13 · Evidence Observatory
Every number, graded
The Atlas treats evidence quality as a first-class dimension. This page is the audit trail: which fields are reported by operators, which are documented estimates, which are class-band inferences, and where the dataset is simply silent.
Documented parameters
0
fields with a provenance record
Mean field confidence
0.00
of 5 across all parameters
Quality-control checks
0
reconciliations against published data
Checks flagged
0
divergences documented rather than adjusted
Confidence distribution across parameters
Count of documented fields in each confidence band. The bulk of this dataset is modelled, not reported.
Reading the grades
What each confidence level means
4–5 · reported
Published by the operator or a regulator, in a named document.
3 · documented estimate
Derived from a credible published figure with a stated method.
2 · modelled
Computed from documented coefficients applied to this facility.
1 · class-band inference
Assigned from a facility-class or land-use band, not this facility.
0 · no record
No value located. Shown as an em dash, never as a zero.
Structural silence
The field does not exist in the dataset: Scope 1, Scope 3, measured noise, metered energy.
Audit trail
Provenance
| Field / parameter | Unit | Evidence type | Confidence | Derivation | Source | Limitation |
|---|---|---|---|---|---|---|
| Data_Centre_ID | id | Derived | 5.0/5 | ISO3 country code + sequential number. Stable primary key across every sheet. | This workbook | Not an external identifier; do not join to third-party datasets on it. |
| Data_Centre_Name | text | Reported | 4.0/5 | Facility name as published by the operator or a market directory. | Operator sites; datacentermap.com market directories | Naming conventions differ between operators; some campus and building rows share a stem. |
| Operator | text | Reported | 4.0/5 | Owning or operating company at the time of compilation. | Operator disclosures and market directories | Ownership changes frequently (M&A, JV, sale-leaseback). Snapshot as at 2026-08-25. |
| Record_Type | category | Derived | 5.0/5 | Facility = a single data centre building. Campus = a multi-building parent, EXCLUDED from all sums. | This workbook | Campus rows exist only to preserve the parent-child structure. |
| Country / Region / State_Province / City | text | Reported | 4.0/5 | Administrative geography of the facility's area. | Facility addresses | City is the administrative city of the AREA, which for metro-fringe sites may differ from the marketed city. |
| Latitude / Longitude | decimal degrees | Proxy | 2.0/5 | Centroid of the facility's industrial estate, business park or suburb. | Area reference table, sheet 52 | AREA CENTROID, NOT a rooftop coordinate. +/-1-3 km typically, +/-5-15 km for city-centroid records. |
| Coordinate_Precision | category | Derived | 5.0/5 | Declares which of the two precision bands the coordinate belongs to. | This workbook | — |
| Status | category | Reported/Inferred | 3.0/5 | Operational, Under construction, Announced, Planned or Unknown. | Market directories and operator announcements | Directories do not always disaggregate status; some Operational flags are inferred from listing context. |
| Year_Online | year | Reported or Modelled | 1.0/5 | Reported commissioning year where found; otherwise back-cast (Methodology section 6). | Operator announcements; sourced national capacity curves | Confidence 1 for every modelled year. Do not use for individual-site vintage claims. |
| Capacity_MW_IT_2026 | MW | Reported or Allocated | 2.0/5 | Energised IT-load capacity. Reported where published; otherwise market-anchor allocation (Methodology section 3). | Market anchors in sheet 51; operator disclosures | An allocation for 1094 of 1098 facilities. Individual values are indicative; market and cluster sums are the reliable level. |
| Capacity_MW_Ultimate | MW | Allocated | 2.0/5 | Capacity at full build-out of the identified pipeline, before the status energisation factor. | As above | Assumes announced and planned schemes complete at the same class weight as built stock. |
| PUE | ratio | Reported or Proxy | 2.0/5 | Operator fleet/design PUE where published; otherwise the country-climate benchmark. | Operator sustainability reports; Uptime Institute regional averages | A fleet average applied to one site. Real site PUE varies with load factor, age and season. |
