AI & Sustainability: Bi-Weekly Update on AI Infrastructure, Innovation, and Inequality

From Britain’s drought-hit AI buildout to Europe’s ambition to lead in sustainable infrastructure, this edition explores the growing tension between digital expansion and environmental stewardship. We examine the UK government’s attempt to reconcile AI ambition with ecological limits, assess water risks from Khazna and Eni’s new facility in Italy, and expose the hidden labour behind the models we rely on. Plus: new research on how to make AI dramatically more efficient, and what the Clarity AI–ecolytiq deal means for the future of green fintech.

1. The Water-Energy AI Nexus: Britain’s Balancing Act

Britain is attempting to position itself as a global leader in AI infrastructure – a logical move for a services-based economy looking to capture the gains of digital transformation. Amazon, Microsoft, Google, Oracle, and IBM have announced investments totalling £15.7 billion. Amazon alone plans to invest £8 billion over five years, projected to contribute £14 billion to GDP while supporting 14,000 jobs annually (Al Saqqaf, 2025).

However, this growth is colliding with environmental reality. UK reservoirs are currently at just 77% capacity – dramatically below the seasonal norm of 93% – yet hyperscale data centres are being approved across the country, each consuming as much water as 100,000 households (Ibid). Thames Water has proposed technical measures such as flow restrictors and peak-time surcharges, but government policy appears skewed toward prioritising infrastructure expansion. In May 2025, Deputy Prime Minister Angela Rayner overturned a council decision to block a 96MW data centre in Hertfordshire, citing updated national planning guidance that obliges local authorities to accommodate strategic infrastructure (Ibid).

The issue is compounded by the fact that these facilities are often built in densely populated, resource-stressed areas like London, which has over 5,700 people per km² (ONS, 2020). In contrast, Scotland’s population density is just 70 people per km² (Ibid), and it produces more renewable electricity than it consumes (Scottish Government, 2024), yet it hosts relatively few data centres. Locating AI infrastructure in regions like Scotland would not only ease pressure on overstretched systems but also bring it closer to abundant renewable energy. The government’s AI Growth Zone initiative aims to correct this imbalance by directing investment toward resource-rich areas such as Scotland, Wales, and Northern England. If implemented effectively, it could help mitigate the water-energy strain while stimulating rural economic development (Al Saqqaf, 2025).

2. Global Implications: Khazna, Eni, and Water Security in Italy

Europe is also investing in its digital future. Khazna Data Centres and Italian energy giant Eni have announced plans to co-develop a 500MW AI-optimised data centre in Ferrera Erbognone, Lombardy. The facility will be powered by “Blue Power”—a form of low-carbon energy generated via Combined Cycle Gas Turbines with integrated CO₂ capture (Sustainability Magazine, 2025). The project is framed as part of Europe’s ambition to become a sustainable AI hub and will also contribute to strengthening Italy’s ageing digital infrastructure, especially in underserved rural areas (Ibid).

However, the environmental narrative remains incomplete. While emissions may be addressed, the water implications are less clear. The site lies in the Po Valley, an agriculturally vital but climate-vulnerable region that, due to intensifying drought over the past two decades, increasingly resembles the arid conditions of Ethiopia and the Horn of Africa (The Guardian, 2023). Given that data centres require vast quantities of water for cooling, questions persist about how the project will impact local farmers, ecosystems, and long-term water security. Without adequate safeguards, even green-powered data centres risk exacerbating resource conflicts at the local level.

3. Ethical Blind Spots: The Hidden Human Cost of AI

Amid debates over water and emissions, a quieter crisis is unfolding: the human cost of training AI. Thousands of content moderators in the Global South—often employed by third-party outsourcing firms—are exposed to distressing content daily. Interviews by (Equidem, 2025) reveal that 53% of surveyed workers in Kenya, Ghana, Colombia, and the Philippines suffer from PTSD, depression, or anxiety. Many work 8–12-hour shifts reviewing traumatic material, including violence and abuse, all under strict NDAs that prevent them from seeking redress or even speaking openly (Jacobin, 2025).

This opaque labour regime reflects what researchers now call technofeudalism: a new form of digital domination not built on land ownership, but on control over data, algorithms, and invisible labour (Equidem, 2025). It’s an uncomfortable truth that could have consequences beyond ethics—growing scrutiny from ESG-conscious investors may ultimately lead to reputational and regulatory risks for AI developers reliant on these practices.

4. Smarter AI, Smaller Footprint

New research from (UNESCO, 2025) and UCL suggests that AI’s environmental footprint can be significantly reduced through targeted design improvements. Key findings include:

• Use smaller models tailored to specific tasks (e.g. translation, summarisation) instead of large general-purpose models. This reduces both energy and water usage while preserving performance.
• Shorten user prompts and model responses, which cuts down on computational load—reducing energy consumption by as much as 50%.
• Apply model compression techniques such as quantisation to reduce the number of bits used in calculations, enabling the same results with less energy.

These changes are not just about decarbonisation—they are also about access. With only 5% of Africa’s AI talent having access to sufficient computing infrastructure, more efficient models could democratise AI by making it usable in low-resource contexts (Ibid).

5. Financial Nudges and Behavioural Changes

Clarity AI’s acquisition of ecolytiq marks a shift in the application of AI to the consumer sphere. ecolytiq offers climate engagement tools—such as carbon tracking, access to green loans, and behavioural nudges—within banking apps. Now bolstered by Clarity AI’s analytics and Visa’s distribution network, these tools may help mainstream climate-conscious financial behaviour at scale (Fintech Futures, 2025).

By embedding environmental impact metrics into daily financial decisions, this merger offers a promising example of AI enabling systemic behaviour change from the ground up.

Final Thought

As nations race to lead the AI revolution, a simple truth remains: infrastructure choices matter, irrespective of geographic or socio-economic specificities. How and where we build data centres will shape the environmental, ethical, and social consequences of AI for decades to come. With smarter models, distributed growth, and transparent labour practices, AI can become an ally in sustainable development. Without them, it risks becoming another extractive industry in digital form.