TECH TALK
Are we really building a superpower?
In the UK, we have spent years talking about becoming an AI superpower. I do not doubt the ambitions of the UK AI Ambition Plan released at the start of 2025. And it is true that we have extraordinary universities, serious research talent and a strong start-up culture. But ambition without infrastructure is just branding.
When I speak to enterprise teams, the reality is often the same: they want to scale AI, certainly, but they cannot find domestic capacity that gives them confidence on cost, power stability and long-term availability. That is when they default to overseas cloud. And it is where the risk is.
What you cannot always do is secure timely grid connection in the South East or lock in predictable electricity pricing for the next five to ten years.
And that is where projects stall. Not because the models are not useful, but because the physical foundation is missing.
Power and infrastructure
We have leaned heavily on hyperscale models as if compute is infinitely elastic, but it is not. Power capacity in key regions is constrained. Planning processes are slow. Grid upgrades take time. Yet the expectation remains that any company can spin up massive AI infrastructure at will.
Bottom long caption and Bottom long caption.
I understand why. It is quick, familiar and feels safe. But it is also expensive and, over time, it creates dependency. We are encouraging organisations to adopt AI at speed while quietly accepting that the underlying infrastructure will sit elsewhere. That is not a sustainable strategy for a country that wants to compete at the highest level.
If you look beyond the UK, you can see a more coordinated approach emerging. In the Nordics, countries such as Norway, Finland and Sweden are leaning into their natural advantages in hydro and wind, pairing renewable power with clear data centre strategies and faster planning pathways.
In parts of the US, state-level incentives and proactive grid investment are being aligned with large-scale AI deployments, so power and compute move in step rather than in conflict. Even smaller European markets are positioning themselves around energy security and available land, recognising that AI infrastructure is as much an energy policy issue as a technology one.
The common thread is intent backed by execution. They are not just talking about AI leadership; they are building the physical capacity to support it.
AI workloads are heavy and we cannot get away from that. They demand dense racks, serious cooling and reliable grid access. You can procure GPUs and you can raise capital.
The reality is often the same: they want to scale AI, certainly, but they cannot find domestic capacity that gives them confidence on cost, power stability and longterm availability.
From where I sit, the conversation needs to shift from hype to basics. If we are serious about AI leadership, we need to treat land and power as strategic assets. That means thinking about where data centres are built, how they connect to renewable generation and how we secure longterm energy agreements that make the economics of AI viable.
It also means recognising that infrastructure is capital intensive and requires patient alignment between industry and government.
We have seen platform shifts reshape markets. Fibre replaced copper. Virtualisation changed server utilisation. Cloud transformed procurement. Each transition rewarded the players who understood the underlying infrastructure, not just the application layer. AI is no different.
Sovereignty is threatened
There is a sovereignty dimension here that we need to be honest about. Sovereign AI is often framed as a question of where data sits. From my perspective, it is also about who builds and operates the infrastructure.
If the UK does not develop and retain the engineering capability to design, deploy and run high-performance GPU environments, then calling it sovereign becomes a stretch.
This is not an argument for isolation. It is an argument for balance – for being a world leader. We should absolutely collaborate globally. But
48
INTELLIGENT CIO EUROPE www. intelligentcio. com