Cloud Sustainability 2024

IT leaders seeking information on the sustainability of cloud services are reliant on the assurances of vendors about the sustainability of their services. Those with queries and questions will be directed towards ESG reports or specific websites. These reports and pages are great at providing background and communicating targets and assurances but they're typically short on one thing that really matters - data.

All cloud vendors will cherry pick data to varying degrees to show themselves in the greenest possible light. This is why Computing now scrutinises sustainability claims closely.

In 2022, Computing conducted research for the first time into the sustainability credentials of the three hyperscale cloud giants and some other smaller cloud service providers.

A series of sustainability metrics were created and grouped into four categories: Standards and Policies, Emissions/Energy/Water, Waste/Circular Economy/Recycling and Transparency.

The research is updated annually, and the matrix is updated to reflect learnings from the previous year.

Matrices for Cloud Sustainability 2022 and Cloud Sustainability 2023 are available here. A full explanation of each criterion and explanation of weighting can be found further down the page beneath the matrix.

2024 Sustainability Matrix

Explanation of Sustainability Matrix

Standards & Policies

Vendors should provide clarity about which sustainability standards have been signed up to. Computing chose what we consider to be robust targets and assessed the degree of compliance or alignment. There are value judgements involved where Computing weighs up factors such as the ambition of climate targets based on the size and power of the organisation. These criteria are strong indicators of the sincerity and commitment to reducing environmental impact.

• UN Sustainable Development Goals: There are 17 UN Sustainable Development Goals, a full list of which is available from https://sdgs.un.org/goals. The goals cover a huge area, including the elimination of poverty and hunger, the reduction of gender inequalities and sustainable cities and communities as well as affordable and clean energy, climate action and life on land. Cloud vendors are assessed on the degree of alignment, the quality of reporting and extent of progress. 1 point
• Policy & Advocacy: This criterion measures the extent of engagement with governments, intergovernmental organisations and NGOs. It helps to assess the degree to which Big Tech is using its power and voice for the benefit of everyone. 1 point
• Science Based Targets Initiative: The Science Based Targets Initiative (SBTi) provides a clearly defined pathway for organisations to reduce their GHG emissions. Targets are considered to be Science Based if they are in line with the goals of the Paris Agreement, limiting global warming to below 2°C and preferably 1.5°C. The names of companies who have made commitments, along with detail on whether these are near term, long term and/or Net Zero commitments are available via https://sciencebasedtargets.org . Vendors are judged on whether they had signed up to a target, which of the five stages of target they are at (Commit, Develop, Submit, Communicate, Disclose) and the detail of that target. 2 points
• Customer ecosystem engagement inc. Scope 3 reporting tools: This criterion is a wide-ranging category based partly on the extent to which cloud vendors engage with their customer ecosystem and the extent to which tools are made available to customers to enable them to assess and reduce the carbon footprints generated by their cloud infrastructure. 2 points
• Upstream supply chain sustainability: This criterion is solely focused on upstream emissions, and the extent to which vendors attempt to quantify and reduce these emissions. 3 points
• Ambition of emissions reduction target: This is a value judgement of the commitment of vendors to reducing their environmental impact. It is informed by the preceding criteria as well as a judgement of the ambition of sustainability targets in comparison to the other vendors within the scope of this analysis. 1 point

Energy/Emissions/Water

This quarter of the matrix is where the metrics provided by cloud giants are compared against their stated ambitions. It assesses the sustainability of datacentres based on metrics such GHG emissions, energy efficiency (including how much energy used is generated from renewable) and water efficiency. The data used to inform these assessments tells us how sustainable an organisation truly is. Large gaps between ambitions and data indicate the application of greenwashing.

• GHG emission reduction progress against Net Zero Target: This is a criterion with a wide scope. The GHG emissions of vendors are analysed not just in absolute terms, but in terms of their reduction over the last few years and in terms of progression to Net Zero emissions. Some vendors already claim to be Net Zero operations, so it is vital to understand whether this is due to a genuine reduction in the amount of GHGs being pumped out, or creative carbon accounting. Maximum points can only be achieved by a demonstrable commitment to quantify and remove historical carbon emissions. 5 points
• Carbon Usage Effectiveness (CUE): This metric determines how well a datacentre uses carbon resources. It is determined by dividing scope 1 & 2 emissions by energy output in kWh or mWh. 2 points
• Renewable energy generation: This criterion examines what proportion of energy consumed by datacentres come from renewable sources. Most cloud vendors can only generate a fraction of the energy they consume from their own onsite renewable sources, so they offset their Scope 2 emissions (indirect emissions caused by the purchase of energy, steam, heat or cooling) with the purchase of renewable energy from other sources such as Power Purchase Agreements (PPAs) or Renewable Energy Certificates (RECs). The more additional renewable energy s generated for datacentres, the higher score will be.3 points
• Built emissions: The majority of the overall carbon footprint of a datacentre is incurred at the beginning of the building lifecycle. Embodied carbon includes emissions created by extraction, manufacturing and transport of materials, and those created during the actual process of building. Quantifying these emissions is tricky but this criterion assesses progress. 2 points
• Cooling methods: Cooling datacentres is potentially a huge source of energy - and water consumption. Datacentres in temperate climes can utilise fresh air cooling if they are equipped accordingly, but many do not. Older style refrigeration cooling is energy intensive. Evaporative cooling is less so but can be water intensive. Closed systems and the use of rainwater, condensate or grey water can ameliorate this. Points are given for quality and quantity of information. 3 points
• Datacentre innovation for efficiency: All hyperscale providers are innovating to create more efficient and more sustainable datacentres, because it serves them as well as everyone else to be able to create more capacity within the same space. This is a value judgement based on information provided about innovations in chip design, datacentre design and construction, cooling etc. 2 points
• Water withdrawal: Many datacentres are built in areas of water scarcity. This metric considers the overall withdrawal of water and also consumption which takes into account water discharge. Points are awarded based on reductions in both. 2 points
• WUE (Water Usage Effectiveness): This metric is obtained by dividing the water consumption in litres by the IT Equipment Energy (ITEE) in kWh. The lower a WUE ratio is, the more efficiently that organisations datacentres are using water resources.1 point

