Editor’s note
Welcome to another edition of the Jisc Digital Sustainability Newsletter. If there’s a thread running through this month’s stories, it’s the growing tension between the pace of datacentre infrastructure expansion and our collective ability to understand, let alone govern, its environmental impact.
Planning authorities in both Edinburgh and Buckinghamshire have pushed back on data centre approvals, City Hall is developing a new policy for London, and MPs have launched a fresh inquiry into the tech sector’s environmental impacts. At the same time, research suggests most UK firms still can’t accurately measure the carbon their AI workloads produce, which makes meaningful progress difficult to demonstrate in either direction.
However, there are reasons for at least a little cautious optimism too. Wind and solar overtook fossil fuels in EU electricity generation for the first time last year, and a Norwegian data centre piping waste heat to a neighbouring trout farm is a small but pleasing example of what a digital sustainability circular economy can look like in practice.
As ever, I hope you enjoy this edition of the newsletter and don’t forget to subscribe for future editions by signing up to the DIGITAL-SUSTAINABILITY Jiscmail list. Please do get in touch if you would like to contribute any stories, blog posts, journal articles, or other resources that our readers may find interesting.
— Cal Innes, Digital Sustainability Specialist, Jisc (cal.innes@jisc.ac.uk)
Digital sustainability news
Here’s a quick roundup of this month’s biggest digital sustainability news headlines:
Government admits it wrongly approved £1bn data centre next to M25
The government has conceded in court that planning permission for a large data centre in Buckinghamshire was granted in error, after campaign groups successfully argued its electricity demands and environmental impacts had not been properly assessed.
Microsoft’s water use set to nearly double by 2030 despite earlier conservation pledges
Internal forecasts obtained by the New York Times show Microsoft expects its data centres to consume around 18 billion litres of water annually by 2030. This figure is more than double its 2020 usage as AI-driven construction outpaces the company’s own sustainability commitments.
Most UK firms cannot measure the emissions from their AI systems, research finds
A survey of UK IT decision-makers has found that more than half are unable to accurately measure the carbon emissions generated by their AI workloads, despite the vast majority believing AI is helping them meet their sustainability targets.
AI data centre boom is driving a record surge in gas power development, report warns
Global plans for new gas-fired power stations are set to grow existing capacity by nearly 50% this year, with the US leading the way as tech firms race to meet the electricity demands of AI; a trend researchers warn will lock in decades of carbon emissions.
Wind and solar overtook fossil fuels in EU electricity generation for the first time last year
Wind and solar power produced 30% of the EU’s electricity in 2025, edging ahead of fossil fuels for the first time on record, according to data from energy think-tank Ember, with renewables and nuclear together now accounting for 71% of the bloc’s power supply. A 19% jump in solar capacity drove the record renewable output; solar now supplies more than a fifth of electricity in Hungary, Spain and the Netherlands.
The European Commission has adopted the world’s first voluntary certification methodologies for permanent carbon removals, setting a global benchmark aimed at accelerating the deployment and scaling of carbon‑removal technologies.
Microsoft claims 100% renewable electricity target met, but critics question what that really means
Microsoft says it has matched all of its global electricity consumption with renewable sources. However, the claim relies on power purchase agreements that add clean energy to the grid overall without guaranteeing it physically powers any given facility. Under current rules its Scope 2 emissions are reported as near zero, though actual grid-based calculations would put the figure at around 25 million metric tonnes of CO2 equivalent between 2020 and 2025. The accounting standards underpinning that claim are currently under review by the Greenhouse Gas Protocol; a process Microsoft is itself actively participating in, lobbying for PPAs to remain a recognised tool.
City Hall confirms new London policy on data centres amid concerns over energy and water use
The Greater London Authority has confirmed it is developing a new policy on data centres, as officials acknowledge that London’s concentration of such facilities is creating significant environmental pressures and undermining the mayor’s climate goals. Every 10 power-hungry data centres is estimated to produce 2.7 million tonnes of carbon emissions, and housing projects in west London have already been stalled due to data centres consuming all available grid capacity. 60 of the 100 UK data centres of this scale currently in the pipeline are earmarked for London.
Edinburgh councillors reject plans for large AI data centre despite planners’ recommendation
City of Edinburgh councillors have voted down a proposed hyperscale AI data centre on the former Royal Bank of Scotland site in South Gyle, overruling their own planning officers amid concerns about emissions and what actually constitutes a “green” data centre, especially its use of diesel back-up generators. The decision sits awkwardly alongside the UK government’s push to designate data centres as critical national infrastructure and introduce fast-track national approval routes that bypass local planning authorities
Norwegian data centre pipes waste heat to nearby trout farm in circular economy trial
A data centre operator in Norway has begun supplying excess heat to a neighbouring fish farm, in a project its partners say reduces energy consumption for both businesses and could be scaled up significantly.
