ASHRAE 90.4 Energy Standard for Data Centers

Executive Summary

ANSI/ASHRAE Standard 90.4, Energy Standard for Data Centers, is a performance-based standard that establishes the minimum energy-efficiency requirements for the design, construction, and planned operation and maintenance of data centres. It exists because data centres behave unlike any other commercial building: their IT equipment runs at near-constant load around the clock, the cooling and power systems dominate non-IT energy use, and the technology evolves faster than a prescriptive code can track. Standard 90.1 governs the building envelope, lighting, and service water heating; Standard 90.4 takes over for the specialised mechanical and electrical systems that make a building a data centre.

Compliance rests on two calculated metrics. The Mechanical Load Component (MLC) captures the efficiency of cooling, fans, pumps, and heat rejection. The Electrical Loss Component (ELC) captures losses through the power chain that feeds the IT equipment. Each is computed and then compared against a maximum value that varies by climate zone. A project complies when both calculated values are at or below their respective maxima — or when a documented tradeoff between the two, permitted under the standard’s alternative compliance path, achieves the same overall result.

What ASHRAE 90.4 Is — and Is Not

Standard 90.4 is a code-intended, performance-based energy standard developed as a companion to ANSI/ASHRAE/IES Standard 90.1. “Code-intended” means it is written in mandatory, enforceable language so that jurisdictions can adopt it directly into building energy codes. “Performance-based” means it sets a numerical efficiency target a design must meet, rather than dictating specific equipment — designers are free to reach the target however the project economics and site conditions allow.

Three things 90.4 is frequently mistaken for, and is not:

It is also worth being precise about what the standard governs. 90.4 regulates the efficiency of the systems that serve the data centre — cooling and the electrical distribution chain — not the efficiency of the IT equipment itself. The standard takes IT load as a given input and asks how efficiently the facility delivers power and cooling to it.

Why a Separate Data-Centre Standard Exists

Until the 2010s, data centres were treated as a category of commercial building under Standard 90.1. That fit poorly. A general office building is occupied during the day, lightly loaded at night, and dominated by envelope and lighting loads. A data centre is the opposite: it is effectively unoccupied, runs its IT load continuously, and is dominated by cooling and power-conversion losses. Applying envelope-and-lighting logic to a facility whose energy story is almost entirely mechanical and electrical produced requirements that were both ill-fitting and, in places, counterproductive to innovation.

ASHRAE initiated Standard 90.4P in 2013 to recognise the unique energy-performance profile of data centres, opening it for advisory public review that November. The first published edition followed in 2016. Since 2019, Standard 90.1 has referenced 90.4 as an alternative compliance path for computer rooms with an IT equipment load larger than 10 kW — formalising the division of labour between the two standards.

Scope and Applicability

Standard 90.4 applies to a data centre — or to a data-centre addition or alteration that requires new mechanical or electrical systems — that meets both of two thresholds:

A facility — or a discrete computer room within a larger building — that exceeds a power density of 20 watts per square foot of conditioned floor area and an IT equipment load of 10 kW falls within scope. Both conditions must be satisfied; a high-density room below 10 kW, or a large low-density space below 20 W/ft², is outside 90.4’s mandatory reach and is handled under 90.1.

The standard reaches three project types:

Edition History: 2016 → 2025

Standard 90.4 follows ASHRAE’s continuous-maintenance process, which allows interim changes through published addenda between full editions. Reading an older edition as though it were current is the most common technical error in the field, because both the compliance methodology and the maximum values have tightened materially over time.

The 2022 edition is the pivot point in the standard’s history: it is where the calculation methodology became genuinely part-load-aware on both the mechanical and electrical sides, and where the efficiency bar was raised across every climate zone. The 2025 edition is, by contrast, primarily a strategic-scope expansion — it opens the door to non-energy sustainability metrics in future editions without yet imposing them.