| Load_Factor | fraction | Modelled | 2.0/5 | Average IT-load utilisation by facility class (Hyperscale 0.80, Wholesale 0.65, Mid colo 0.60, Small 0.50). | Operator disclosures; JLL/CBRE colocation occupancy; LBNL 0.50 fleet assumption | Single annual constant; ignores ramp-up on newly commissioned halls. |
| Energy_Consumption_MWh_yr | MWh/yr | Modelled | 2.0/5 | Capacity x 8,760 x Load_Factor x PUE. | Derived in this workbook | No site is metered here. Error compounds capacity error, load-factor error and PUE error. |
| Energy_Intensity_MWh_per_MW_yr | MWh/MW/yr | Derived | 2.0/5 | 8,760 x Load_Factor x PUE. | Derived | Varies only by class and country in this build; it is not a site-specific efficiency measure. |
| Grid_Electricity_Dependence_pct | % | Proxy | 1.0/5 | Set to 100 for every facility. | Absence of published behind-the-meter generation data | No public evidence of material on-site generation was located at any facility. Treat as an upper bound. |
| Renewable_Grid_pct | % | Reported | 3.0/5 | National grid-average renewable share of GENERATION. | Ember / national statistical agencies | A physical grid mix, NOT contracted or matched procurement. Do not confuse with the next field. |
| Renewable_Matched_pct | % | Reported | 4.0/5 | Operator-published renewable matching / carbon-free-energy percentage, where one exists. | Operator sustainability reports | Annual matching, not hourly delivery, except where the operator states otherwise (Google reports hourly CFE). |
| Water_Consumption_m3_yr | m3/yr | Modelled | 2.0/5 | Energy_MWh x WUE. | Derived | On-site cooling water only. Excludes upstream water embodied in electricity generation, which is often larger. |
| WUE | m3/MWh (= L/kWh) | Reported or Proxy | 2.0/5 | Operator fleet WUE where published; otherwise country-climate benchmark. | Operator reports; source workbook benchmarks | UK, EU and Irish values are climate proxies with no primary national data-centre figure. Singapore's is a regulatory CEILING. |
| Water_Stress_Score / Label | 0-5 / category | Proxy | 2.0/5 | WRI Aqueduct-informed baseline water-stress band for the facility's basin. | WRI Aqueduct Water Risk Atlas, area-level interpretation | NOT a real Aqueduct basin lookup against the facility coordinate. Assigned per reference AREA. |
| Water_Stress_Weighted_Demand_m3 | m3/yr | Derived | 2.0/5 | Water_m3 x (water-stress score / 5). A scarcity-adjusted demand. | Derived | Linear scarcity weighting is a convention, not a hydrological model. |
| Cooling_Water_Dependence | category | Modelled | 2.0/5 | Banded from WUE: >1.5 evaporative-dominant, 0.5-1.5 hybrid, <0.5 air-cooled/closed-loop. | Derived from WUE | Inferred from a water ratio, not from the installed plant. |
| Carbon_Emissions_Location_tCO2e_yr | tCO2e/yr | Modelled | 2.0/5 | Energy_MWh x national grid emission factor. | National grid emission factors, sheet 51 | LOCATION-BASED Scope 2. National averages hide large sub-national variation (~6x across US eGRID subregions). |
| Carbon_Emissions_Market_tCO2e_yr | tCO2e/yr | Modelled | 3.0/5 | Energy_MWh x grid factor x (1 - operator matched-renewable %). | Operator disclosures | Only populated where the operator publishes a matching percentage. Annual matching is not hourly decarbonisation. |
| Emission_Intensity_tCO2e_per_MWh | tCO2e/MWh | Reported | 3.0/5 | The national grid emission factor applied to that facility. | National regulators / EPA / CEA / UBA / EMA | Uniform within a country by construction. |
| Waste_Heat_MW_thermal / _TJ_yr | MWth / TJ/yr | Derived | 3.0/5 | Energy_MWh / 8,760, and the same energy expressed as TJ per year. | Energy-balance proxy, same method as the supplied workbook | An average annual heat-rejection RATE, not a measured local temperature effect. |
| Cooling_Technology | category | Modelled | 1.0/5 | Inferred from WUE band. | Derived from WUE | A guess at plant type from a water ratio. Replace with planning-application data where available. |
| Mean_Annual_Temperature_C | deg C | Proxy | 3.0/5 | Long-run mean annual air temperature for the facility's area. | Climate normals, area-level | Annual mean; ignores the summer peak that actually drives cooling and heat-island effects. |
