megawattindexAPI access

Data Center PUE by Operator and Campus: Published Benchmarks

Power usage effectiveness (PUE) is a data center's total energy divided by the energy that reaches its IT equipment, so 1.0 would mean no overhead at all. Lawrence Berkeley National Laboratory put the U.S. average at 1.40 in 2023. Among operators that publish a fleet figure for 2024 or later, PUE runs from 1.08 (Meta) to 1.59 (CyrusOne), and the lowest single-campus figure published is 1.04 (Google, Central Ohio (Lancaster), Ohio).

Below: a PUE calculator, the most efficient campuses by published PUE, every operator's fleet average, and 168 published PUE and water figures from 12 operators, each with its source and period.

Megawatt Index Editorial Team

Updated October 16, 2026Dataset and calculator168 published figures

U.S. average PUE
1.40
LBNL estimate, 2023
Lowest fleet PUE
1.08
Meta, 2024 annual
Lowest campus PUE
1.04
Google, Central Ohio (Lancaster), Ohio
Operators covered
12
168 published figures
Your PUE1.40Yearly cost of everything above the IT load$363,600At the U.S. average (LBNL, 2023)'s PUE of 1.40, the same IT load would carry$363,600 of overhead, $0 less a year

The default price is the U.S. average industrial rate for January to August 2026 (EIA). Starting energy values are an example.

Most efficient data centers by published PUE

Data center campuses ranked by the latest PUE their operator publishes
#CampusOperatorPUESource
1Central Ohio (Lancaster), OhioGoogle1.04Source
2Fort Wayne, IndianaGoogle1.05Source
3Omaha, NebraskaGoogle1.05Source
4Central Ohio (Columbus), OhioGoogle1.06Source
5Fredericia, DenmarkGoogle1.06Source
6The Dalles, Oregon (2nd facility)Google1.06Source
7Central Ohio (New Albany), OhioGoogle1.07Source
8Eemshaven, NetherlandsGoogle1.07Source
9Council Bluffs, Iowa (2nd facility)Google1.08Source
10Dublin, IrelandGoogle1.08Source
11Loudoun County, VirginiaGoogle1.08Source
12Loudoun County, Virginia (2nd facility)Google1.08Source
13Quilicura, ChileGoogle1.08Source
14St. Ghislain, BelgiumGoogle1.08Source
15The Dalles, Oregon (3rd facility)Google1.08Source

Campus figures only (regions and fleets are averages over many sites). Of the operators covered, only Google publishes PUE campus by campus, so every row is Google; the full list of 35 campuses is in the table further down. Fleet averages by operator: Meta 1.08, Google 1.09, Amazon Web Services 1.14, Microsoft 1.17, Digital Realty 1.20, Equinix 1.37, NTT DATA (Global Data Centers) 1.38, QTS 1.38, Iron Mountain Data Centers 1.41, CyrusOne 1.46.

Fleet averages operators publish

OperatorMetricValuePeriodSource
MetaPUE1.082024 annualSource
GooglePUE1.09trailing twelve months ending Q4 2025 (calendar 2025)Source
GooglePUE1.09trailing twelve months ending Q2 2026Source
Amazon Web ServicesPUE1.142025 annual (January 1 to December 31, 2025)Source
MicrosoftPUE1.17FY25 (July 1, 2024 to June 30, 2025)Source
SwitchPUE1.18not stated (undated web page claim marked with an asterisk that the page does not explain)Source
Digital Realty196 MW of IT capacity delivered in 2025 (subset)PUE1.2capacity delivered in 2025Source
Vantage Data CentersPUE1.262023 (latest ESG report found)Source
EquinixPUE1.372025 annualSource
Digital RealtyPUE1.382025 annualSource
NTT DATA (Global Data Centers)PUE1.38FY2024 (NTT DATA fiscal year; months not restated on page)Source
QTSPUE1.382025 annualSource
LBNL (US national estimate)United States, all data centersPUE1.42023 annualSource
Iron Mountain Data CentersPUE1.412025 annualSource
CyrusOnePUE1.462025 annualSource
CyrusOneStandard Dry Facilities (76% of critical load capacity)PUE1.462025 annualSource
CyrusOneNonstandard Wet Facilities (21% of critical load capacity)PUE1.472025 annualSource
CyrusOneNonstandard Dry Facilities (3% of critical load capacity)PUE1.592025 annualSource
CyrusOneStandard Dry Facilities (76% of critical load capacity)WUE02025 annualSource
CyrusOneNonstandard Dry Facilities (3% of critical load capacity)WUE02025 annualSource
Digital RealtySites with non-evaporative cooling (subset)WUE0.072025 annualSource
Amazon Web ServicesWUE0.122025 annual (January 1 to December 31, 2025)Source
MetaWUE0.192024 annualSource
CyrusOneWUE0.252025 annualSource
MicrosoftWUE0.27FY25 (July 1, 2024 to June 30, 2025)Source
LBNL (US national estimate)United States, all data centersWUE0.36through 2023Source
QTSWUE0.442025 annualSource
Digital RealtyWUE0.592025 annualSource
NTT DATA (Global Data Centers)WUE0.72FY2024 (NTT DATA fiscal year; months not restated on page)Source
Iron Mountain Data CentersWUE0.822025 annualSource
EquinixWUE0.912025 annualSource
CyrusOneNonstandard Wet Facilities (21% of critical load capacity)WUE1.352025 annualSource
EquinixData centers with evaporative cooling (subset)WUE1.412025 annualSource
NTT DATA (Global Data Centers)Evaporative sites only (subset)WUE1.42FY2024 (NTT DATA fiscal year; months not restated on page)Source
Digital RealtySites with evaporative cooling (subset)WUE1.732025 annualSource

