may see PUE shift as cooling demands per rack increase, independent of any change in operational efficiency.
The most useful way to apply these benchmarks isn’t to chase a single “good” number, but to identify where your facility currently sits, then use that starting point to prioritize the cooling and infrastructure upgrades that will move you into the next tier. That’s where the connection between PUE and your cooling system becomes the most actionable part of this conversation.
Data Center Infrastructure Efficiency (DCiE) and Related Benchmarks
While PUE is the most widely used efficiency metric, it isn’t the only one worth understanding. Data center infrastructure efficiency, or DCiE, is essentially the inverse of PUE, expressed as a percentage rather than a ratio.
DCiE = (IT Equipment Energy ÷ Total Facility Energy) x 100
Or, more simply:
DCiE = 1 ÷ PUE
Where PUE tells you how much total energy a facility uses relative to its IT load, DCiE flips that relationship to show what percentage of total energy usage is actually reaching computing equipment. A facility with a PUE of 1.5, for example, has a DCiE of about 67%, meaning roughly two-thirds of the energy consumed goes toward IT equipment, and the remaining third supports cooling, lighting, and power distribution.
Neither metric is inherently better than the other. Some teams prefer DCiE because a percentage is easier to communicate to non-technical stakeholders, while others stick with PUE ratio reporting since it’s the more widely recognized industry standard. Both carry the same limitation: they’re most useful when tracked over time for one data center, rather than compared across different facilities.
Other Data Center Efficiency Metrics Worth Tracking
PUE and DCiE cover energy, but they don’t capture the full sustainability picture. A handful of related metrics have emerged to round out that view:
- Water usage effectiveness (WUE) measures the amount of water consumed to cool IT equipment, an increasingly important number for facilities using evaporative or liquid cooling systems.
- Carbon usage effectiveness (CUE) measures the ratio of carbon emissions tied to total energy consumption against the energy used by IT equipment, useful for teams reporting against sustainability targets.
- Energy reuse effectiveness (ERE) accounts for facilities that redirect waste heat for other uses, such as heating adjacent buildings.
These metrics matter most for organizations with formal sustainability reporting requirements, but for the majority of commercial, industrial, and institutional operators, PUE remains the most practical and actionable number to track day to day. It’s the metric most directly tied to operating costs, and as the next section covers, it’s also the metric most directly shaped by cooling system design.
How Your Cooling System Design Shapes PUE Performance
Cooling is consistently the largest source of non-IT energy consumption in a data center, which means it’s also the single biggest lever for improving PUE. For commercial, industrial, and institutional facilities in particular, where budgets and space constraints often differ from hyperscale campuses, the right cooling system decisions can make the difference between a mediocre PUE and a genuinely competitive one.
A few core cooling decisions have an outsized effect on where a facility lands on the PUE scale:
Chiller selection and sizing: Oversized chiller plants run inefficiently at partial load, wasting energy used to cool equipment that isn’t there yet or workloads that never materialize. Right-sizing chillers to actual and near-term IT load, rather than theoretical maximum capacity, is one of the most direct ways to lower overhead energy consumption.
Economization: Air-side and water-side economizers use outside air or water temperatures to reduce or eliminate mechanical cooling during favorable conditions. Facilities that can take advantage of more economizer hours throughout the year typically see a meaningful reduction in PUE, since mechanical cooling is one of the most energy-intensive systems in the building.
Airflow management and containment: Hot and cold aisle containment prevents supply and exhaust air from mixing, which lets cooling systems run at higher setpoints without risking equipment overheating. Poor airflow management forces cooling systems to work harder than necessary, negatively affects equipment reliability, and wastes energy cooling air that never reaches the servers that need it.
Liquid cooling: As rack densities climb, particularly with AI and high-performance computing workloads, liquid cooling is becoming a practical option for facilities that need to manage heat loads that air cooling can’t handle efficiently. Liquid cooling won’t automatically improve PUE on its own, but for high-density deployments, it can reduce the energy penalty that comes with trying to force traditional air cooling to keep pace.
Controls and monitoring: Modern building management systems (BMS) paired with smart PDUs give facility teams rack-level visibility into power consumption and cooling performance in real time. Without this visibility, inefficiencies in specific zones or equipment can go unnoticed for months.
The table below summarizes how each of these cooling decisions typically affects PUE:
| Cooling Strategy | Primary PUE Impact | Best Suited For |
| Right-sized chiller plants | Reduces energy waste from partial-load inefficiency | Facilities with known, stable IT load growth |
| Economization | Lowers mechanical cooling energy during favorable outdoor conditions | Facilities in temperate or cool climates |
| Hot/cold aisle containment | Reduces energy used to overcool due to air mixing | Nearly all commercial and industrial data centers |
| Liquid cooling | Manages high-density heat loads more efficiently than air alone | High-density compute, AI, and HPC workloads |
| BMS and smart PDU controls | Identifies waste and inefficiency in real time | Facilities looking to sustain improvements long-term |
None of these strategies work in isolation. A facility with excellent containment but an oversized, poorly controlled chiller plant will still carry a higher PUE than necessary. The facilities that consistently perform best treat cooling system design as an integrated strategy, not a series of individual upgrades.