| Heat_Island_Proxy | MWth x k-persons/km2 | Derived | 1.0/5 | Waste_Heat_MW x resident density / 1,000. | Derived | A composite index with no physical units meaning. Comparative only. |
| Noise_dB | dB(A) @ ~50 m | Modelled | 1.0/5 | 50 + 10 x log10(Capacity_MW), clipped to [45,85]. | Engineering proxy anchored to the Prince William County 60 dB(A) ordinance threshold | One of the two weakest fields here. No facility in this workbook has a measured noise value. |
| Residential_Distance_m | m | Modelled | 1.0/5 | Land-use class band. | Anchored to Virginia JLARC 2024 (~1/3 of VA data centres within 200 ft of residential zoning) | A class band, not a measured setback. Real setbacks vary by an order of magnitude within a class. |
| Land_Use_Context / Sensitivity | category / 0-1 | Proxy | 2.0/5 | Land-use class of the facility's area and its community-sensitivity weight. | Area reference table, sheet 52 | Assigned per area, so every facility in an estate shares a value. |
| Population_1km / Population_5km | persons | Modelled | 1.0/5 | Density band x pi x r^2, with a 0.60 radial decay factor applied at 5 km. | Area density bands, sheet 52 | NOT a raster zonal statistic. Replace with GHSL or WorldPop against real coordinates before publication. |
| Community_Complaints | count/yr | Not Available | 0.0/5 | 0 for every facility. | No public record located | 0 MEANS 'no publicly documented facility-specific complaint was found'. It does NOT mean none exist. |
| Regulatory_Planning_Constraint | 0-1 | Reported | 3.0/5 | Cluster-level score for restrictive instruments in force (moratoria, connection bans, zoning frameworks). | Regulator and grid-operator publications, cited per cluster | Cluster-level, so it cannot distinguish two facilities in the same market. |
| Documented_Objection_Record | 0-1 | Reported | 3.0/5 | Cluster-level score for the strength of the public objection / litigation record. | Legislative audits, planning decisions, litigation reporting | Strong for the USA and Ireland; near-absent for India, Brazil and Singapore - an evidence gap, not an absence of impact. |
| Local_Infrastructure_Pressure_pct | % | Derived | 2.0/5 | Facility capacity as a share of its cluster's total capacity. | Derived | A concentration measure, not a grid-headroom measure. |
| Employment_Estimate_FTE | jobs | Modelled | 1.0/5 | Capacity_MW x 5 permanent roles per MW. | Industry rule of thumb | Highly uncertain and varies enormously between colocation and hyperscale operating models. |
| P1..P6 pillar scores | 0-100 | Derived | 2.0/5 | Weighted sum of normalised sub-indicators (Methodology section 5). | Derived | Inherits the confidence of its weakest input; P5 and P6 are Confidence 1-2 throughout. |
| Composite_Score / Risk_Class | 0-100 / category | Derived | 2.0/5 | Weighted mean of the six pillars; class = quartile of this dataset. | Derived | Relative to this dataset only. Not an absolute risk statement about any facility. |
| Record_Confidence | 0-5 | Derived | 5.0/5 | Mean of the field confidences for capacity, PUE, WUE, coordinates, status, year and context. | Derived | A summary; always check the per-field confidence columns before relying on a single value. |
| Grid emission factor (India) | tCO2e/MWh | Reported | 3.0/5 | 0.710 | CEA CO2 Baseline Database for the Indian Power Sector (weighted average operating margin ~0.71 tCO2/MWh) | National annual average; masks large sub-national and hourly variation. |
| Grid renewable share (India) | % | Reported | 3.0/5 | 22.0 | Ember / CEA generation mix ~2024 (renewables incl. large hydro as % of generation) | Physical generation mix, not contracted procurement. |
| Benchmark PUE (India) | ratio | Proxy | 2.0/5 | 1.60 | Uptime Institute Global Data Center Survey - Asia-Pacific regional average; same value as the source workbook | Regional/industry average applied to individual sites. |
| Benchmark WUE (India) | m3/MWh | Proxy | 2.0/5 | 2.15 | Source workbook India WUE (2.15 m3/MWh), derived from Indian colocation disclosures | See the source note - several of these are climate proxies or regulatory ceilings, not measurements. |
| Grid emission factor (USA) | tCO2e/MWh | Reported | 3.0/5 | 0.386 | US EPA eGRID national annual average generation emission factor | National annual average; masks large sub-national and hourly variation. |
