If you haven’t already, be sure to subscribe to our DIGITAL-SUSTAINABILITY JiscMail mailing list to receive future editions of the Jisc Digital Sustainability Newsletter directly to your inbox.
Editor’s Note
Welcome to the first Digital Sustainability Newsletter of 2026.
What strikes me about this month’s headlines is that there seems to be growing pressure on the tech industry from communities opposing data centre expansion. One environmental group is now suing a US regulator over energy demand secrecy, while Microsoft has responded to years of local opposition with a genuine shift in approach, pledging to cover infrastructure costs rather than pass them on. Whether that becomes industry standard or remains an outlier is one to watch.
This month has also seen the publication of a strong crop of academic papers examining everything from the material footprint of GPUs to the embodied carbon in storage media.
Meanwhile, research keeps sharpening our understanding of AI’s true costs. The latest AI Energy Score findings from Hugging Face I found particularly eye-opening. Reasoning models (the ones that take a little longer to “think” through problems before answering) can use up to 700 times more energy than standard models for the same task. That’s not a typo. It’s driven largely by the sheer volume of tokens these models generate during their internal monologue. A useful reminder that the most powerful option isn’t always the right one, and that being deliberate about model choice is becoming a sustainability decision in itself.
Plenty more to dig into below. As always, I’d love to hear what you think.
— 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:
Microsoft has announced a “Community-First AI Infrastructure” initiative, promising to pay for electricity grid upgrades and water infrastructure needed for its data centres, after more than $64 billion worth of data centre projects were cancelled over the past two years due to community objections about utility costs and resource scarcity. Microsoft’s five commitments include: bearing the cost of electric grid transmission updates and new generation resources; replenishing water in host communities; training and hiring local residents; and refraining from requesting special tax breaks.
Environmental group sues US regulator over Meta data centre energy demand secrecy
An environmental advocacy group is suing Wisconsin’s Public Service Commission for refusing to release electrical load projections for Meta’s planned $1 billion AI data centre in Beaver Dam, arguing the public has a right to know about the facility’s energy demands. The PSC is reportedly treating the energy figures as a trade secret at the request of either Meta or a public utility.
Researchers Sasha Luccioni and Boris Gamazaychikov have launched an updated AI Energy Score leaderboard, revealing that AI models with reasoning capabilities – where the model works through problems step-by-step (when we see “thinking…”) in an internal monologue before answering – use on average 30 times more energy than standard models, with some using up to 700 times more energy for the same task.
An Essex couple have become the first in the UK to heat their home using a small data centre installed in their garden shed, cutting their monthly energy bills from £375 to as low as £40. The system captures heat generated by over 500 matchbox-sized Raspberry Pi computers arranged in modules of up to 56 units. As they process data, the heat generated is captured by oil and transferred to the hot water system.
SAP has released a whitepaper outlining how AI can help businesses achieve sustainability goals, such as generating reports 80% faster and optimising supply chains, while also addressing how to develop AI itself more sustainably through energy efficiency measures and ethical governance.
Mastercard creates real-time dashboard to track carbon footprint of individual products and services
Mastercard has developed a patent-pending management dashboard that tracks the energy consumption and carbon intensity of each of its products and services in real time, helping the company grow revenue by 12% in 2024 while cutting overall emissions by 7%.
Norway extends Arctic deep-sea mining ban to 2029 and cuts funding for seabed mapping
Norway’s newly elected government has halted all deep-sea mining activity in its Arctic waters until at least 2029 and cut public funding for state-led seabed mineral mapping, marking a significant reversal from the country’s previous pro-mining stance.
AI’s 2025 carbon emissions equivalent to New York City’s annual output, study finds
Research published in the journal Patterns estimates that AI systems produced up to 80 million tonnes of CO2 in 2025 – equivalent to New York City’s annual emissions – while AI-related water use reached 765 billion litres, exceeding total global bottled water demand.