Practical Ways to Reduce Data Center Power Through Smarter Cooling
Understanding the theory behind PUE and cooling is one thing. Actually improving your number requires a clear, prioritized set of actions. Below are practical steps commercial, industrial, and institutional operators can take to reduce data center power consumption and move toward a better PUE.
- Audit your current airflow management before spending on new equipment. Many facilities carry a higher PUE simply because hot and cold air are mixing somewhere in the room. Before investing in new chillers or controls, walk the floor and identify gaps in containment, unsealed cable cutouts, or misaligned perforated tiles. These low-cost fixes often deliver fast, measurable improvement.
- Right-size cooling capacity to actual IT load. If your facility was built for future growth that hasn’t materialized yet, your cooling system is likely running far less efficiently than it should. Reassess chiller and CRAC/CRAH capacity against current and near-term equipment load, not original design specifications.
- Expand economizer hours where climate allows. If your facility isn’t already taking advantage of air-side or water-side economization, this is often one of the highest-return upgrades available. Even facilities in moderate climates can typically capture more free cooling hours than they currently do.
- Implement or upgrade hot/cold aisle containment. For facilities without full containment, this remains one of the most cost-effective ways to lower PUE. Even partial containment, applied to the highest density rows first, can meaningfully reduce the energy needed to keep equipment within safe operating temperatures.
- Add rack-level monitoring and controls. Smart PDUs and BMS integration give facility teams the visibility needed to catch inefficiencies before they compound. Without rack-level data, it’s difficult to know whether cooling upgrades are actually translating into a lower PUE ratio or simply shifting where energy is being used.
- Evaluate liquid cooling for high-density zones. Facilities supporting AI, HPC, or other high-density workloads should evaluate whether liquid cooling makes sense for those specific zones, rather than the entire facility. A hybrid approach, air cooling for standard racks and liquid cooling for high-density rows, often delivers the best balance of cost and PUE improvement.
- Build a maintenance and controls calibration schedule. Cooling systems, even well-designed ones, drift out of optimal calibration over time. Regularly scheduled reviews of setpoints, sensor accuracy, and controls sequencing help ensure a facility doesn’t slowly regress to a higher PUE after an initial improvement.
A quick reference for prioritizing these upgrades:
| Priority Level | Action | Typical Cost | Typical Impact on PUE |
| High, low cost | Airflow audit and containment gap fixes | Low | Fast, measurable improvement |
| High, moderate cost | Hot/cold aisle containment buildout | Moderate | Significant, sustained improvement |
| Medium | Rack-level monitoring and controls | Moderate | Improves visibility, supports long-term gains |
| Medium to high | Economizer expansion | Moderate to high | Strong improvement in favorable climates |
| Project-level | Chiller right-sizing or replacement | High | Substantial, long-term improvement |
| Situational | Liquid cooling for high-density zones | High | High impact for specific high-density workloads |
These upgrades don’t need to happen all at once. Most facilities see the best results by starting with lower-cost airflow and containment fixes, then building toward larger capital projects like chiller replacement or expanded economization as budget allows.
Building a Long-Term Strategy for Sustainable Energy Efficiency
Improving PUE isn’t a one-time project. It’s an ongoing discipline that requires consistent measurement, regular reassessment, and a willingness to treat cooling system performance as a core part of facility operations, not a background utility function.
The most successful commercial, industrial, and institutional operators approach energy efficiency the same way they approach any other critical infrastructure priority:
- Establish a baseline and track it consistently. A single PUE calculation tells you very little. What matters is the trend over months and years, measured using consistent methodology so improvements (or regressions) are visible and attributable.
- Set realistic, facility-specific efficiency goals. A facility in a hot climate with an aging building envelope shouldn’t be held to the same target as a newly built, temperate-climate site. Benchmark against your own facility’s history first, industry averages second.
- Prioritize cooling system upgrades based on data, not assumptions. Rack-level monitoring and BMS data should guide which upgrades come first. The highest-impact opportunity is rarely the newest or most attractive equipment on the market. It’s usually the airflow gap, oversized chiller, or drifted setpoint that’s been quietly wasting energy for years.
- Revisit calibration and controls on a set schedule. Cooling systems don’t stay optimized on their own. Building in regular reviews protects the gains already made.
- Treat efficiency as a team-wide responsibility. Facility, IT, and operations teams all influence energy consumption in ways that show up in the PUE number. Cross-team visibility into how decisions affect cooling load helps prevent avoidable waste.
For operators serious about long-term efficiency goals, the path forward isn’t a single upgrade or a single number. It’s a sustained strategy where cooling system design, controls, and maintenance work together to keep total facility energy as close as possible to what’s actually needed to run IT equipment efficiently.
Data center PUE will likely remain the industry’s primary efficiency benchmark for years to come, and for good reason. It’s simple to understand, easy to track over time, and directly tied to the cooling decisions facility teams make every day. Getting cooling system design right isn’t just about lowering a number. It’s about running a more reliable, cost-effective, and efficient facility overall.
Talk to APA Technologies About Your Cooling Strategy
Improving PUE starts with understanding exactly how your cooling system is performing today. APA Technologies works with commercial, industrial, and institutional facilities to evaluate cooling infrastructure, identify where energy is being lost, and design upgrades that deliver measurable improvement, from containment retrofits to full chiller plant modernization.
If your facility’s PUE isn’t where it should be, our team can help you find out why and build a practical plan to fix it. Schedule a consultation today.