Governance and Maintenance

Standard 90.4 is developed and maintained by ASHRAE’s Standing Standard Project Committee (SSPC) 90.4, operating under a documented programme for the regular publication of addenda and revisions. The committee draws representation from across the data-centre ecosystem — owners, developers, design engineers, and equipment manufacturers — reflecting the consensus model that ANSI accreditation requires.

The standard carries dual designation as an ANSI/ASHRAE American National Standard, meaning each edition and each addendum is approved through ANSI’s consensus process before publication. Between full editions, the continuous-maintenance process lets the committee issue addenda — lettered a, b, g, and so on — that amend specific sections. Several addenda to 90.4-2022 were approved by ASHRAE and ANSI during 2024 and 2025, and the substance of those addenda fed into the 2025 edition.

The Two Compliance Metrics: MLC and ELC

Everything in 90.4’s compliance machinery reduces to two numbers. Both are ratios, both are compared to climate-zone maxima, and both must pass — unless the tradeoff path is invoked to balance one against the other.

The conceptual division is clean: the MLC answers “how much mechanical power does it take to remove the heat this IT load produces, in this climate?” and the ELC answers “how much of the electricity is lost on its way to the servers?” Cooling efficiency depends heavily on climate, so MLC maxima vary by zone; electrical-chain losses do not, so a single ELC bar applies everywhere.

Mechanical Load Component (Section 6)

The design Mechanical Load Component is defined as the sum of all cooling, fan, pump, and heat-rejection design power divided by the data-centre IT design power. It is, in effect, a measure of how much mechanical overhead the cooling system imposes per watt of IT load. A perfectly loss-free, free-cooled facility would approach an MLC of zero; a heavily mechanically cooled facility in a hot climate carries a higher MLC.

Design MLC versus annualised MLC

The standard distinguishes two MLC calculations:

• Design MLC — a peak-condition snapshot, evaluating the mechanical system at its design point.
• Annualised MLC — an energy-weighted figure that captures how the system actually performs across the year, including economiser hours and part-load operation.

Before the 2022 edition, the annualised calculation assumed the IT equipment always ran at 100%. The 2022 edition replaced that assumption with a part-load evaluation: the HVAC system’s energy use is now calculated at 25%, 50%, 75%, and 100% of the IT load, because cooling equipment efficiency changes substantially with load and a 100%-only assumption flattered systems that perform poorly at part load. The annualised MLC, once computed, is compared against the maximum annualised MLC for the climate zone, evaluated at 100% ITE load.

An important informative note in the standard frames the MLC maxima around realistic thermal assumptions: the values were developed for air pulled through IT equipment across a defined temperature rise and a design return-air temperature consistent with ASHRAE TC 9.9‘s thermal guidelines. Operating within the wider allowable temperature ranges those guidelines permit is one of the most direct levers a designer has on the MLC, and the 2022 edition was explicit that compliance can be achieved without an economiser when the thermal envelope is managed well.

Electrical Loss Component (Section 8)

The Electrical Loss Component measures losses through the electrical chain that delivers power to the IT equipment. It is calculated using the worst-case losses of each segment of the power chain, so that the figure demonstrates a guaranteed minimum level of electrically efficient design rather than a best-case marketing number.

The power-chain segments

Historically the ELC summed the losses of three segments: the incoming service segment, the UPS segment, and the IT-equipment distribution segment. The 2022 edition removed the incoming-service segment from the ELC calculation, on the grounds that it has minimal impact on overall distribution efficiency and that designers have limited ability to define it for a typical project — public-utility transformers and feeders are already accounted for in most projects. The ELC therefore now centres on the segments the designer actually controls.

Part-load congruence and transformer efficiency

The 2022 edition aligned the ELC methodology with the MLC by requiring all segment loss calculations to be made at 25%, 50%, 75%, and 100% of design load. This part-load congruence means a power chain must demonstrate efficiency across its operating range, not just at full load. The same edition brought transformer efficiency into the calculation: previously transformers needed only to meet U.S. Department of Energy minimums defined at 35% loading — an unrealistic operating point for most data centres — so the standard now evaluates them at the loads they actually see.

