CSO approaches to IT emissions: help us forge the future of sustainable AI

Article written in collaboration with Joseph Walton, Research Fellow in Arts, Climate & Tech at the University of Sussex

Digitisation is an essential aspect of sustainability management – but the increased use of software and cloud solutions to assess and reduce carbon footprints is contributing to a rapid increase in IT emissions. So how are Chief Sustainability Officers approaching the climate impacts of IT, the cloud and AI? The University of Sussex and Greenpixie have teamed up to find out, in partnership with CSO Futures, in a new Innovate UK-funded project. 

For most Chief Sustainability Officers, digital emissions remain an obscure part of carbon accounting – and this could quickly become problematic as they begin to integrate more and more power-hungry AI applications to support their work.

It’s a challenge for any company to measure their own cloud carbon emissions, never mind engaging the suppliers of their AI tools around their cloud carbon. So how are sustainability professionals interacting with providers, when it comes to AI and the cloud? 

Answering this questionnaire takes less than five minutes, and it will help to inform the next generation of tools and standards to support sustainable AI. 

Designing the future of cloud carbon emissions data

The research project is called Designing the Future of Cloud Carbon Emissions Data. It’s a bold title — research projects need to be, these days — so here’s a little more context. It might just persuade you that boldness is just what is needed.

One often quoted study estimates that IT contributes 2.1%–3.9% of our global carbon footprint. That may not sound like much. But it’s already more than all global aviation put together. And the carbon footprint of IT is growing fast. In fact, since the study was published in 2021, we have witnessed an unprecedented expansion in AI deployment. Tools like ChatGPT and Gemini are already rewiring the way we do business. It looks like AI video is next — and generating video is a lot more carbon intensive than generating text. 

But let’s not forget, AI can also unlock a lot of sustainability benefits. AI is part of how the planet will reach net zero. It is already creating efficiencies across energy, transport, agriculture, and other sectors. But as things stand now, nobody really knows how to weigh up the environmental benefits of AIs against their environmental costs. 

One reason is that some promised gains from AI may turn out to be wishful thinking. Another reason is that the big cloud providers, whose data centres are used to train and deploy AI tools, aren’t yet being transparent enough about how they run those data centres. 

Granular data often isn’t available to calculate carbon emissions accurately. That makes managing cloud emissions via GreenOps more difficult than it needs to be. Even when the data is available, calculations can get very complex very quickly. It is challenging to model the interactions between embodied carbon, network, storage, and compute costs, in data centres fed by a variable mix of energy across space and time. 

The dark side of cloud computing

Tools for carbon-aware computing provided by the three cloud giants, Amazon Web Services, Google Cloud Platform, and Microsoft Azure, are certainly steps in the right direction. But they need to improve rapidly. Right now it is third-party tools from specialist cloud emissions data providers, as well as open source tools such as Cloud Carbon Footprint, that are leading the way. There are also ongoing debates around the effectiveness of existing carbon-aware computing frameworks.

51% percent of companies say that sustainability is moving them to a public cloud first approach. Yet this is also consistent with a growing recognition that the public cloud, by itself, won’t be enough to green global ICT. The recent documentary Clouded II, from Dark Matter, reveals thought-provoking examples of cloud repatriation. Cloud repatriation is when businesses move some of their computational workloads out of the public cloud, and back onto their on-premise infrastructure. Cost is a big driver, but transparency and sustainability are part of the picture too.

Better data needed from IT supply chains

Then there is the question of supply chain. The dream would be to be able to quickly and reliably rank the sustainability of different AI tools, along with other popular IT products and services. While there are some relevant certifications and standards out there, we are very far off from a comprehensive and reliable framework for sustainable IT procurement.

We need better data around these challenges, which is where your input to the Sustainable Cloud Survey comes in. Addressing these challenges will also mean coordinating across sustainability, procurement, IT, and innovation functions. The future has a way of approaching from many angles at once. 

Businesses don’t want to fall behind on AI. But nor do they want to fall behind on sustainability. How do these visions of the future relate? That’s the conversation we need to be having.