Academic papers
Now for a summary of some of the most important new digital sustainability research:
Imposing AI: Deceptive design patterns against sustainability
Authors: Anaëlle Beignon, Thomas Thibault, Nolwenn Maudet
A fascinating paper documenting how tech companies are using interface design to push users towards generative AI features whether they want them or not. The authors argue this constitutes a form of deceptive design with real environmental consequences. Analysing 90 screenshots across 53 applications including Google, Microsoft, Meta and Adobe, the researchers identify two main strategies: making AI features visually dominant at the expense of existing non-AI functions, and using “magic” and “assistant” metaphors that obscure both the limitations and environmental costs of the technology. The paper calls for regulation linking interface design choices directly to environmental accountability.
When digital technologies stumble: Exnovating for conservation science and practice
Authors: Danilo Urzedo, Sabrina Chakori
A thought-provoking perspective piece arguing that conservation science needs to get better not just at adopting new digital technologies, but at deliberately phasing out or replacing ones that aren’t working (a process the authors call “exnovation”). While AI and other digital tools are increasingly central to biodiversity conservation, the paper documents how poorly designed or contextually inappropriate technologies can entrench injustice, generate biased decisions, and cause real harm to the communities and ecosystems they’re meant to protect.
Understanding efficiency: Quantization, batching, and serving strategies in LLM energy use
Authors: Julien Delavande, Regis Pierrard, Sasha Luccioni
A technically detailed but practically important paper from Hugging Face researchers examining how the way you deploy an LLM, not just which model you choose, can make orders-of-magnitude differences to energy consumption. Testing across multiple models on NVIDIA H100 GPUs, the authors find that common assumptions about efficiency (particularly around quantisation) don’t always hold up in practice, and that simply managing how requests arrive at a server can cut per-request energy significantly.
Carbon topography representation: Improving impacts of data centre lifecycle
Authors: Olivier Weppe, David Bekri, Thibaut Marty, Loïc Guibert, Louise Aubet, Jean-Christophe Prévotet, Maxime Pelcat, Sébastien Rumley
A methodological paper proposing a new way to visualise the carbon footprint of servers; one that captures how emissions shift depending on how busy the server is and how clean its electricity supply is. By combining real-world power measurements across four different servers with lifecycle assessment data covering manufacturing emissions, the authors produce a kind of “carbon map” for each server. Their findings challenge some common assumptions about when it makes sense to replace old hardware, extend server lifetimes, or prioritise renewable energy.
Resource corner
Each month, we share one or two digital sustainability reports, tools, or resources that we think is worth your time. This month we’re highlighting two resources that take very different, but complementary, approaches to understanding the environmental impact of AI.
The AI climate hoax – Report by Ketan Joshi
Our first resource is a new report by climate and energy analyst Ketan Joshi, developed in collaboration with organisations including the Green Web Foundation, Stand.earth, Beyond Fossil Fuels, and Friends of the Earth.
The report interrogates a claim that has become increasingly common in tech narratives: that AI will help solve the climate crisis. By analysing public claims, datasets and case studies, the report argues that the evidence supporting these claims is often weak or overstated.
A key finding is that many of the climate benefits attributed to AI come from older, lightweight forms of machine learning, rather than the rapidly expanding generative AI systems driving the current boom in data centre construction. Meanwhile, the environmental impact of this infrastructure expansion, particularly energy use and associated emissions, is rising rapidly.
The report suggests that the “AI for climate” narrative can sometimes function as a distraction from growing emissions linked to digital infrastructure, especially when companies are simultaneously missing their climate targets.
You can read the full report and access the supporting datasets and charts here.
Framework for assessing the environmental impact of AI – ITU
Our second resource takes a more methodological approach. Developed jointly by the International Telecommunication Union and ETSI, this report provides one of the first standardised frameworks for assessing the environmental impact of AI systems.
The guidance is built on life cycle assessment methodology and encourages practitioners to evaluate the full life cycle of AI systems, from hardware manufacturing and model training to deployment, operation and end-of-life.
The framework distinguishes between two perspectives that are often conflated:
- AI for sustainability – using AI to reduce emissions or environmental impact in other sectors.
- Sustainability of AI – understanding and minimising the environmental impact of AI systems themselves.
The recommendation outlines how to measure impacts such as energy use, greenhouse gas emissions, water consumption and resource extraction across different AI technologies. For organisations experimenting with or deploying AI systems, the framework provides practical guidance on defining functional units, setting system boundaries, and allocating impacts when using shared infrastructure such as cloud data centres.
You can explore the report here.
