Methodology

This page describes the methodology used for data analysis and calculations in this dashboard.

Data Sources

• Our primary source for the reported data is from the Dutch authorities, namely the RVO.
• For calculating grid-based emissions, we use the annual average for the grid in the Netherlands provided by the European Environmental Agency (EEA, 2023)
• For calculating the manufacturing emissions of servers, we use the data provided by Datavizta from Boavizta.
• For emissions related to the burning of diesel fuel we use the data provided by the Environmental Protection Agency (EPA, United States).
• DataCenterMap.com to verify how many data centers are missing from the published RVO data.

Data Collection

RVO has released the raw data as Excel files for every data center operator. Each Excel file may contain multiple buildings. We have consolidated this data and imported it into a SQL database for data analysis in this dashboard.

In some cases, the Excel file contains only consolidated numbers for multiple buildings. In those cases, we have distributed the consolidated values evenly across the mentioned buildings.

For the estimations that we perform on data centers in the RVO set that reported incomplete data, we verify our estimations against publicly available data for each building, such as the megawatt capacity of the building or the POE mentioned on the website of the operator.

For the missing data centers, we consulted datacentermap.com to identify which data centers may be missing from the dataset, as well as other public sources and media information. We verified the collected data for each missing data center using the publicly available information on the operators' websites.

Calculation Methods

• Backup Generators: The reported data contains the electricity produced by backup generators. For our calculations, we assume all the backup generators are diesel generators and use diesel fuel type 2. To produce a kilowatt‑hour of electricity using DF2, we know that it requires approximately 10.7 liters of fuel. This allows us to convert electricity produced to fuel usage. We then multiply the fuel usage by the emission factor provided by the EPA (2.69kg CO₂-eq per liter) to get the total emissions from backup generators.
• Complete Data: We consider the reported data sets for each building as complete when they contain at least the total floor space and the total energy consumption.
• Owner Country: For each reported data center building, we research the ultimate beneficiary owner of the operating entity to be able to include the country of origin for the operator of the building.
• Performance Indicators: To calculate data center performance metrics such as DCiE (The Green Grid), ERF (EN 50600), PUE (EN 50600), and WUE (EN 50600), we used the corresponding standards to calculate them.
• Grid Emissions: To calculate the emissions from energy consumption of the data center buildings according to the self‑reported total energy consumption, we used the emission factor provided by the European Environmental Agency (256 gCO₂/kWh for Netherlands 2023) and multiplied the total energy consumption by the provided average annual carbon intensity of the power grid in the Netherlands.
• Emissions from Server Manufacturing: For calculating the carbon emissions occuring during the manufacturing of the server, we take into account the lifetime carbon emissions (5 years). To do so, we assume all servers are the same, and define the standard server as: 2 CPUs with 32 Cores each, 12 x 32 GB of memory (384 GB), 4 x 1 TB SSD disks and 2 PSUs. Using the standard server, we calculate the manufacturing emissions using the Datavizta tool. You can change our assumption (the GWP value of manufacturing) under 'Configure assumptions'.

Assumptions

Throughout the dashboard, several configurable assumptions are used for estimating missing data. These can be adjusted via the "Configure Assumptions" interface at the top of the main dashboard. The default values are:

• Power Density by Era: We assume different power densities per rack based on when a data center was built:
  • Before 2000: 1 kW/rack
  • 2000-2020: 4 kW/rack
  • After 2020: 8 kW/rack
  • For Hyperscale, we customize the estimated power density to 4-5 kW/Rack
• Square Meters per Rack: 4 sqm per rack (used for converting floor space to rack count)
• Utilization Rate: 50% (used for estimating actual energy consumption from capacity)
• IT Space Ratio: 70% (percentage of total floor space that is usable IT space)
• Default PUE: 1.6 (used when a data center's design PUE is not reported)
• Server Power Consumption: 1 kW per server
• Server Manufacturing GWP: 1365.0 kg CO₂-eq per server (Global Warming Potential over 5-year lifetime)

Estimation Methods

When you toggle "With non-reported data" (internally: assumptions=1), the dashboard applies estimation algorithms to fill in missing values for incomplete data center records. This provides a more complete picture of the Netherlands' data center landscape, but the numbers should be understood as estimates based on our methodology.

What happens when assumptions=1 is enabled:

1. Capacity Estimation (Floor Space and Power)

For data centers with partial capacity information, we estimate missing values using the following cascade logic:

• IT Floor Space: If total floor space is known but IT floor space is missing, we estimate it using the IT Space Ratio (default 70%): IT Floor Space = Total Floor Space × 0.70
• Total Floor Space: If IT floor space is known but total is missing: Total Floor Space = IT Floor Space / 0.70
• IT Power Capacity: Estimated using multiple approaches:
  • If total power capacity is known: IT Power = Total Power × Utilization Rate (50%)
  • If IT floor space is known: IT Power = (IT Floor Space / 4 sqm per rack) × Power Density (based on year built of the data center building)
• Total Power Capacity: Estimated from IT power capacity:
  • If IT power is known: Total Power = IT Power / Utilization Rate (50%)
  • If floor space is known: Total Power = IT Power Estimate × PUE

2. Energy Consumption Estimation

For annual reports missing energy consumption data, we estimate using capacity and utilization:

• IT Energy Consumption: If IT power capacity is known: IT Energy (kWh/year) = IT Power Capacity (kW) × Utilization Rate (50%) × 8,760 hours
• Total Energy Consumption: Derived from IT energy using PUE:
  • If data center has design PUE: Total Energy = IT Energy × Design PUE
  • Otherwise: Total Energy = IT Energy × Default PUE (1.6)
• Reverse Calculation: If total energy is reported but IT energy is missing, we work backwards: IT Energy = Total Energy / PUE

3. Server Count Estimation

When server counts are missing or reported as 0 or -1:

• If actual IT energy consumption is known (more accurate): Servers = IT Energy (kWh/year) / 8,760 hours / Server Power (1 kW)
• Otherwise, estimate from IT power capacity: Servers = IT Power Capacity (kW) / Server Power (1 kW)

4. Missing Data Centers

In addition to estimating missing values for reported data centers, when assumptions are enabled, the dashboard also includes data centers that we identified as missing from the RVO reports entirely. For these facilities:

• We gather publicly available information (floor space, power capacity from operator websites)
• We create synthetic annual reports with empty values for unreported metrics for when we cannot find public information
• We apply the same estimation cascade as above to fill in capacity and energy values

5. What Gets Included vs Excluded

The toggle affects different metrics differently:

• Always excluded (even with assumptions enabled): Data centers with absolutely no usable data points
• Included when assumptions are turned on:
  • Data centers with at least one capacity or floor space value (for estimation)
  • Data centers marked as missing from RVO reports
  • Annual reports with partial data (we fill in the gaps)
• Estimation flags: Our system internally tracks which values were estimated vs. reported.

6. Accuracy Considerations

Important notes about the reliability of estimated values:

• Estimates based on reported total energy consumption (working backwards to IT energy) are generally more accurate than estimates from capacity
• Power density assumptions vary significantly: newer facilities (post-2020) are assumed to have 5× the density of pre-2000 facilities
• Utilization rate of 50% is a conservative industry average, but actual utilization varies widely by operator and facility type and is likely lower.
• The default PUE of 1.6 reflects typical data center efficiency, but modern cloud, hyperscale and co-location facilities often achieve much lower values (1.1-1.2). Our value is derived from the EU Code of Conduct for Data Centers.
• For missing data centers, we rely heavily on publicly stated capacities, which may reflect design capacity rather than operational capacity