An analysis by Rest of World mapping global data centre locations against climate data found that nearly 7,000 of the world’s 8,808 operational data centres sit outside the recommended temperature range for efficient operation, with around 600 located in areas considered too hot, where cooling systems must work harder and energy costs increase.
Finnish researchers develop carbon capture compound that releases Co2 without extreme heat
Researchers at the University of Helsinki have created a new compound that captures 156 milligrams of CO2 per gram from ambient air and releases it when heated to just 70°C for 30 minutes, compared to the 900°C typically required by conventional carbon capture materials.
The University of Southern Denmark has launched a new AI and high-performance computing data centre in Sønderborg, developed with Danfoss and HPE, which uses advanced cooling systems to capture and reuse all surplus heat while running on Danish-developed open-source cloud technology.
Academic Papers
Now for a summary of some of the most important new digital sustainability research:
Author: Rianne Riemens
Summary: A fantastic analysis of how major tech companies use sustainability campaigns, reports, and advertising to position themselves as responsible actors in the climate crisis. Through case studies of Amazon’s Climate Pledge, Apple’s Carbon Neutral campaign, and Microsoft’s AI for Good programme, the author identifies four legitimisation strategies underpinned by an “ecomodernist” worldview. The paper argues this “tech-on-climate discourse” is a form of greenwashing that allows tech companies to present themselves as essential environmental agents while avoiding fundamental changes to their business models or broader societal structures.
From FLOPs to Footprints: The Resource Cost of Artificial Intelligence
Authors: Sophia Falk, Nicholas Kluge Corrêa, Sasha Luccioni, Lisa Biber-Freudenberger, Aimee van Wynsberghe
Summary: This study moves beyond energy and water consumption to quantify the material footprint of AI training by analysing the elemental composition of an Nvidia A100 GPU, identifying 32 elements with heavy metals making up around 90% of the hardware. The findings highlight that incremental performance gains between AI models come at disproportionately high material costs.
Authors: Nour Rteil, Rich Kenny, Deborah Andrews, Kristina Kerwin
Summary: This research examines the embodied carbon emissions of data storage devices, finding substantial gaps and inconsistencies in current industry assessments. A key finding is that SSDs have significantly higher embodied carbon than HDDs – on average about eight times higher for equivalent capacity – due to the energy-intensive nature of semiconductor fabrication. The authors also found major discrepancies between manufacturers’ own assessments, with supposedly similar SSDs from HP and Dell showing close to a 4x difference in reported impact, highlighting the urgent need for standardised life cycle assessment frameworks for storage technologies.
Authors: Tianqi Xiao, Francesco Fuso Nerini, H. Damon Matthews, Massimo Tavoni, Fengqi You
Summary: This study models the predicted combined energy, water, and climate impacts of AI server deployment across the United States between 2024 and 2030. Critically, the study finds the AI server industry is unlikely to meet net-zero aspirations by 2030 without substantial reliance on uncertain carbon offset and water restoration mechanisms. While best practices could reduce emissions and water footprints by up to 73% and 86% respectively, their effectiveness is constrained by current energy infrastructure limitations.
Investigating the Potential of Kepler Towards Power Observability for Sustainable Cloud Computing
Authors: Zouhir Bellal, Laaziz Lahlou, Nadjia Kara, Timothy Murphy, Tan Phat Nguyen, Arif Ahmed, Mario Perez-Jimenez
Summary: This paper evaluates Kepler, an open-source tool for measuring power consumption at the container level in cloud environments. The researchers developed a new framework for testing the accuracy of such tools and found significant limitations: Kepler can overestimate power consumption by up to 15 times, with accuracy highly sensitive to factors like CPU configuration and power state transitions. The findings highlight the need for improved power monitoring tools if cloud computing is to become more sustainable.
Resource corner
Each month, we share a digital sustainability report, tool, or resource that we think is worth your time.
Government Digital Sustainability Alliance (GDSA) digital scope 3 insights report
This month’s featured resource comes from the Government Digital Sustainability Alliance (GDSA), a collaborative body bringing together UK government, its ICT supply chain, academics and third sector organisations to drive progress on digital sustainability. Established at COP27 in 2022 and chaired by Chris Howes, Defra’s Chief Digital and Information Officer, the GDSA operates through working groups focused on Scope 3 emissions, Circular Economy, and Planetary Impact.