Campus and regional figures

Showing 133 of 133

Campus and regional PUE and WUE published by operators
Amazon Web ServicesEurope (Ireland), IEWUE0.02
Amazon Web ServicesEurope (Stockholm), SEWUE0.02
MicrosoftEurope, Middle East & AfricaWUE0.03
Amazon Web ServicesCanada (West), CAWUE0.04
Amazon Web ServicesEuropeWUE0.05
Amazon Web ServicesAsia-Pacific (Melbourne), AUWUE0.06
Amazon Web ServicesU.S. East (Northern Virginia)WUE0.06
Amazon Web ServicesU.S. East (Ohio)WUE0.06
Amazon Web ServicesCanada (Central), CAWUE0.06
Amazon Web ServicesNorth AmericaWUE0.08
Amazon Web ServicesCentral/South AmericaWUE0.09
Amazon Web ServicesSouth America (Sao Paulo), BRWUE0.09
Amazon Web ServicesAsia-Pacific (Sydney), AUWUE0.10
Amazon Web ServicesEurope (Spain), ESWUE0.12
Amazon Web ServicesU.S. West (Oregon)WUE0.12
Amazon Web ServicesEurope (London), GBWUE0.14
Amazon Web ServicesEurope (Frankfurt), DEWUE0.17
Digital RealtyEMEAWUE0.19
MicrosoftAsia PacificWUE0.25
MicrosoftAmericasWUE0.34
Amazon Web ServicesU.S. West (Northern California)WUE0.39
Digital RealtyNorth AmericaWUE0.68
GoogleCentral Ohio (Lancaster), OhioPUE1.04
GoogleCentral Ohio (Lancaster), OhioPUE1.04
GoogleOmaha, NebraskaPUE1.05
GoogleFort Wayne, IndianaPUE1.05
GoogleOmaha, NebraskaPUE1.05
GoogleCentral Ohio (Columbus), OhioPUE1.06
GoogleCentral Ohio (New Albany), OhioPUE1.06
GoogleThe Dalles, Oregon (2nd facility)PUE1.06
GoogleCentral Ohio (Columbus), OhioPUE1.06
GoogleFredericia, DenmarkPUE1.06
GoogleThe Dalles, Oregon (2nd facility)PUE1.06
Amazon Web ServicesEurope (Spain), ESPUE1.06
GoogleEemshaven, NetherlandsPUE1.07
GoogleFredericia, DenmarkPUE1.07
GoogleCentral Ohio (New Albany), OhioPUE1.07
GoogleEemshaven, NetherlandsPUE1.07
Amazon Web ServicesAsia-Pacific (Melbourne), AUPUE1.07
GoogleCouncil Bluffs, Iowa (2nd facility)PUE1.08
GoogleDublin, IrelandPUE1.08
GoogleLoudoun County, VirginiaPUE1.08
GoogleLoudoun County, Virginia (2nd facility)PUE1.08
GoogleQuilicura, ChilePUE1.08
GoogleSt. Ghislain, BelgiumPUE1.08
GoogleCouncil Bluffs, Iowa (2nd facility)PUE1.08
GoogleDublin, IrelandPUE1.08
GoogleLoudoun County, VirginiaPUE1.08
GoogleLoudoun County, Virginia (2nd facility)PUE1.08
GoogleQuilicura, ChilePUE1.08
GoogleSt. Ghislain, BelgiumPUE1.08
GoogleThe Dalles, Oregon (3rd facility)PUE1.08
GoogleBerkeley County, South CarolinaPUE1.09
GoogleBristow, VirginiaPUE1.09
GoogleDouglas County, GeorgiaPUE1.09
GoogleHenderson, NevadaPUE1.09
GoogleMontgomery County, TennesseePUE1.09
GooglePapillion, NebraskaPUE1.09
GoogleBristow, VirginiaPUE1.09
GoogleDouglas County, GeorgiaPUE1.09
GoogleHamina, FinlandPUE1.09
GoogleHenderson, NevadaPUE1.09
GoogleMontgomery County, TennesseePUE1.09
GooglePapillion, NebraskaPUE1.09
Amazon Web ServicesEurope (Stockholm), SEPUE1.09
GoogleHamina, FinlandPUE1.10
GoogleJackson County, AlabamaPUE1.10
GoogleLenoir, North CarolinaPUE1.10
GoogleMidlothian, TexasPUE1.10
GoogleThe Dalles, OregonPUE1.10
GoogleBerkeley County, South CarolinaPUE1.10
GoogleInzai, JapanPUE1.10
GoogleLenoir, North CarolinaPUE1.10