| Grid renewable share (USA) | % | Reported | 3.0/5 | 23.0 | EIA / Ember US generation mix ~2024 | Physical generation mix, not contracted procurement. |
| Benchmark PUE (USA) | ratio | Proxy | 2.0/5 | 1.50 | Uptime Institute Global Data Center Survey - North America regional average | Regional/industry average applied to individual sites. |
| Benchmark WUE (USA) | m3/MWh | Proxy | 2.0/5 | 0.37 | LBNL 2024 US Data Center Energy Usage Report: 17.4 bn gal direct cooling water / 176 TWh (2023) | See the source note - several of these are climate proxies or regulatory ceilings, not measurements. |
| Grid emission factor (UK) | tCO2e/MWh | Reported | 3.0/5 | 0.207 | UK DESNZ/DEFRA Greenhouse Gas Conversion Factors, grid electricity generation factor | National annual average; masks large sub-national and hourly variation. |
| Grid renewable share (UK) | % | Reported | 3.0/5 | 45.0 | DESNZ Energy Trends / Ember GB generation mix ~2024 | Physical generation mix, not contracted procurement. |
| Benchmark PUE (UK) | ratio | Proxy | 2.0/5 | 1.50 | Uptime Institute Global Data Center Survey - Europe regional average | Regional/industry average applied to individual sites. |
| Benchmark WUE (UK) | m3/MWh | Proxy | 2.0/5 | 1.00 | Climate-based WUE proxy carried from the source workbook - WEAK EVIDENCE, no primary UK national figure | See the source note - several of these are climate proxies or regulatory ceilings, not measurements. |
| Grid emission factor (Ireland) | tCO2e/MWh | Reported | 3.0/5 | 0.280 | SEAI Energy in Ireland - electricity CO2 intensity | National annual average; masks large sub-national and hourly variation. |
| Grid renewable share (Ireland) | % | Reported | 3.0/5 | 40.0 | EirGrid / SEAI renewable share of electricity generation ~2024 | Physical generation mix, not contracted procurement. |
| Benchmark PUE (Ireland) | ratio | Proxy | 2.0/5 | 1.40 | Cool-maritime-climate benchmark; free-cooling-dominant Dublin hyperscale fleet | Regional/industry average applied to individual sites. |
| Benchmark WUE (Ireland) | m3/MWh | Proxy | 2.0/5 | 0.80 | Cool-climate WUE proxy (free cooling reduces evaporative demand) - MODELLED, no Irish national figure | See the source note - several of these are climate proxies or regulatory ceilings, not measurements. |
| Grid emission factor (Germany) | tCO2e/MWh | Reported | 3.0/5 | 0.380 | German Environment Agency (UBA) specific CO2 emissions of the German electricity mix | National annual average; masks large sub-national and hourly variation. |
| Grid renewable share (Germany) | % | Reported | 3.0/5 | 55.0 | AGEB / Fraunhofer ISE renewable share of public net electricity generation ~2024 | Physical generation mix, not contracted procurement. |
| Benchmark PUE (Germany) | ratio | Proxy | 2.0/5 | 1.45 | Uptime Institute Europe regional average adjusted for the Frankfurt free-cooling window | Regional/industry average applied to individual sites. |
| Benchmark WUE (Germany) | m3/MWh | Proxy | 2.0/5 | 0.90 | Temperate-climate WUE proxy - MODELLED, no German national data-centre figure | See the source note - several of these are climate proxies or regulatory ceilings, not measurements. |
| Grid emission factor (Brazil) | tCO2e/MWh | Reported | 3.0/5 | 0.110 | MCTI / ONS Brazilian national grid emission factor (hydro-dominant system) | National annual average; masks large sub-national and hourly variation. |
| Grid renewable share (Brazil) | % | Reported | 3.0/5 | 87.0 | ONS / EPE Brazilian generation mix ~2024 | Physical generation mix, not contracted procurement. |
| Benchmark PUE (Brazil) | ratio | Proxy | 2.0/5 | 1.70 | Source workbook Brazil PUE (1.7) - warm-humid climate colocation benchmark | Regional/industry average applied to individual sites. |
| Benchmark WUE (Brazil) | m3/MWh | Proxy | 2.0/5 | 2.15 | Source workbook Brazil WUE (2.15 m3/MWh) | See the source note - several of these are climate proxies or regulatory ceilings, not measurements. |
| Grid emission factor (Singapore) | tCO2e/MWh | Reported | 3.0/5 | 0.400 | Singapore EMA grid emission factor (gas-dominant system) | National annual average; masks large sub-national and hourly variation. |
| Grid renewable share (Singapore) | % | Reported | 3.0/5 | 5.0 | EMA Singapore Energy Statistics - solar share of generation | Physical generation mix, not contracted procurement. |
