Is universities' AI embrace undermining their net-zero targets?
University AI adoption now clashes with net zero pledges
Universities across the UK are racing to integrate artificial intelligence into teaching, research, and administration. However, this rapid adoption is creating a significant problem. AI systems require substantial amounts of electricity, generate considerable cooling demands, and produce indirect emissions that many institutions have not yet measured. Consequently, universities risk undermining their own climate commitments while promoting AI as a strategic priority.
The issue has moved beyond theoretical concern. Higher education leaders and sustainability professionals now recognize that AI carries a material environmental cost that belongs in carbon accounting frameworks. Moreover, the emissions often remain hidden in cloud contracts and aggregated IT budgets, making them easy to exclude from standard reporting.
For UK businesses, particularly those that supply universities or compete for public sector contracts, this debate signals a broader shift. Digital infrastructure and cloud computing are becoming carbon accounting issues, not just operational considerations. Therefore, companies that understand and measure the emissions profile of their digital services will hold a clearer advantage in procurement processes.
Computing infrastructure drives the emissions problem
AI relies on energy-intensive infrastructure, specifically data centres and high-performance computing clusters. These facilities consume electricity continuously and require substantial cooling systems to maintain operational temperatures. Additionally, the emissions split across two categories: direct energy use on campus and indirect emissions from cloud services and third-party tools.
Data centres globally used an estimated 415 terawatt-hours of electricity in 2024. Projections suggest this figure will reach 945 TWh by 2030. In the EU, data centres consumed between 45 and 65 TWh in 2022, representing approximately 1.8 to 2.6 percent of total EU electricity demand. Notably, Ireland's data centres alone consumed 21 percent of all metered grid electricity in 2023.
The environmental cost extends beyond electricity. Water consumption for cooling, hardware production emissions, and electronic waste all contribute to the total footprint. UNESCO has warned that as universities integrate large language models into routine platforms, the cumulative demand from thousands of daily queries by students and staff can scale rapidly. Furthermore, research estimates that training GPT-3 consumed 1,287 megawatt-hours of electricity and produced approximately 552 metric tons of carbon dioxide equivalent.
One study found that a single user making 50 GPT-3 queries daily for a year could consume as much electricity as a modern refrigerator. Similarly, training a single large model can emit carbon equivalent to hundreds of long-haul flights or several cars over their entire lifetimes. These figures illustrate why AI cannot be treated as a neutral productivity tool when institutions are pursuing net zero targets.
Sector guidance now demands vendor transparency
In May 2025, Times Higher Education reported that universities face accusations of overlooking the hidden environmental impact of AI while publicly committing to sustainability. The article framed this as an institutional contradiction: rapid technology adoption without corresponding carbon measurement.
By June 2026, the sector response had sharpened considerably. New guidance called on universities to demand emissions data from AI vendors, audit their own AI usage, and educate staff and students about the technology's environmental impact. This shift matters because it moves responsibility from abstract concern to concrete procurement and governance requirements.
Academic research has reinforced the case for action. A 2026 systematic review in the journal Sustainability concluded that operational electricity consumption, carbon emissions, and water demand are the most frequently reported impacts of generative AI in education. However, the review also noted that evidence in educational settings remains limited and inconsistent. Nevertheless, it stated clearly that computing energy and data centre emissions should be included in institutional greenhouse gas audits and climate action plans.
For UK SMEs, this development carries practical implications. Universities represent a significant market for software, cloud services, and IT infrastructure. As institutions begin requiring emissions data from suppliers, companies that cannot provide transparent carbon metrics for their digital products will face procurement barriers. Therefore, businesses should prepare to document the energy consumption and emissions profile of their services.
Why measurement gaps create reputational risk
Universities promote AI as essential for research productivity, student support, and administrative efficiency. At the same time, many advertise themselves as climate leaders with ambitious net zero targets. This combination creates a credibility problem when AI-related emissions go unmeasured.
The risk is both practical and reputational. If institutions do not account for AI emissions, they undercount their real carbon footprint and weaken the integrity of net zero pledges. Moreover, this gap undermines public trust in university sustainability commitments, particularly when students and staff increasingly scrutinize institutional climate action.
A governance problem compounds the measurement challenge. AI emissions typically hide within cloud contracts or aggregated IT budgets rather than appearing as distinct line items in sustainability reports. Consequently, procurement decisions about digital services proceed without meaningful carbon oversight, even though their cumulative impact may be substantial.
This pattern extends beyond universities. UK businesses in any sector that operate under net zero commitments face similar risks if they expand cloud computing and AI use without measuring the associated emissions. Furthermore, companies that supply public sector organizations should anticipate that carbon accounting for digital services will become standard procurement criteria.
Essential facts about AI energy consumption
- Global data centres consumed approximately 415 terawatt-hours of electricity in 2024, with projections reaching 945 TWh by 2030.
