This year’s Spending Review highlighted a critical issue that is often buried in the hype of AI. Huge budgets are being piled into AI development – including the Government’s latest promise of £2 billion in public funds – but without equal investment in data center capacity, power availability and sustainability, the country risks falling short of its AI and R&D ambitions.
While AI demand is scaling exponentially, our infrastructure isn’t. The UK’s AI ambitions rest on compute power, but compute doesn’t exist in a vacuum – it lives in data centers. These infrastructures therefore stand as the physical foundation of digital innovation, yet they’re often left as an afterthought.
Politicians talk about AI breakthroughs, but rarely divulge the power, cooling and land needed to support them. Agreed, it’s not the ‘sexiest’ part of global AI discussions, but it’s a vital part, nevertheless.
In the worst-case scenario – which is one we’re rapidly heading towards – this gap in infrastructure could become a national vulnerability. The industry is innovating, but retrofitting the past is harder than building the future. We need policy frameworks to support both.
VP of Professional Services for EMEA at Salute.
Infrastructure bottlenecks
The UK’s digital ambitions are running up against physical limitations, especially on the national grid. Without urgent upgrades, we’ll hit a hard ceiling on our ability to scale. These grid constraints are a major bottleneck, especially in London and the South East, where power shortages are already a consequence of underpowered infrastructure. Some providers are reporting that greater power availability in these high-demand areas won’t be available under 2035 or later.
The trouble is, factors like planning delays and general uncertainty in the sector around energy availability are slowing down the development of new data centers. More broadly, it’s also clear the UK lacks a coordinated national strategy for this digital infrastructure, unlike some of our global peers.
If we don’t act, the UK could become a net importer of compute power, outsourcing both innovation and control to more established countries. Are we really prepared to sit by and watch our AI innovation migrate abroad, simply because they have invested more in their infrastructure?
Economic and regional inequalities could also worsen if infrastructure gaps persist. Without greater data center investment outside London and the South East, underserved regions may miss out. The UK’s geopolitical position will also suffer a serious setback due to lack of innovation, control and AI capacity and capabilities.
Growing pressures for greener, scalable capacity
Fortunately, there is growing recognition of the need to build greener, more scalable infrastructure. Data center operators and developers are increasingly considering renewable energy sources, waste heat reuse and modular construction. However, many of these efforts are constrained by legacy decisions, particularly the locations historically chosen for data center builds, which often don’t align with the availability of clean energy or heat reuse networks.
At the same time, there’s a growing shift toward modular, energy-efficient builds which can reduce environmental impact and support more flexible deployment. However, retrofitting legacy sites remains a challenge for the larger AI workloads, particularly those requiring upwards of 500MW capacity.
While critical collaboration between government, energy providers and infrastructure is improving, it remains fragmented. Energy planning and digital infrastructure are often siloed, despite being deeply interdependent. That has to change. What’s needed is a whole-systems approach that treats energy and digital infrastructure as two sides of the same coin.
Transparency as a whole is improving, but more needs to be done. Consistent reporting standards are needed as well as greater information sharing from manufacturing to construction and into operations. This will allow for more accurate lifetime carbon assessments and improved accountability.
Innovation and infrastructure need to work in tandem
Demand for high-performance compute is growing exponentially, driven by generative AI and large-scale model training. These technologies are power-hungry and need robust infrastructure to support them. Without urgent investment, we risk becoming dependent on foreign infrastructure for critical AI workloads.
At the same time, there is a growing disconnect between the UK’s AI ambitions – and subsequent compute growth – and its sustainability targets. Without clear guidance on how AI infrastructure fits into the UK’s carbon budget, we risk undermining current and future commitments.
A national digital sustainability framework is urgently needed. This should include fast-track planning for sustainable data centers in underserved regions. Incentives should be introduced to support grid upgrades, renewable energy integration and co-location of data centers with district heating networks.
National standards for energy efficiency and environmental impact reporting must become mandatory, not optional. A framework like this would ensure that new infrastructure is built not only for scale, but also for sustainability.
A future-ready, AI-capable data centre ecosystem will be modular, regionally distributed and powered by clean energy. It will be closely integrated with local communities, sharing excess heat, providing skilled jobs and expanding digital access. It will be governed by clear, transparent standards that prioritize both innovation and environmental responsibility.
To achieve this vision, several things must happen now. We need faster grid connection timelines, clearer planning frameworks for edge and modular data centers and incentives for sustainable builds. It’s time to treat data centers like the critical national infrastructure it is.
We’ve featured the best green web hosting.
This article was produced as part of TechRadarPro’s Expert Insights channel where we feature the best and brightest minds in the technology industry today. The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here:
