top of page
Search

AI’s Carbon Cost: Can Banks Stay on Track for Net Zero?

  • Dec 16, 2025
  • 8 min read

Banks are adopting artificial intelligence (AI) at unprecedented levels to improve operations and customer service. Recent data from the European Banking Authority shows that nearly all EU banks now use AI in some form, applying it to tasks such as profiling customers, automating internal processes, assessing creditworthiness, and strengthening AML and fraud detection systems. The same EBA note reports that 55% of banks already deploy general-purpose or agentic AI in consumer-facing processes, most notably for fraud alerts and customer support, showing how quickly these tools have become part of everyday banking (European Banking Authority, 2025).


Profiling or clustering of clients or transactions 

Creating customer groups by analysing behaviour and transactional patterns to uncover hidden similarities.


Optimisation of internal processes 

Automating repetitive tasks such as document classification and minute preparation to increase operational efficiency. 


Creditworthiness assessment and credit scoring 

Using large datasets to more accurately predict an individual’s likelihood of repaying credit. 


AML/CFT and Fraud detection 

Identifying suspicious behaviours through anomaly detection and enhanced verification processes to prevent financial crime.


Risk modelling

Detecting unusual transaction trends and customer sentiment changes that could indicate emerging risks.


Customer support 

Deploying AI-powered chatbots to provide faster and more personalised customer service. 

 


However, these digital gains come with a steep environmental cost. Training and running large AI models demands massive data-centre computing power, which drives up electricity consumption and carbon emissions. Analysts warn that AI workloads could account for as much as 35 - 50% of data-centre power usage by 2030 (Gabbatiss, 2025). This raises concerns that the sector’s tech-driven innovation might inflate banks’ carbon footprints just as they strive to reduce them. 

 

Major European financial institutions have publicly committed to align with the Paris Agreement’s climate goals, pledging net-zero emissions by 2050 with interim targets for 2025 and 2030 (Sustainable Finance Observatory, 2025). Yet, a late review by Green Central Banking (2025) found that banks were making very little progress on these commitments. As COP31 - the UN climate summit to be hosted by Turkey in 2026 - approaches, the spotlight is on whether these institutions can bridge the gap between digital innovation and decarbonization (WWF Australia, 2025).


In this post, we explore what the intersection of AI innovation and climate ambition means for the future of European banking. 

 

Net-Zero Banking by 2025: Are We on Track?  



Climate change is a global challenge that cannot be solved by individual nations alone. It is a problem that asks for collective action and unified solutions across countries and sectors. Recognizing this urgency, leaders of all EU countries came together at the UN Climate Conference in Paris on 12 December 2015 to reach a breakthrough agreement to combat the crisis: the Paris Agreement (United Nations, n.d.). The agreement officially came into effect on 4 November 2016, after it was approved by over 55 nations responsible for at least 55% of global greenhouse gas emissions. Key requirements of the Paris Agreement include the following (Council of the European Union, 2025): 

  • The global rise in average temperature should be kept below 1.5°C by the end of the century 

  • Countries are required to report publicly on their progress for transparency and global accountability 

  • Each country must commit to reducing its greenhouse gas emissions through a national climate action plan 

  • Wealthier nations should provide climate finance to help vulnerable countries cut emissions and adapt to climate change 

  • Every five years, governments are expected to submit updated plans that reflect even more ambitious climate targets 


Banking on the Paris Agreement: Where We Are Now  



According to RMI (2023), which draws on the ECB’s climate alignment analysis, the current picture is concerning:  

  • Nearly 90% of euro-area banks’ corporate loan portfolios are misaligned with the goals of the Paris Agreement. 

  • Over 70% of euro-area banks have made climate pledges, but their corporate lending portfolios remain misaligned with those goals. For example, many still finance carbon-intensive sectors like fossil fuels while postponing investments in renewable energy, thus increasing their exposure to regulatory and litigation risks. 

  • Electric vehicle (EV) financing is one of the rare areas showing alignment, thanks to strong projected growth in EV production. 

  • The EU regulators are expected to make transition plans for euro-area banks mandatory next year. One powerful tool gaining traction is PACTA (Paris Agreement Capital Transition Assessment), which shows whether a bank’s lending is aligned with climate goals. It uses forward-looking production data and company-level insights to guide banks’ own climate strategies and assess the credibility of their clients’ transition plans. For instance, banks can use PACTA to assess whether their financing of sectors like automotive or power is shifting fast enough toward low-carbon technologies. 

  • Besides a challenge, the climate transition is also a massive investment opportunity for banks. To meet the EU’s Green Deal goals, around €520 billion more will need to be invested each year this decade, coming from both public and private sources, with banks playing a key role in mobilizing that capital. Banks can seize this moment via transition finance, offering tailored products and support to help clients decarbonize. Tools like PACTA aid in identifying which clients are best positioned to transition and where financial support can have the most strategic impact. 

 

To sum up, most euro-area banks are still off track, but the tools, incentives, and regulatory pressure now in place offer a path to align their portfolios with climate commitments. 

 

What Drives Emissions? 