Illustrative depiction of where controllable electrical losses concentrate in the power chain — not the standard’s tabulated loss limits, which are read from the edition in force at 25/50/75/100% design load.

Climate Zones and Maximum Values

Because cooling opportunity depends on climate, the maximum MLC values are tabulated by ASHRAE climate zone — the same zone framework used throughout Standard 90.1. A data centre in a cool, dry zone with abundant economiser hours faces a tighter (lower) maximum MLC than one in a hot, humid zone, because free cooling is genuinely available there and the standard expects designs to exploit it.

The ELC, by contrast, applies a single maximum that does not vary by climate, because electrical-chain losses are essentially independent of outdoor conditions.

The practical consequence is that a design that complies in one climate zone may fail in another with no change to the equipment — the same MLC calculation is measured against a different bar. Climate-zone determination is therefore an early, consequential step in any 90.4 compliance exercise, and it follows the 90.1 zone definitions rather than introducing a separate scheme.

The Tradeoff / Alternative Compliance Path (Section 11)

Section 11 contains the standard’s defining flexibility: an alternative compliance path that allows tradeoffs between the MLC and the ELC. A design that exceeds its maximum ELC can still comply if it offsets the excess with a sufficiently better-than-required MLC, and vice versa — provided the combined result meets the maximum overall systems design value.

The logic is additive. The maximum MLC for the zone and the maximum ELC sum to a maximum overall systems design value. A design whose ELC runs slightly over its standalone limit can compensate with a mechanical system whose annualised MLC sits comfortably below its limit, so long as the two calculated values together stay at or below the overall maximum.

The 2022 edition also refined the scoping of the tradeoff so that a minor change — adding a single piece of cooling equipment — no longer forces a recalculation of MLC and ELC across the entire facility. Tradeoffs and sub-target allowances can be applied at the level of the affected segment or two, rather than triggering a full-facility re-evaluation.

Why 90.4 Does Not Use PUE for Compliance

Power Usage Effectiveness — total facility power divided by IT power — is the data-centre industry’s most recognised efficiency metric. It is also, by design, not the compliance metric in 90.4. Understanding why is central to understanding the standard.

PUE is an operational, whole-facility outcome measured on a running data centre. It is excellent for benchmarking and for operational reporting, but it is poorly suited to design-stage code compliance for several reasons. It depends on actual IT load, which is unknown and variable at design time. It can be gamed: loading up the IT equipment improves PUE arithmetically without making any system more efficient. And it bundles mechanical and electrical performance into a single number, hiding which subsystem is actually inefficient.

90.4 decomposes the problem instead. By separating the MLC (mechanical) from the ELC (electrical), it lets a code official see, and a designer demonstrate, exactly where efficiency is being delivered — and it sets each bar at a level that is verifiable from design documents before the facility is ever loaded.

Renewable Energy and Heat-Recovery Credits

The 2022 edition introduced, and the 2025 edition carries forward, a methodology under Section 11 for applying credits to the combined MLC and ELC. Two credit types are central.

On-site renewable energy credit

A design that deploys on-site renewable generation can apply a credit to the combined MLC and ELC, incentivising renewable deployment while still requiring the underlying mechanical and electrical systems to be energy-efficient in their own right. The credit is a complement to efficiency, not a substitute for it — a design cannot meet the standard purely by bolting on solar while running inefficient cooling.

Recovered-heat (heat-reclaim) credit

The standard gives specific credit for recovered heat shared with non-data-centre spaces — for example, using rejected data-centre heat to warm an adjacent building. The credit-calculation method is deliberately specific to prevent double-counting the same recovered energy. Heat recovery turns what is otherwise a pure thermal liability into a useful output, and the standard rewards designs that capture it.