The newly published Digital Scope 3 Insights Report provides practical guidance for public sector teams to understand, measure and reduce digital scope 3 emissions – the carbon embedded in hardware manufacturing, transport, end-of-life treatment and cloud services. This matters because scope 3 typically represents 70-90% of an organisation’s total digital carbon footprint, yet remains poorly understood and often excluded from decision-making.
The report walks readers through establishing baselines, profiling suppliers, enabling sustainable procurement, and engaging users – backed by real case studies.
You can download the full report for free here. To learn more about the GDSA, visit the UK Government Sustainable ICT blog or the GDSA page on GOV.UK.
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:
Jo Lindsay Walton digs into a claim that has become central to arguments that AI will be a net climate benefit: the idea that AI could improve solar and wind load factors by up to 20%, saving around 1.8 GtCo2e annually by 2035. His investigation reveals the evidence base for this widely-cited figure is surprisingly shaky.
Balancing AI innovation and sustainability: our presentation at HM Treasury ID25
A reflection on Oliver Cronk and Suzanne Angell of Scott Logic’s presentation to government and industry leaders at HM Treasury’s Innovation Day, challenging the framing that positions AI innovation and sustainability as competing forces. Their core message: “Innovation without sustainability is short-lived. Sustainability without innovation is stagnant.”
Practical guide for Green IT strategy
Natallia Kachura distils insights from the GreenIO Paris Conference into really useful actionable strategies for IT sustainability. Having served as Master of Ceremony, she translates three days of presentations into a practical roadmap covering cloud computing, AI governance and measurement tools.
Data centre sustainability roundup – a year in review
Antonia Collings, CEO of Enaxiom, offers a comprehensive look back at 2025. Her assessment: this wasn’t the year sustainability became important for data centres; it was the year it became unavoidable, showing up as a real constraint on where facilities can be built, how quickly they can scale, and what “good” increasingly looks like.
Blind spot sustainability: making AI’s environmental impact measurable
Lena Winter and Theresa Züger highlight a fundamental problem: we still lack the data, measurement methods and binding standards needed to properly assess AI’s environmental footprint. Their article introduces a new five-year research project, Impact AI, which aims to systematically evaluate 15 AI initiatives to understand both the sustainability of AI systems themselves and their contribution to environmental goals.
From enthusiasm to execution: why operational rigor will define climate tech’s next chapter
With climate tech attracting over $312 billion globally between 2014 and 2024, Christopher R. LeWand and colleagues at FTI Consulting argue that climate tech investment has outpaced operational discipline. This piece outlines key areas investors must address to de-risk climate technologies in an increasingly uncertain policy environment.
AI and environmental justice in 2026 – what might come next
Mary Stevens, writing for Friends of the Earth, reflects on the major developments in AI and environmental justice during 2025 and outlines what campaigners should watch for in the year ahead. She argues that while the power of Big Tech may seem unassailable, cracks are beginning to show as resistance movements grow and legal challenges mount.
Blockchain-based mineral tracking: essential infrastructure requirements for 2025
Muflih Hidayat provides an extensive technical overview of blockchain-based mineral tracking systems in Discovery Alert, examining the infrastructure, costs and challenges involved in implementing distributed ledger technology for supply chain verification in mining. The piece is detailed but notably cautious, warning that premature blockchain deployment without proven operational foundations may create new risks rather than reduce them.
What will it take to achieve net-positive AI energy by 2030?
Maria Basso, James Mazurek and Eric Enselme outline a World Economic Forum framework for ensuring AI’s energy gains outweigh its consumption. Drawing on over 130 real-world use cases across more than 15 countries, they present a blueprint for aligning AI growth with energy transition goals – arguing that power availability, not processing power, is fast becoming the key limit to AI’s expansion.