Waste/Recycling/Circular Economy

All cloud vendors hold a different position on the spectrum between basic recycling and embedding the principles of circular economy into core business processes, and infrastructure. This category of assessment is all about the extent to which cloud vendors are squeezing various categories of waste out of their value chain, recycling and refurbishing server components and networking equipment and closing the recycling loop.

• Use of refurbished parts in datacentres: What proportion of server and networking equipment is refurbished? There is ample evidence to indicate that refurbished hardware can perform as well as new in certain scenarios. Points are awarded for data provision and the fact of whether this proportion has increased. 2 points
• Total waste generation: Points are awarded for the provision of clear data and a reduction in the overall quantity of waste generated from year to year. 2 points
• E-waste generation: Electronic waste is anything consisting of electronic components, plugs and cables. E-waste matters because it can contain toxic compounds such as mercury and lead which is problematic to safely dispose of. However, many of the non-ferrous (not containing iron) metals such as copper, gold and aluminium can be re used. Points are given for separating e-waste as a reporting category and the proportion recycled. 2 points
• Landfill Diversion Rate: This is a useful environmental impact metric which simply measures how much waste is kept out of landfills, often via recycling and reuse schemes that already exist. Landfill diversion rate is arrived at by dividing the overall waste figure by the proportion reused/recycled/otherwise diverted from landfill and multiplying by one hundred. 2 points
• Evidence of circular economy: This is a value judgement based partly on the data available in other criteria in this category but also on information provided in ESG reporting. 2 points

Transparency

This category is concerned with transparency of data. Whilst Net Zero ambitions (and increasingly water positivity) are cornerstones of sustainability profiles, data quality varies. Carbon data is often presented selectively, sometimes with misleading visual prompts. Carbon accounting is a dark art because GHG reporting protocols are sufficiently ambiguous as to leave room for interpretation. Carbon offset projects vary enormously in their effectiveness. The final criterion in this category is a value judgment based on the ease (or otherwise) with which journalists can find the hard data required.

• Quality of scope 3 reporting: Scope 3 reporting is very challenging, because its very nature is indirect. These are emission which occur outside of the direct control of the reporting organisation so they can be very difficult to quantify. Points are awarded based on how far a vendor has gone to quantify scope 3 emissions – even when more accurate reporting leads to having to report increased emissions. Scopee 3 categories should be clearly laid out. 2 points
• Clear, target based GHG emissions reporting: This concerns how emissions data is presented. GHG emissions should be presented in full as well as net of offsets and compared to progress against targets set. Data should be easy to view in comparison to previous years to facilitate a clear understanding of progress over a longer period. If baselines have been adjusted for any reason, this should be explained.1 point
• Carbon offset quality & transparency: This criterion concerns itself primarily with projects used to offset indirect emissions. ESG reports often contain considerable content about offsetting projects, but sometimes fail to set out how exactly how they meet the critical metrics of additionality, leakage prevention, permanence and verifiability.
  
  Additionality means that offsetting project should actively remove CO² from the atmosphere and that these carbon savings would not occur regardless of the project. Permanence refers to the requirement that savings in GHGs brought about by the project cannot be reversed. Reforestation projects won't work if the forest gets burned to the ground or cut down at a later date. Equally, leakage occurs when, for example, reforestation in one part of a country leads to deforestation in another.
• Verifiability means that carbon offset projects, and the carbon credits that they link to should be verified by a third party such as Verified Carbon Standard (VCS): 2 points
• Accessibility and transparency of data overall: This is partly a value judgement of researchers, but also partly informed by the preceding criteria. It is a judgement of the quality of data, how easy (or otherwise) it is to access and how transparent the vendor is overall. Points are removed for making less favourable data harder to access, for omission of context and for misleading use of graphs. 3 points
• Use of third-party datacentres: Included to reflect the fact that if cloud hyperscalers fail over to third party datacentres, multiple other metrics are affected. It's an open industry secret that hyperscalers use third parties so transparency is critical if customers are to understand the impact on their emissions. However, confidentiality and security concerns make this very difficult. Points are awarded for acknowledging use of third-party datacentres and trying to quantify the impact. 2 points