Digital sustainability articles
Here is a selection of our favourite articles on digital sustainability from the last month. Click on the title link to be redirected to the full article:
Enabled emissions: How AI helps to supercharge oil and gas production
Global Witness lays out a case, via senior campaigner Hannah Sharpe and communications advisor Lars Sellien, that the conversation about AI’s environmental impact has been looking in the wrong direction, focusing on data centre energy use while largely ignoring the far larger emissions unlocked when AI is used to boost fossil fuel extraction. The piece introduces the concept of “enabled emissions” and argues that current reporting frameworks are structurally blind to them.
What is the energy cost of an AI query?
Anne Currie cuts through the noise around AI energy consumption to argue that asking how much power an individual query uses is not only unanswerable, it’s the wrong question entirely. The more useful question, she contends, is what commitments your AI provider has made to running on clean energy, and whether you’re making supplier choices accordingly.
New technologies are stepping up the global fight against wildlife trafficking
Eve Bohnett, writing for The Conversation, surveys a range of emerging digital tools being deployed against wildlife trafficking and finds that AI and data-driven approaches are beginning to shift enforcement from reactive to proactive, even as the scale of the problem remains enormous.
Why companies should be sustainable and how IT can help
Carolyn Heinze, in a feature for TechTarget, makes the case that sustainability and ESG strategy has moved from a nice-to-have to a core business function, and IT teams are uniquely positioned to drive it, given their control over data infrastructure, procurement, and digital transformation. The piece draws on recent Deloitte and KPMG research to map both the current state of corporate sustainability ambition and the practical levers available to technology leaders.
Why making AI sustainable by design is key to a greener future
David Costa, NTT DATA’s Chief Sustainability Officer, sets out his case in the World Economic Forum that sustainability needs to be built into AI systems from the ground up rather than bolted on afterwards, and that the industry currently lacks the measurement frameworks to even know how it’s doing.
The world needs China’s climate technology. That’s the dilemma
Justin Worland, writing for Time, traces the profound tension at the heart of global climate politics: China’s dominance in clean technology manufacturing is arguably the single biggest factor bending the world’s emissions curve downward, yet geopolitical rivalry and trade protectionism risk squandering that advantage precisely when it’s most needed.
GreenOps: Greener cloud usage multiplies with Kubernetes optimisation
CloudBolt COO Yasmin Rajabi, writing for Computer Weekly’s developer blog, makes the case that the biggest barrier to reducing cloud carbon footprints in Kubernetes environments isn’t the availability of tools, but instead misaligned incentives between the teams who write the code and the teams who pay the bill, and that GreenOps initiatives are beginning to address this through shared accountability rather than purely financial arguments.
Despite its steep environmental costs, AI might also help save the planet
Nir Kshetri, professor of management at the University of North Carolina, offers in The Conversation a sector-by-sector survey of cases where AI is already delivering measurable environmental benefits, arguing that the technology’s impact is more complex than either its critics or cheerleaders tend to acknowledge, and that real efficiency gains are happening now rather than being speculative.
AI and power: Where does green HPC go from here?
Kirk Cameron, co-founder of the GREEN500 list and faculty lead at Virginia Tech’s Institute for Advanced Computing argues that generative AI is recreating the energy efficiency crisis that HPC spent two decades solving, and that the community is at risk of repeating its early mistakes. Alex Woodie covers Kirk Cameron’s presentation for HPCwire.
Why Microsoft and Amazon are turning to nuclear power for AI
Robert Rapier, a chemical engineer and energy sector contributor at Forbes, examines the strategic logic behind Microsoft and Amazon’s moves into nuclear power — arguing that AI’s insatiable demand for continuous, reliable electricity is forcing tech companies to think and act like long-term energy infrastructure planners for the first time.
A social critique of AI amid the climate crisis
Paul Schütze, Visiting Senior Research Fellow at King’s College London’s Department of Digital Humanities, argues in this piece on the KCL blog that framing AI’s environmental impact as a technical problem to be fixed with greener infrastructure fundamentally misdiagnoses it, and that AI functions as an ideological system that actively reproduces the socio-economic conditions making serious climate action impossible.
Podcast pick
AI sustainability podcast episode 12 – a sustainability leader’s guide to using AI responsibly
This month’s pick is from the AI Sustainability Podcast, hosted by Nina Benoit. In this episode, Benoit speaks with Alison Taylor, clinical professor at NYU Stern School of Business, about what responsible AI adoption actually looks like for sustainability professionals. The conversation covers governance structures, stakeholder analysis, and the evolving skill sets sustainability leaders need as AI becomes harder to ignore.
You can listen to the episode here.
Get involved:
We want to hear from you! Share your comments, suggestions, and digital sustainability highlights. Contact our Subject Specialist for Digital Sustainability, Cal Innes, at cal.innes@jisc.ac.uk
And don’t forget to subscribe to our DIGITAL-SUSTAINABILITY JiscMail mailing list for future editions of the Digital Sustainability Newsletter.