For organisations whose corporate climate reporting depends on procured renewable electricity, it is worth separating two distinct accounting worlds. The 90.4 on-site renewable credit is a design-compliance instrument. Whether that same renewable generation can be claimed against corporate Scope 2 emissions is governed by entirely different rules — the market-based method in the GHG Protocol Scope 2 Guidance and the sourcing criteria in standards like RE100. A credit that helps a building pass 90.4 is not automatically a defensible Scope 2 reduction, and vice versa.

Relationship with ASHRAE 90.1

Standard 90.4 cannot be applied in isolation. It is a code-intended companion to ANSI/ASHRAE/IES Standard 90.1, and 90.1 is a normative reference within 90.4. The division of responsibility is deliberate and kept clean to avoid overlap:

Since the 2019 edition of 90.1, a computer room with an IT equipment load larger than 10 kW may be designed to 90.4 instead of 90.1’s prescriptive computer-room provisions. The two standards are revised on coordinated cycles — the 2022 edition of 90.4 cross-references the corresponding edition of 90.1 — so practitioners must hold both in view and confirm the editions are a matched pair under the adopted code. Treating 90.4 as a complete standalone code is one of the most consequential scoping errors a design team can make.

Relationship with the IECC and Code Adoption

The International Energy Conservation Code (IECC) is the model energy code adopted by most U.S. jurisdictions, and it interacts with 90.4 through Standard 90.1, which is the backbone of the IECC’s commercial provisions. The IECC references Section 6 and Section 8 of 90.4 as the compliance path for data centres, but the referenced edition lags the latest published edition, and the technical approach differs across IECC cycles.

The practical result is a patchwork. The edition of 90.4 that legally applies to a given project is determined by:

• Which edition of the IECC or ASHRAE 90.1 the jurisdiction has adopted;
• Which edition of 90.4 that code edition references;
• Any state or local amendments layered on top.

Some U.S. states reference 90.4 directly in their energy codes, often by adopting specific sections. Others reach it indirectly through the IECC. Because adoption lags publication, a newly published edition such as 90.4-2025 typically does not become enforceable in a jurisdiction until a future code cycle references it. The Authority Having Jurisdiction has the final word on which edition and which amendments apply.

What Changed in 90.4-2025

The headline change in the 2025 edition is not a tightened number — it is a widened purpose. The edition expands the standard’s purpose statement to enable the future inclusion of resources and considerations beyond direct energy use that contribute to data-centre sustainability and decarbonization, naming greenhouse-gas emissions, environmental impact, and water consumption as examples.

This is a strategic positioning move. Through the 2022 edition, 90.4 regulated energy efficiency and nothing else. The expanded 2025 purpose statement does not yet impose GHG, water, or environmental-impact limits, but it creates the standing for future editions or addenda to do so — aligning 90.4 with ASHRAE’s declared positions on decarbonizing the built environment and eliminating GHG emissions from it.

For sustainability teams, the signal is clear: the standard that governs data-centre system efficiency is being repositioned to eventually speak to the same carbon and water concerns that corporate disclosure regimes already demand. The energy-efficiency core remains, but the boundary of the standard is being drawn wider.

Compliance Workflow and Required Submittals

Demonstrating 90.4 compliance is a structured, document-driven exercise. The typical workflow for a new data centre runs as follows.

