Prime Minister Keir Starmer has made his ambition to make the UK an AI powerhouse clear, placing advanced technology at the heart of future economic growth. Delivering on that ambition will require a rapid expansion of digital infrastructure, particularly data centres capable of supporting increasingly energy-intensive AI workloads. But while the UK shares the goal, it should not assume that success depends on following the United States down the path of vast, resource-hungry data centre cities.
Growth without guardrails
The American model is already visible in Loudoun County, Virginia, dubbed “Data Center Alley,” home to the world’s largest concentration of data centres – responsible for roughly 70% of global internet traffic. It’s an impressive feat of infrastructure development, but one that is built largely without environmental guardrails. Individual large data centres in the region can consume up to five million gallons of water per day – equivalent to the water use of a town of tens of thousands of people.
As local communities push back against rising water use, power demand, heat and noise, the limits of this approach are becoming clear. As the UK stands on the brink of a data centre construction boom worth billions, the question isn’t whether we can replicate America’s model of scale at all costs, but whether we should. The answer is clear: we cannot and should not. But this constraint could become our competitive advantage. Done right, the UK’s constraints could underpin a more efficient and sustainable approach to building the AI infrastructure of the future.
The UK’s unique challenge
The UK government has committed significant resources to data centre expansion, designating them as Critical National Infrastructure in September 2024. Capacity is expected to grow from 1.6 GW in 2024 to between 3.3 GW and 6.3 GW by 2030, with spending on new facilities set to rise to £10 billion by the end of the decade.
The UK, however, faces constraints that the Americans do not. The most obvious is land availability – the UK could fit into the United States roughly 40 times over. On top of that, the UK maintains stronger commitments to the environment than across the pond. Planning laws also give notable influence to communities – a hint of NIMBYism is often enough to stop a development.
These constraints create a dilemma. The UK needs infrastructure expansion to remain competitive in the AI age, but we must deliver it within tighter spatial, environmental and regulatory limits. The American approach is neither realistic nor desirable. Yet AI’s infrastructure demands are real, and without a different strategy, they risk undermining the sustainability commitments that distinguish the UK from that model.
The green data centre dilemma
Heat generation remains one of the most pressing technical challenges. Every data centre is essentially a huge heat engine, requiring either air or water-cooling systems that themselves consume considerable energy and resources.
We’ve seen some progress like BREEAM-rated facilities and projects such as the Met Office’s research centre in Exeter demonstrate that environmental standards can be embedded into design. But these remain exceptions rather than the rule
At scale, the fundamental problem persists, achieving adequate power supply and effective cooling whilst maintaining an attractive price point for data storage is extraordinarily difficult. Unlike the US, we can’t simply build our way out of the problem by sprawling across available land.
Brownfield or Greenfield?
Location choices will shape the UK’s digital infrastructure for decades. Brownfield development offers the opportunity to regenerate previously developed land and connect into existing infrastructure. However, there could be challenges with power connectivity. Greenfield sites might offer easier development but at the cost of consuming valuable land and potentially damaging ecosystems.
This isn’t merely a technical decision. Prioritising brownfield development aligns with broader sustainability goals, preserves agricultural land and creates opportunities to integrate data centres into existing communities. When done well, this can deliver tangible local benefits, from employment to district heating schemes that capture and redistribute waste heat.
AI as the solution to its own problem
The narrative, however, can become more hopeful. AI itself may provide the tools we need to manage this challenge more effectively.
Small language models are proving more efficient than their larger counterparts – a significant shift from the “bigger is better” mentality that drove initial AI development. More importantly, AI can be deployed to clean up the data wastage that’s driving explosive storage demands. Through intelligent data retention policies and compression, organisations could dramatically reduce their storage footprint. We need collaboration between organisations that retain data and those that can help them become more efficient with it.
Digital twins offer another promising avenue. By mapping power infrastructure and data centre operations in a virtual environment, we can model efficiency improvements and resilience measures before implementing them physically. This could revolutionise how we approach power demand and cooling systems.
Smarter infrastructure and real leadership
To get this right, the UK must first accept a simple reality: data centres are no longer passive pieces of digital infrastructure. They are, and will increasingly be, active components of the wider energy ecosystem, with direct implications for power generation, grid resilience, water use and local communities.
Meeting that challenge will require more than individual projects or piecemeal regulation. The UK needs a coherent, integrated and adaptable policy approach – one that sets clear objectives and frameworks for how data centres are planned, powered, regulated and operated in the AI era.
Crucially, these approaches must be shaped through meaningful engagement between key stakeholders, including UK and devolved governments, planners and landowners, investors, utilities, IT and telecoms providers, data centre operators, regulators and wider society. Only through coordinated decision-making can competing pressures around energy, land, growth and sustainability be properly balanced.
If done well, this approach offers the UK a way to support its AI and digital ambitions in a more risk-balanced, regulated and genuinely sustainable manner. More importantly, it provides a route to building the infrastructure the future demands while avoiding, or at least minimising, unintended consequences for ecosystems and communities.
The opportunity is not simply to build more data centres. It is to build them better.