How to balance AI adoption with ESG commitments
Veronica Hannon, founder of Transform Comms, writes in People Management about the tension HR leaders face between driving digital transformation and maintaining environmental commitments. She argues that responsible AI adoption should be treated as a cultural shift rather than just a technological one, noting that while many understand AI’s benefits, far fewer grasp its environmental impact – and offers practical steps from vendor assessment to employee education.
‘Just an unbelievable amount of pollution’: how big a threat is AI to the climate?
Ajit Niranjan takes a deep dive into AI’s climate impacts for The Guardian. The article explores whether the technology’s potential benefits can outweigh its mounting environmental costs.
Beyond municipal water: alternative solutions for data centre cooling needs
Christopher Tozzi explores alternatives to municipal water supplies for data centre cooling in Data Centre Knowledge. With municipal water reportedly accounting for about 95% of all data centre water use. Tozzi examines five options that could reduce strain on public systems, such as utilising water from rivers and lakes, and wastewater from municipal drainage.
The ‘greener’ the data center, the greater the potential savings
Tricia Crimmins reports in Tech Brew on how data centres prioritising clean energy and efficiency can reduce costs while cutting emissions. The piece profiles several companies taking different approaches – from building next to underutilised renewable farms to holistic building design to hardware optimisation.
The 90% reality: making the responsible choice the easy choice
Baran Bedir interviews Robert Keus, CEO of GreenPT about building a privacy-first, sustainable AI platform hosted entirely in Europe. Keus’s central argument is striking: 90% of AI tasks can be handled by open-source models, yet most organisations default to the biggest, most resource-hungry options when they actually don’t need to.
This infrastructure would save the planet, but cloud giants refuse to build it
Wilco Burggraaf makes a provocative but compelling argument on Medium that the biggest source of waste in modern infrastructure isn’t inefficient hardware, it’s unnecessary work. And the infrastructure that could fix this won’t come from hyperscalers because it directly contradicts their business model of monetising consumption.
GreenOps & FinOps series brief, counting the cost of cloud
Adrian Bridgwater explores the convergence of GreenOps and FinOps in Computer Weekly, noting that as GreenOps becomes a core requirement of IT operations alongside security and performance, the financial and environmental dimensions of cloud computing can no longer be treated as separate concerns.
The grid isn’t a cluster: what technologists get wrong about energy
Dave Masselink tackles common misconceptions about carbon-aware computing on the Compute Gardener blog, arguing that technologists often apply mental models from software to electrical grids…and get it wrong. The core problem, according to Masselink is that grids aren’t fixed-capacity compute clusters with zero-sum allocation; they’re economic systems with markets, slack, and operators who adapt to changing patterns.
Rethinking AI: why starting smaller might be smarter
Alex Bardell, Chair of the BCS Green IT Specialist Group, questions whether large language models are the right tool for every job, or whether we’re using a mallet to crack a nut. With rising energy demands and environmental concerns, Bardell makes the case for small, task-specific models as a more sustainable alternative.
Podcast pick
Each month we highlight a podcast episode that brings fresh insight into digital sustainability and climate tech.
Beyond the Hype: Will sustainable AI ever be a reality?
This month’s recommendation comes from Beyond the Hype, the podcast from Scott Logic. Host Oliver Cronk is joined by colleagues James Camilleri and Hélène Sauvé to explore whether AI can ever be genuinely sustainable – or whether “green AI” is more marketing hype than reality.
The conversation spans the environmental impact of AI from carbon emissions to water usage, examining whether industry narratives match what’s actually happening on the ground. The trio delve into the trade-offs between energy and water consumption, the importance of lifecycle analysis (looking beyond just operational emissions to manufacturing and disposal), and practical approaches that could help – including federated learning, smaller models, and edge computing.
What makes this episode particularly valuable is its focus on what organisations can actually do. Rather than getting stuck on the scale of the problem, the discussion offers concrete steps for making AI more sustainable in practice. It’s a thoughtful, technically-grounded conversation that cuts through the noise.
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.