| Benchmark PUE (Singapore) | ratio | Proxy | 2.0/5 | 1.40 | IMDA Green Data Centre Standard / Singapore new-build PUE requirement (<=1.3 for new; 1.4 fleet benchmark) | Regional/industry average applied to individual sites. |
| Benchmark WUE (Singapore) | m3/MWh | Proxy | 2.0/5 | 2.00 | Singapore regulatory WUE CEILING (<=2.0 m3/MWh) - a ceiling, not a measured mean; likely overstates actual use | See the source note - several of these are climate proxies or regulatory ceilings, not measurements. |
| Operator disclosure (Google) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.09 | WUE 0.99 | matched renewable 64.0 | Google Environmental Report - fleet-wide average annual PUE, average WUE and carbon-free energy percentage | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (Meta) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.09 | WUE 0.2 | matched renewable 100.0 | Meta Sustainability Report - fleet PUE/WUE and 100% renewable-energy matching claim | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (Microsoft) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.18 | WUE 0.3 | matched renewable 100.0 | Microsoft Environmental Sustainability Report - design PUE/WUE and 100% renewable purchase claim | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (Amazon AWS) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.15 | WUE 0.18 | matched renewable 100.0 | AWS Sustainability - global WUE 0.18 L/kWh and 100% renewable-matched claim | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (Equinix) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.42 | WUE None | matched renewable 96.0 | Equinix Sustainability Report - global portfolio PUE and renewable-coverage percentage | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (Digital Realty) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.45 | WUE None | matched renewable 65.0 | Digital Realty ESG Report - global portfolio PUE and renewable-energy coverage | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (STT GDC India) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.5 | WUE None | matched renewable None | STT GDC India sustainability disclosures - India portfolio design PUE | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (CtrlS Datacenters) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.42 | WUE None | matched renewable None | CtrlS published design PUE for its Indian Tier-IV fleet | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (Yotta Data Services) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.5 | WUE None | matched renewable None | Yotta NM1 published facility PUE (datacentermap facility specification) | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Operator disclosure (AirTrunk) | PUE / WUE / CFE | Reported | 4.0/5 | PUE 1.25 | WUE None | matched renewable None | AirTrunk published design PUE for its APAC hyperscale fleet | A FLEET-WIDE figure applied to each of that operator's individual facilities in this workbook. |
| Market capacity anchor (IN-Mumbai Metropolitan Region) | MW IT load | Reported / Derived | 3.0/5 | 595 MW operational | Cushman & Wakefield India Data Centre Update H1 2025 - Mumbai metro operational IT load (~53% of national) | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IN-Chennai) | MW IT load | Reported / Derived | 3.0/5 | 191 MW operational | Cushman & Wakefield India Data Centre Update H1 2025 - Chennai operational IT load | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IN-Delhi NCR) | MW IT load | Reported / Derived | 3.0/5 | 146 MW operational | Cushman & Wakefield India Data Centre Update H1 2025 - Delhi-NCR operational IT load | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IN-Pune) | MW IT load | Reported / Derived | 3.0/5 | 112 MW operational | Cushman & Wakefield India Data Centre Update H1 2025 - Pune operational IT load | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IN-Hyderabad) | MW IT load | Reported / Derived | 3.0/5 | 152 MW operational | Residual of the Cushman & Wakefield national total (1,280 MW, H1 2025) after the six named metros - DERIVED, not directly reported | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IN-Bengaluru) | MW IT load | Reported / Derived | 3.0/5 | 76 MW operational | Cushman & Wakefield India Data Centre Update H1 2025 - Bengaluru operational IT load | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IN-Kolkata) | MW