GoogleThe Dalles, OregonPUE1.10
Amazon Web ServicesAsia Pacific (excl. China)WUE1.10
Amazon Web ServicesEurope (Ireland), IEPUE1.10
GoogleCouncil Bluffs, IowaPUE1.11
GoogleInzai, JapanPUE1.11
GoogleCouncil Bluffs, IowaPUE1.11
GoogleJackson County, AlabamaPUE1.11
GoogleMayes County, OklahomaPUE1.11
GoogleMidlothian, TexasPUE1.11
Amazon Web ServicesEuropePUE1.11
GoogleMayes County, OklahomaPUE1.12
GoogleSingapore, SGPUE1.12
GoogleSingapore, SGPUE1.12
Amazon Web ServicesU.S. East (Ohio)PUE1.12
Amazon Web ServicesU.S. West (Oregon)PUE1.12
GoogleChangua County, TaiwanPUE1.13
GoogleChangua County, TaiwanPUE1.13
Amazon Web ServicesCanada (West), CAPUE1.13
GoogleSingapore (2nd facility), SGPUE1.14
GoogleStorey County, NevadaPUE1.14
GoogleRed Oak, TexasPUE1.14
Amazon Web ServicesNorth AmericaPUE1.14
Amazon Web ServicesCentral/South AmericaPUE1.14
Amazon Web ServicesAsia-Pacific (Sydney), AUPUE1.14
Amazon Web ServicesSouth America (Sao Paulo), BRPUE1.14
GoogleSingapore (2nd facility), SGPUE1.15
GoogleStorey County, NevadaPUE1.15
Amazon Web ServicesU.S. East (Northern Virginia)PUE1.15
MicrosoftAmericasPUE1.16
MicrosoftEurope, Middle East & AfricaPUE1.16
Amazon Web ServicesU.S. West (Northern California)PUE1.18
Amazon Web ServicesAfricaPUE1.19
Amazon Web ServicesAfrica (Cape Town), ZAPUE1.19
Amazon Web ServicesCanada (Central), CAPUE1.19
Amazon Web ServicesAsia-Pacific (Tokyo), JPWUE1.22
Amazon Web ServicesEurope (London), GBPUE1.23
Amazon Web ServicesEurope (Frankfurt), DEPUE1.24
Amazon Web ServicesAsia Pacific (excl. China)PUE1.25
Amazon Web ServicesChina, CNPUE1.25
Amazon Web ServicesMiddle East (UAE)PUE1.25
Amazon Web ServicesAsia-Pacific (Tokyo), JPPUE1.25
Amazon Web ServicesChina (Ningxia), CNPUE1.25
Amazon Web ServicesMiddle East (Tel Aviv), ILPUE1.27
GoogleMesa, ArizonaPUE1.28
MicrosoftAsia PacificPUE1.28
Amazon Web ServicesMiddle EastPUE1.29
Amazon Web ServicesAsia-Pacific (Jakarta), IDPUE1.29
Amazon Web ServicesAsia-Pacific (Singapore), SGPUE1.30
Digital RealtyEMEAPUE1.31
Amazon Web ServicesMiddle East (Bahrain), BHPUE1.33
Amazon Web ServicesEurope (Paris), FRPUE1.35
Amazon Web ServicesAsia-Pacific (Bangkok), THPUE1.38
Digital RealtyAPACPUE1.38
Amazon Web ServicesAsia-Pacific (Mumbai), INPUE1.40
Digital RealtyNorth AmericaPUE1.41
Amazon Web ServicesAsia-Pacific (Hyderabad), INPUE1.43
Amazon Web ServicesAsia-Pacific (Osaka), JPPUE1.46
Amazon Web ServicesAsia-Pacific (Singapore), SGWUE1.57
Digital RealtyAPACWUE1.66
Amazon Web ServicesAsia-Pacific (Jakarta), IDWUE2.85
Coverage
Fleet, regional and campus figures from 12 operators' own reports and pages, mostly for 2024 and 2025, plus LBNL's national estimates.
Sources
Google Data Centers: Power usage effectiveness (efficiency page, 2026 and 2025 PUE yearly reports), Google 2026 Environmental Report (covers calendar year 2025), Meta Platforms: Meta 2025 Environmental Data Index (covers 2024), Microsoft Datacenters: Measuring datacenter energy and water use (PUE and WUE by region)
Last updated
, 168 published figures