1. Confirm applicability. Verify the facility exceeds both the 20 W/ft² and 10 kW thresholds, and confirm with the AHJ which edition of 90.4 the adopted code enforces.
2. Determine the climate zone. Establish the ASHRAE climate zone (per the 90.1 zone framework) to identify the applicable maximum MLC.
3. Establish 90.1 compliance for non-data-centre aspects. Envelope, lighting, and service water heating are demonstrated under 90.1 in parallel.
4. Calculate the annualised MLC. Model HVAC energy at 25%, 50%, 75%, and 100% of ITE load; compute the annualised MLC and compare it to the climate-zone maximum.
5. Calculate the ELC. Sum worst-case segment losses (UPS and ITE distribution) at the four part-load points, including transformer efficiency; compare to the maximum ELC.
6. Apply tradeoffs and credits if needed. Where one metric exceeds its standalone limit, use the Section 11 path and any on-site renewable or recovered-heat credits to meet the maximum overall systems design value.
7. Assemble submittals. Document the basis of design and calculations, an electrical single-line diagram, ELC calculations showing as-designed and compliance values, and floor plans identifying the areas served by each distribution path.
8. Plan for operation and maintenance. The standard addresses planned operation and maintenance, so the compliance package includes the operational plan, not just the design snapshot.

Common Misinterpretations

Common Compliance Errors

1. Using the wrong edition. Designing to the latest published edition when the adopted code references an earlier one — or designing to an old edition out of habit when a newer one is enforced.
2. Calculating at full load only. Omitting the 25/50/75/100% part-load evaluation required from the 2022 edition for both MLC and ELC.
3. Mis-determining the climate zone. Applying the wrong maximum MLC because the climate zone was set incorrectly or inconsistently with the 90.1 framework.
4. Ignoring 90.1. Treating 90.4 as complete and omitting the parallel 90.1 demonstration for envelope, lighting, and service water heating.
5. Using best-case rather than worst-case ELC inputs. The ELC is built from worst-case segment losses by design; cherry-picking favourable points understates losses and invalidates the demonstration.
6. Including the incoming-service segment in a 2022+ ELC. That segment was removed from the ELC calculation in the 2022 edition; including it applies an outdated methodology.
7. Double-counting credits. Claiming the same on-site renewable generation or recovered heat in more than one place in the compliance demonstration, contrary to the anti-double-dipping credit methodology.
8. Incomplete submittals. Missing the single-line diagram, as-designed ELC values, or the four-point MLC calculation — the most frequent cause of failed plan review.

Interaction with GHG Accounting and Scope 2

90.4 is an energy-efficiency design standard, not a carbon-accounting standard — but for operators with corporate climate commitments, the two domains meet. The electricity a data centre consumes is, for most operators, the single largest line in the corporate greenhouse-gas inventory, and it sits almost entirely in Scope 2.

The connection runs through energy intensity. A facility designed to a tighter MLC and ELC consumes less non-IT energy per unit of IT load, which directly reduces the purchased electricity behind its Scope 2 emissions. The conversion from that electricity to emissions is then governed by grid emission factors and the dual-reporting requirements of the GHG Protocol Scope 2 Guidance — location-based factors reflecting the grid where the facility sits, and market-based factors reflecting contractual instruments such as renewable procurement. Operators model that conversion with tools like the Scope 2 electricity calculator and reference current grid intensities in the EPA eGRID dataset.

For data centres specifically, 90.4 also sits alongside the dedicated facility-design standards covered in the ISO/IEC 22237 and EN 50600 data-centre design reference — those address physical infrastructure, availability, and broader facility design, where 90.4 addresses the specific energy-efficiency compliance metrics. Read together, they give a design team the energy, resilience, and infrastructure picture for a modern data centre.

Criticisms and Limitations

90.4 is the most widely recognised data-centre energy standard, and it is not beyond legitimate critique. Four limitations recur.

It regulates the facility, not the IT

By design, 90.4 takes IT load as a given input and regulates only how efficiently the facility serves it. The largest single energy consumer — the IT equipment itself — sits outside the standard’s reach. A facility can fully comply while running highly inefficient servers, because server efficiency is not within scope. This is a deliberate boundary, but it means 90.4 compliance is not a complete statement about a data centre’s energy performance.

Adoption lag blunts its impact

Because the operative edition is set by locally adopted codes that lag publication, the efficiency gains in a new edition take years to become enforceable in most jurisdictions. The 2022 tightening, for instance, only binds projects in jurisdictions whose adopted code references it. The standard improves faster than the codes that give it legal force.