IT load | Reported / Derived | 3.0/5 | 110 MW operational | Built up from facility-level disclosures (NTT Kolkata 1 = 40 MW; STT Kolkata DC2 = 34 MW) plus six further operating facilities. The Cushman & Wakefield city figure of 8 MW is irreconcilable with these two disclosures alone and is therefore NOT used - see the 62_QC_Reconciliation sheet | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IN-Other India) | MW IT load | Reported / Derived | 3.0/5 | 25 MW operational | Modelled residual for Ahmedabad/Gandhinagar, Kochi, Jaipur, Visakhapatnam, Coimbatore, Indore and Nashik - LOW CONFIDENCE | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (US-Northern Virginia) | MW IT load | Reported / Derived | 3.0/5 | 4000 MW operational | Northern Virginia (Loudoun + Prince William + Fairfax) commissioned data-centre IT load, ~4 GW class, per Dominion Energy connected-load reporting and market trackers | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (UK-London & Slough) | MW IT load | Reported / Derived | 3.0/5 | 1300 MW operational | Greater London / Slough Trading Estate operational IT load, ~1.3 GW class, per UK FLAP-D market trackers | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (IE-Dublin) | MW IT load | Reported / Derived | 3.0/5 | 950 MW operational | Dublin metropolitan operational data-centre IT load, ~0.95 GW class; EirGrid reports data centres at ~21% of national metered electricity demand | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (DE-Frankfurt Rhein-Main) | MW IT load | Reported / Derived | 3.0/5 | 950 MW operational | Frankfurt/Rhein-Main operational IT load, ~0.95 GW class, the largest FLAP-D market | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (BR-Sao Paulo State) | MW IT load | Reported / Derived | 3.0/5 | 600 MW operational | Sao Paulo state operational IT load, ~0.6 GW class (Ascenty, Scala, ODATA, Equinix, Microsoft) | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Market capacity anchor (SG-Singapore) | MW IT load | Reported / Derived | 3.0/5 | 1400 MW operational | Singapore national operational IT load ~1.4 GW; consistent with the source workbook's 2024 Singapore capacity value | The single largest source of uncertainty in every downstream figure. All facility capacities in this market scale linearly with it. |
| Facility load factors | fraction | Modelled | 2.0/5 | Hyperscale 0.80 | Wholesale 0.65 | Mid colo 0.60 | Small 0.50 | IT-load utilisation factors. Hyperscale 0.80 reflects operator-reported high average utilisation; colocation 0.60-0.65 reflects Indian and European colocation occupancy and diversity factors reported by JLL/CBRE (India vacancy ~4-8%, but contracted-vs-drawn load is lower). LBNL uses 0.50 for whole-fleet facility power; that value is retained for the smallest/enterprise class. | Single annual constants; no seasonality, no commissioning ramp. |
| Facility class weights | unitless | Modelled | 2.0/5 | Hyperscale 30 | Wholesale 12 | Mid colo 6 | Small 1.5 | Judgement-based ratios reflecting typical built IT load by facility type, calibrated so that market sums reconcile to the published anchors. | Judgement, not measurement. Changing these ratios redistributes capacity between operators within a market. |
| Facility identity, operator, address | text | Reported | 4.0/5 | Facility name, operator and street/estate address | Operator websites and press releases; datacentermap.com market directories for India, USA (Northern Virginia), UK, Ireland, Germany, Brazil and Singapore | Directory coverage is good for commercial colocation and hyperscale, weaker for enterprise-owned and government facilities. Ownership changes are not tracked after the compilation date. |
| Original national benchmark series 2020-2026 | various | Reported / Modelled / Proxy | 2.0/5 | All 44 country-year rows and both audit-trail sheets, preserved verbatim | The supplied workbook 'Data_Centre_Global_Dataset_2020_2026.xlsx' and its own Sources_Methods sheets | Carried through unchanged, INCLUDING its own stated limitations. Its India carbon and renewable-share values are inconsistent with standard national factors - see 62_QC_Reconciliation. |
Field dictionary
Definitions and the limitation to read first
42 fields, each with its formula, primary source and the specific way it can mislead.