- EU data centres used an estimated 45 to 65 TWh in 2022, equal to roughly 1.8 to 2.6 percent of total EU electricity consumption.
- Ireland's data centres accounted for 21 percent of all metered grid electricity in 2023, illustrating concentration effects in specific regions.
- Training GPT-3 required an estimated 1,287 megawatt-hours of electricity and generated approximately 552 metric tons of carbon dioxide equivalent.
- Research suggests that a single user making 50 daily queries to GPT-3 for one year could consume electricity equivalent to a modern refrigerator's annual use.
- Training a single large AI model can produce emissions comparable to hundreds of long-haul flights or several cars over their operational lifetimes.
How universities must adapt procurement and policy
The solution does not require universities to abandon AI entirely. Instead, institutions need to treat AI as a material sustainability issue requiring the same rigor applied to buildings, travel, and energy supply. Specifically, this means establishing carbon accounting processes for digital infrastructure and cloud services.
Procurement represents the most direct intervention point. Universities can require AI vendors to disclose the energy consumption and emissions profile of their products and services. Additionally, institutions can prioritize suppliers that use renewable energy for data centres or demonstrate measurable efficiency improvements. This approach creates market pressure for greener computing infrastructure across the supply chain.
Internal policy must also evolve. Universities should audit their current AI usage to establish a baseline understanding of where and how these tools are deployed. Furthermore, institutions can set usage guidelines that discourage unnecessary queries or excessive model training when simpler alternatives would suffice. Education plays a role here as well, since staff and students who understand the carbon cost of AI may use these tools more selectively.
Carbon reporting frameworks need updating to include digital services explicitly. Many institutions currently track emissions from electricity, heating, and transport but lack categories for cloud computing and AI. Therefore, finance and sustainability teams must collaborate to ensure that IT procurement decisions route through carbon approval processes, not just budget approval.
For UK businesses, these institutional changes signal a broader market shift. Companies that provide transparent emissions data, demonstrate energy efficiency, and align with customer net zero targets will gain competitive advantage. Conversely, suppliers that cannot or will not disclose the carbon footprint of their digital products will face increasing procurement obstacles, particularly in the public sector and among organizations with strong climate commitments.
At SBS, we help businesses navigate exactly this kind of compliance and procurement challenge. Our net zero program supports organizations in measuring emissions from all sources, including digital infrastructure, and developing credible reduction strategies that stand up to scrutiny.
Implications for public sector suppliers
The university sector's focus on AI emissions reflects a wider trend in public procurement. Government frameworks increasingly require suppliers to demonstrate carbon measurement and reduction across their operations. Digital services represent an emerging frontier in this requirement, particularly as cloud computing and AI become more prevalent.
UK businesses that supply central government, local authorities, NHS trusts, or universities should anticipate questions about the emissions profile of their digital products. This includes software as a service, cloud hosting, data analytics platforms, and any AI-enabled tools. Consequently, companies need systems to measure and report these emissions accurately.
The challenge extends to supply chain transparency. A business may purchase cloud services from a third party and then resell those services or incorporate them into its own products. In this scenario, the business must obtain emissions data from its upstream suppliers to provide credible information to downstream customers. Therefore, carbon accounting for digital services requires collaboration across the entire value chain.
Procurement frameworks such as PPN 06/21 already require suppliers bidding for central government contracts above certain thresholds to publish a carbon reduction plan. While current guidance focuses primarily on traditional emission sources, the direction of travel is clear. Digital infrastructure will increasingly feature in these requirements as measurement methodologies mature and sector guidance evolves.
Businesses that prepare now will hold a competitive advantage. This preparation involves establishing energy and emissions baselines for digital operations, engaging with cloud and software vendors about their environmental data, and developing internal policies that prioritize energy-efficient computing choices. Moreover, companies should document these efforts in formats that align with emerging reporting standards, making it straightforward to respond to procurement questionnaires and tender requirements.
Our compliance support service helps businesses understand and meet these evolving requirements, including carbon reporting for public sector tenders and supply chain emissions measurement.
Where to find authoritative guidance
The Department for Energy Security and Net Zero provides policy direction on UK carbon reduction targets and reporting frameworks. Additionally, the government's Procurement Policy Notes set out requirements for public sector suppliers, including carbon reduction plans.
For sector-specific guidance on emissions measurement, the Institute of Environmental Management and Assessment offers technical standards and professional development resources. Universities and businesses alike can benefit from IEMA's frameworks for calculating Scope 1, 2, and 3 emissions, which increasingly include digital infrastructure.
The Times Higher Education continues to report on sustainability developments in the university sector, providing valuable context for businesses that work with higher education institutions. Finally, academic journals such as Sustainability publish peer-reviewed research on AI environmental impacts, offering evidence-based perspectives that inform policy and practice.