Calculation complexity

The part-load methodology introduced in 2022, while more accurate, raised the modelling burden — four-point evaluations for both MLC and ELC, transformer efficiency at realistic loads, and careful tradeoff accounting. For smaller projects the analytical effort can be disproportionate to the facility size, even though the 2022 scoping refinements eased the worst of the full-recalculation triggers.

Energy-only focus, until recently

Through 2022 the standard said nothing about carbon, water, or embodied impact — increasingly the metrics that matter most to operators and regulators. The 2025 purpose-statement expansion acknowledges this gap but does not yet close it with enforceable requirements.

Future Evolution

Three trajectories will shape 90.4 over the coming editions.

The decarbonization scope expansion. The 2025 purpose-statement change is the opening move toward editions that may address GHG emissions, water consumption, and broader environmental impact. The most consequential question for the next cycle is whether — and how — those considerations move from “enabled in the purpose statement” to “imposed as compliance requirements.”

Liquid cooling and high-density IT. The rapid growth of AI workloads is driving rack densities and direct-to-chip liquid cooling far beyond what air-cooled assumptions anticipated. Addenda to the 2022 edition already began regulating data-centre cooling systems more broadly in anticipation of water-cooled IT, and future editions will need to keep the MLC methodology meaningful as the thermal architecture shifts.

Tighter coordination with 90.1 and the IECC. As both parent documents evolve on their own cycles, keeping 90.4’s cross-references and the IECC’s adopted editions aligned remains an ongoing maintenance task — and the principal determinant of when a given edition’s efficiency gains actually reach projects.

Frequently Asked Questions

Sources and References

Every numerical claim and methodological statement on this page reconciles to the primary and authoritative sources below. Where ASHRAE has published a definitive document, it is cited directly; trade and engineering commentary is used only to corroborate interpretation.

Primary ASHRAE documents

• ANSI/ASHRAE Standard 90.4-2025, Energy Standard for Data Centers, ASHRAE, October 2025.
• ANSI/ASHRAE Standard 90.4-2022, Energy Standard for Data Centers, ASHRAE, 2022, and its published addenda (including Addenda a, b, and g, approved 2024–2025).
• ANSI/ASHRAE Standard 90.4-2019 and 90.4-2016, prior editions.
• ASHRAE, Standard 90.4-2022 Fact Sheet (Government Affairs Advocacy Toolkit), ashrae.org.
• BSR/ASHRAE Standard 90.4P public-review draft (definitions of design MLC, design ELC, and design PUE).
• ANSI/ASHRAE/IES Standard 90.1, Energy Standard for Sites and Buildings Except Low-Rise Residential Buildings — normative reference for envelope, lighting, and service water heating.

Corroborating engineering and code sources

• Consulting-Specifying Engineer, coverage of ASHRAE 90.4 compliance, MLC/ELC methodology, and IECC / 90.1 / 90.4 coordination.
• Data Center Dynamics, reporting on 90.4 edition updates and the scope definition (20 W/ft², 10 kW).
• International Energy Conservation Code (IECC), commercial provisions referencing ASHRAE 90.1 and Sections 6 and 8 of Standard 90.4.
• ASHRAE TC 9.9, Thermal Guidelines for Data Processing Environments, referenced by 90.4 for allowable temperature ranges.

Related GreenCalculus reference pages

• ISO/IEC 22237 and EN 50600 — Data-centre design and infrastructure
• GHG Protocol Scope 2 Guidance
• RE100 Technical Criteria
• ISO 50001 — Energy management
• EPA eGRID
• GHG Protocol Corporate Standard

What changed in this revision

Published 19 June 2026. Initial publication. Reflects ANSI/ASHRAE Standard 90.4-2025 (October 2025) as the latest published edition, the 90.4-2022 edition and its 2024–2025 addenda, and the normative relationship with ANSI/ASHRAE/IES Standard 90.1.