| Field | Unit | Definition / formula | Limitation — read this |
|---|---|---|---|
| Data_Centre_ID | id | ISO3 country code + sequential number. Stable primary key across every sheet. | Not an external identifier; do not join to third-party datasets on it. |
| Data_Centre_Name | text | Facility name as published by the operator or a market directory. | Naming conventions differ between operators; some campus and building rows share a stem. |
| Operator | text | Owning or operating company at the time of compilation. | Ownership changes frequently (M&A, JV, sale-leaseback). Snapshot as at 2026-08-25. |
| Record_Type | category | Facility = a single data centre building. Campus = a multi-building parent, EXCLUDED from all sums. | Campus rows exist only to preserve the parent-child structure. |
| Country / Region / State_Province / City | text | Administrative geography of the facility's area. | City is the administrative city of the AREA, which for metro-fringe sites may differ from the marketed city. |
| Latitude / Longitude | decimal degrees | Centroid of the facility's industrial estate, business park or suburb. | AREA CENTROID, NOT a rooftop coordinate. +/-1-3 km typically, +/-5-15 km for city-centroid records. |
| Coordinate_Precision | category | Declares which of the two precision bands the coordinate belongs to. | — |
| Status | category | Operational, Under construction, Announced, Planned or Unknown. | Directories do not always disaggregate status; some Operational flags are inferred from listing context. |
| Year_Online | year | Reported commissioning year where found; otherwise back-cast (Methodology section 6). | Confidence 1 for every modelled year. Do not use for individual-site vintage claims. |
| Capacity_MW_IT_2026 | MW | Energised IT-load capacity. Reported where published; otherwise market-anchor allocation (Methodology section 3). | An allocation for 1094 of 1098 facilities. Individual values are indicative; market and cluster sums are the reliable level. |
| Capacity_MW_Ultimate | MW | Capacity at full build-out of the identified pipeline, before the status energisation factor. | Assumes announced and planned schemes complete at the same class weight as built stock. |
| PUE | ratio | Operator fleet/design PUE where published; otherwise the country-climate benchmark. | A fleet average applied to one site. Real site PUE varies with load factor, age and season. |
| Load_Factor | fraction | Average IT-load utilisation by facility class (Hyperscale 0.80, Wholesale 0.65, Mid colo 0.60, Small 0.50). | Single annual constant; ignores ramp-up on newly commissioned halls. |
| Energy_Consumption_MWh_yr | MWh/yr | Capacity x 8,760 x Load_Factor x PUE. | No site is metered here. Error compounds capacity error, load-factor error and PUE error. |
| Energy_Intensity_MWh_per_MW_yr | MWh/MW/yr | 8,760 x Load_Factor x PUE. | Varies only by class and country in this build; it is not a site-specific efficiency measure. |
| Grid_Electricity_Dependence_pct | % | Set to 100 for every facility. | No public evidence of material on-site generation was located at any facility. Treat as an upper bound. |
| Renewable_Grid_pct | % | National grid-average renewable share of GENERATION. | A physical grid mix, NOT contracted or matched procurement. Do not confuse with the next field. |
| Renewable_Matched_pct | % | Operator-published renewable matching / carbon-free-energy percentage, where one exists. | Annual matching, not hourly delivery, except where the operator states otherwise (Google reports hourly CFE). |
| Water_Consumption_m3_yr | m3/yr | Energy_MWh x WUE. | On-site cooling water only. Excludes upstream water embodied in electricity generation, which is often larger. |
| WUE | m3/MWh (= L/kWh) | Operator fleet WUE where published; otherwise country-climate benchmark. | UK, EU and Irish values are climate proxies with no primary national data-centre figure. Singapore's is a regulatory CEILING. |
| Water_Stress_Score / Label | 0-5 / category | WRI Aqueduct-informed baseline water-stress band for the facility's basin. | NOT a real Aqueduct basin lookup against the facility coordinate. Assigned per reference AREA. |
