Known limitations of current DCFootprint model

DCFooprint only relies on publicly accessible and open data for its impact factors. Some of those factors may have limitations, listed below:

• Electricity consumption impact factors, coming from Ademe Base Empreinte, are from 2011, which is too old for an accurate evaluation considering the rapid evolution of the electricity grid mixes. Based the last decade evolution of electricity grid mixes, the current version of DCFootprint might, on average, overestimate the GHG emissions due to electricity consumption.
• Water consumption footprint per kWh of electricity consumed comes from the World Resource Institue from a 2020 dataset, which is also old now.

The first version of the DCFootprint model has several limits as well:

• The IT inventory is modeled based on the main parameters given to DCFootprint, this model, described here, is limited as it models a generic and simple IT inventory, not accounting for more modern infrastructures like genAI specialized servers and GPUs. Because of this, the current model might underestimate the manufacturing footprint of IT equipments.
• The manufacturing footprint of the building and the technical environment are computed based on generic impact factors coming from Ademe’s Cloud & Datacenter PCR. Those factors are handy for surface and power based calculations, but limited as they have been computed on other studies made in France. They present average values that are certainly accurate enough for Datacenters installed in France, but not elsewhere. This is a limitation to fix for future versions of DCFootprint.