European banks are rapidly adopting AI across many functions - from fraud and anti-money-laundering screening to credit scoring and customer chatbots. However, each of these functions requires substantial computing. For example, Schwartz, Dodge, Smith, and Etzioni (2024) report that training a single deeplearning AI model can emit as much carbon as five cars over their entire lifetimes. 

 

All this AI work runs on data centres. Data centres are a fundamental element for processing user data, training AI models, and hosting cloud services. Desikan and Neff (2025) report from the University of Cambridge’s Minderoo Centre for Technology and Democracy projects that the number of data centres will increase dramatically in the years to come. Data centres, according to the authors, contribute slightly less than 1.5% of global emissions. This share is predicted to rise by 15 - 30% annually, potentially accumulating to 4% by 2030 and 8% by 2040. AI is a key driver: it already accounts for about 5 - 15% of current datacentre power use (IEA) and could reach 35 - 50% by 2030 (Gabbatiss, 2025). The United States, China, and Europe are major hubs for data centre energy consumption. In those countries, facilities already use approximately 2 - 4% of national electricity. In some regions, the concentration is even more intense - in Dublin, for instance, data centres now account for up to 20% of Ireland’s total electricity use (Desikan & Neff, 2025). Furthermore, the report suggests that emissions linked to major tech firms have surged in recent years. Between 2019 and 2023, Google saw its greenhouse gas emissions jump by 48%, while Microsoft’s emissions rose nearly 30% from 2020 to 2023. Amazon’s carbon footprint also grew by about 40% between 2019 and 2021, and although it has since declined, it remains well above its 2019 levels.  

 

These trends mean that banks’ AI workloads will increasingly draw on datacentre power. In fact, cloud versus on-premises infrastructure makes a big difference. Migrating banking IT to public cloud can dramatically cut emissions: for example, one Microsoft study found that cloud users emit up to 98% less CO₂ than equivalent on-premise systems (World Economic Forum, 2022). Cloud providers achieve this through highly efficient cooling, economies of scale, and renewable power. Conversely, if a bank spins up its own AI servers, their electricity footprint depends on location and technology. Hosting servers where grids are clean (e.g. hydro-rich Norway) yields far lower emissions than on carbon-intensive grids, states the Bank for International Settlements (2023). In practice, many banks now use large cloud data centres operated by third parties (AWS, Azure, Google, etc.), shifting the computation off their books, though not off the planet. 


Scope 2 and Scope 3 Emissions  


In carbon accounting terms, Scope 2 covers the bank’s purchased electricity (e.g. powering its offices or owned data centres), whereas Scope 3 covers outsourced or upstream activities. For AI, this distinction matters. If a bank runs its own server farm, the power it buys is Scope 2. If instead it runs AI on rented cloud platforms, the cloud provider’s emissions count as the bank’s Scope 3 (indirect) footprint (Bernard, de Lange, Lakshmanan, & Likens, 2025). Similarly, all the embodied emissions in servers, networking gear and chips are Scope 3. Regulators are increasingly demanding that these be tracked: EU and US supervisors are considering including indirect ICT emissions in climate stress tests, and major banks (like HSBC and others) expect stricter Scope3 disclosure rules (World Economic Forum, 2022).


How to Ensure AI in Banking Supports Real Sustainability Goals 



Despite the environmental concerns discussed earlier, The European Magazine offers a contrasting view: AI in digital banking can actively support sustainability goals (Kaye, 2025): 

  • Traditional banking has relied heavily on paper processes and energy-intensive data centres, which significantly contribute to its environmental footprint. AI helps reduce this impact by replacing physical call centres with virtual assistants and cutting down unnecessary communication through smarter customer forecasting. What’s more, AI is helping banks cut energy use and inefficiencies by automating key back-office operations. Tasks that once relied on manual oversight, like fraud detection, are now handled faster and more sustainably by smart algorithms. 

  • It’s argued that AI plays an important role in advancing green finance by helping banks respond to growing demand for sustainable investment options. It enables them to assess environmental, social and governance (ESG) factors more effectively by analysing large volumes of complex data. Thus, thanks to AI, banks assess the sustainability of companies and projects with greater precision, helping them allocate capital more responsibly. 

  • AI is helping bridge the global financial inclusion gap by enabling access to essential banking services for people currently excluded from the formal financial system. For example, it can analyse non-traditional data like mobile usage or utility payments to generate fairer credit scores and unlock services for those without conventional credit histories. 


Final Reflections 


Morrison Finance sees the environmental footprint of AI as a core concern and integrates sustainability into how we approach AI-driven transformations. In our role as advisors and implementation partners, we actively share insights and guide clients toward choices that align with greener, more responsible AI practices.  


As we move from COP30 in Belém toward COP31 in Antalya, the pressure on financial institutions to demonstrate credible climate action is mounting. Now is the time for boards and executives to ask practical questions: Do we understand the emissions tied to our AI systems? Are they accounted for in our net-zero plans? Do we assess the energy impact of every new AI project - from fraud detection to customer support? The real challenge isn’t choosing between innovation and climate action - it’s making sure they advance together through measurable, responsible choices. 



References

Comments

bottom of page