| Water_Stress_Weighted_Demand_m3 | m3/yr | Water_m3 x (water-stress score / 5). A scarcity-adjusted demand. | Linear scarcity weighting is a convention, not a hydrological model. |
| Cooling_Water_Dependence | category | Banded from WUE: >1.5 evaporative-dominant, 0.5-1.5 hybrid, <0.5 air-cooled/closed-loop. | Inferred from a water ratio, not from the installed plant. |
| Carbon_Emissions_Location_tCO2e_yr | tCO2e/yr | Energy_MWh x national grid emission factor. | LOCATION-BASED Scope 2. National averages hide large sub-national variation (~6x across US eGRID subregions). |
| Carbon_Emissions_Market_tCO2e_yr | tCO2e/yr | Energy_MWh x grid factor x (1 - operator matched-renewable %). | Only populated where the operator publishes a matching percentage. Annual matching is not hourly decarbonisation. |
| Emission_Intensity_tCO2e_per_MWh | tCO2e/MWh | The national grid emission factor applied to that facility. | Uniform within a country by construction. |
| Waste_Heat_MW_thermal / _TJ_yr | MWth / TJ/yr | Energy_MWh / 8,760, and the same energy expressed as TJ per year. | An average annual heat-rejection RATE, not a measured local temperature effect. |
| Cooling_Technology | category | Inferred from WUE band. | A guess at plant type from a water ratio. Replace with planning-application data where available. |
| Mean_Annual_Temperature_C | deg C | Long-run mean annual air temperature for the facility's area. | Annual mean; ignores the summer peak that actually drives cooling and heat-island effects. |
| Heat_Island_Proxy | MWth x k-persons/km2 | Waste_Heat_MW x resident density / 1,000. | A composite index with no physical units meaning. Comparative only. |
| Noise_dB | dB(A) @ ~50 m | 50 + 10 x log10(Capacity_MW), clipped to [45,85]. | One of the two weakest fields here. No facility in this workbook has a measured noise value. |
| Residential_Distance_m | m | Land-use class band. | A class band, not a measured setback. Real setbacks vary by an order of magnitude within a class. |
| Land_Use_Context / Sensitivity | category / 0-1 | Land-use class of the facility's area and its community-sensitivity weight. | Assigned per area, so every facility in an estate shares a value. |
| Population_1km / Population_5km | persons | Density band x pi x r^2, with a 0.60 radial decay factor applied at 5 km. | NOT a raster zonal statistic. Replace with GHSL or WorldPop against real coordinates before publication. |
| Community_Complaints | count/yr | 0 for every facility. | 0 MEANS 'no publicly documented facility-specific complaint was found'. It does NOT mean none exist. |
| Regulatory_Planning_Constraint | 0-1 | Cluster-level score for restrictive instruments in force (moratoria, connection bans, zoning frameworks). | Cluster-level, so it cannot distinguish two facilities in the same market. |
| Documented_Objection_Record | 0-1 | Cluster-level score for the strength of the public objection / litigation record. | Strong for the USA and Ireland; near-absent for India, Brazil and Singapore - an evidence gap, not an absence of impact. |
| Local_Infrastructure_Pressure_pct | % | Facility capacity as a share of its cluster's total capacity. | A concentration measure, not a grid-headroom measure. |
| Employment_Estimate_FTE | jobs | Capacity_MW x 5 permanent roles per MW. | Highly uncertain and varies enormously between colocation and hyperscale operating models. |
| P1..P6 pillar scores | 0-100 | Weighted sum of normalised sub-indicators (Methodology section 5). | Inherits the confidence of its weakest input; P5 and P6 are Confidence 1-2 throughout. |
| Composite_Score / Risk_Class | 0-100 / category | Weighted mean of the six pillars; class = quartile of this dataset. | Relative to this dataset only. Not an absolute risk statement about any facility. |
| Record_Confidence | 0-5 | Mean of the field confidences for capacity, PUE, WUE, coordinates, status, year and context. | A summary; always check the per-field confidence columns before relying on a single value. |
Where the workbook's bottom-up total diverges from a published market figure, the divergence is recorded in the quality-control table rather than reconciled away. Reading those rows is the fastest way to understand how much weight the totals can bear.