Reducing impact - our what-if scenarios

In this section, we look beyond estimations/reporting and think about what factors affect the emissions estimates of our digital estate. We’re calling this our what-if scenarios.

What would happen if we switched cloud provider? What would happen if we only used refurbed hardware from now on. Here’s what we explored in these sections:

• Employee hardware
• Cloud services

Employee Hardware

What-if question

What if GWF purchased new laptops rather than refurbished ones?

Exploration

Of the four laptops acquired in 2023, two were purchased new and two refurbished. Those refurbished laptops use a condition factor of 15% attribution, the new laptops use 100% attribution.

Returning to the formula we used in the upstream emissions section, we can recalculate the total embodied carbon emissions assuming all four laptops were acquired as new:

With two refurbished and two new laptops, we previously calculated a total of 186 kgCO₂e. If they were all purchased new, that total becomes 437 kgCO₂e. That’s a difference of 251 kgCO₂e.

Purchasing refurbished hardware clearly brings significant carbon savings.

Cloud Services

What-if question

What if GWF uses a cloud service provider that runs on Germany’s or Finland’s regional grid rather than renewable energy sources?

Exploration

Let’s revisit the detailed estimation for Hetzner VMs operational emissions. The resulting detailed operational carbon emissions estimate for the four Hetzner VMs was 0.94 kgCO2e/year.

To estimate those emissions, we used a renewable energy carbon intensity value of 0.0195 kgCO₂e/kWh which represents a mix of hydro and wind power. To model our what-if scenario we can replace this carbon intensity for a region’s carbon intensity.

Germany

For this we’ll assume a replacement cloud service provider with a data centre located in Germany using the local energy grid. We’ve chosen Germany as this is where the majority of the GWF team are based, and it’s common that organisations use suppliers in their own country. Germany’s grid carbon intensity is 0.364 kgCO₂e/kWh, so:

Finland

We also want to include Finland in these scenarios as Hetzner’s data centre is located there. Whilst Hetzner demonstrate they use renewables for their data centre, it is open to debate as to whether the country’s regional grid data is more helpful when considering impact. In simple terms, the reason for the debate is that rarely do data centres operate in isolation from the wider grid. It depends on where you draw your boundaries of concern, and that can justifiably change depending on whether you are concerned with localised or more systemic impacts.

Finland’s grid carbon intensity is 0.079 kgCO₂e/kWh, so:

Outcome

The figures we got from modelling a data centre powered completely by wind/hydro was less than 1 kgCO₂e/year. On the Finnish grid we see just under 4 kgCO₂e/year and on the German grid we see over 17 kgCO₂e/year.

Let’s bear in mind these numbers in isolation are very small, because GWF doesn’t use massive amounts of server power. But the differences between them in percentage terms is very large and that’ll result in much more significant numbers for an organisation using more server power.

It’s a massive saving in carbon emissions to simply consider the energy used by a cloud provider - either directly or through the wider grid they’re situated in. When buying decisions like this are made at scale, the impacts from emission reduction are likely to be significant.

Next - Appendix