How this was built

Power usage effectiveness is total data center energy divided by the energy delivered to IT equipment, as standardized in ISO/IEC 30134-2 and used by The Green Grid; 1.0 would mean every kWh reaches IT equipment (source: Mytton 2021, eq. 1, citing ISO/IEC 30134-2:2016; Lei et al. 2025 citing The Green Grid 2011). Water usage effectiveness, defined by The Green Grid (white paper WP#35, 2011), is annual site water usage divided by IT equipment energy, reported in liters per kWh of IT energy; it counts on-site water only, and The Green Grid's separate WUE source adds the water used to generate the electricity (source: Mytton 2021, eqs. 2 and 3, citing The Green Grid WP#35).

Every figure is copied as the operator publishes it, with its unit, period and basis (measured, design, trailing twelve months). Where a report does not state the unit for WUE, the unit column says so. We do not convert or estimate missing values. To turn a PUE into dollars, use the calculator above or the power cost calculator.

Limits of this data

  • Operators draw their measurement boundaries differently (which buildings, which losses, which period), so figures are not strictly comparable across companies.
  • Hyperscale operators run large owned campuses; colocation operators run many smaller, multi-tenant sites. Their PUEs differ partly for that reason.
  • WUE counts on-site water only. It leaves out the water used to generate the electricity; the water calculator shows both.
  • Several operators publish only a fleet average, and the Uptime Institute's survey average sits behind a registration form, so it is not included.

Sources

  1. 1.Google Data Centers: Power usage effectiveness (efficiency page, 2026 and 2025 PUE yearly reports)
  2. 2.Google 2026 Environmental Report (covers calendar year 2025)
  3. 3.Meta Platforms: Meta 2025 Environmental Data Index (covers 2024)
  4. 4.Microsoft Datacenters: Measuring datacenter energy and water use (PUE and WUE by region)
  5. 5.Microsoft 2026 Environmental Sustainability Report
  6. 6.Amazon Sustainability: AWS Cloud (PUE and WUE by geography and AWS Region)
  7. 7.Equinix 2025 Sustainability Report
  8. 8.Equinix 2025 Sustainability Data Summary
  9. 9.Digital Realty 2025 Impact Report
  10. 10.Iron Mountain Data Centers Sustainability Overview 2025
  11. 11.NTT DATA Group ESG Data: Environment Data
  12. 12.CyrusOne 2026 Sustainability Report (covers 2025)
  13. 13.QTS Data Centers: QTS 2025 Sustainability Report
  14. 14.Switch: Sustainability (web page)
  15. 15.Vantage Data Centers 2023 ESG Report
  16. 16.Lawrence Berkeley National Laboratory: 2024 United States Data Center Energy Usage Report (Shehabi et al., LBNL-2001637, December 2024)

Frequently asked questions

What is a good PUE for a data center?

Lower is better, and 1.0 is the theoretical floor. The lowest fleet figures operators publish are around 1.08, while LBNL estimated the U.S. average at 1.40 for 2023. Climate, size and cooling design all affect what is achievable.

Which data center has the best PUE?

Of the campus figures operators publish, the lowest is 1.04 at Google's Central Ohio (Lancaster), Ohio campus (trailing twelve months ending Q2 2026). Most operators publish only fleet or regional averages, so a campus that is not listed may be as efficient without saying so.

How do you calculate PUE?

Divide the total energy the facility used over a period by the energy its IT equipment used over the same period. Measure both over a full year so seasonal cooling loads are included.