The UK government has set ambitions to become a global leader in AI research, safety and commercialisation.
Over £2.5bn has been committed to compute infrastructure, with new supercomputers being built in Edinburgh and Bristol, and “AI Growth Zones” being launched across regions including Greater Manchester and the West Midlands.
But a growing chorus of voices in the telecoms and infrastructure sectors is warning that these ambitions won’t be realised without a serious upgrade to the UK’s digital and physical backbone.
“We could end up an inference-only nation if we don’t get it right, they will go elsewhere”, Lee Myall, chief executive of Neos Networks, a major UK business fibre provider, told City AM.
The concern is that while the UK may succeed in attracting AI companies, it could fall short in hosting the infrastructure needed to train and operate the next generation of AI models at scale.
The UK’s AI infrastructure ambitions
In November 2023, the UK hosted the inaugural AI Safety Summit at Bletchley Park, where Prime Minister Rishi Sunak said the country would “lead the charge” on AI governance and innovation.
Shortly after, the government confirmed a £2.5bn commitment to building new compute capacity, including the Isambard-AI and Dawn supercomputers.
Since then, the Department for Science, Innovation and Technology (DSIT) has pledged to treat data centres as nationally significant infrastructure, giving projects a clearer path through planning.
What’s more, Keir Starmer has thrown the weight of Whitehall behind AI in a bid to boost growth and position the UK as a world leader in the sector.
He announced the UK’s AI Opportunities Action Plan’ earlier this year, a transformative strategy to establish the UK as a global leader in AI.
He said: “The AI industry needs a government that is on their side, one that won’t slip back and let opportunities slip through its fingers.
“And in a world of fierce competition, we cannot stand by. We must move fast and take action to win the global race.”
The announcement follows significant private sector commitments, including £14bn in AI infrastructure investments from Vantage data centres, NScale and Kyndryl, creating 13,250 jobs across the UK.
These moves are part of a broader strategy to position the UK as a full-stack AI economy – not just a consumer market for global platforms.
“We are building the necessary infrastructure to support AI’s growth”, a DSIT spokesperson said, “ensuring that both compute capacity and digital connectivity can scale as the technology evolves.”
Infrastructure bottlenecks remain
Despite this momentum, industry experts say major bottlenecks in fibre, power and planning still threaten to limit progress.
Myall argues that AI workloads demand a fundamentally different kind of network – one designed for high-throughput, low-latency, enterprise-grade data transfer.
“AI is an incredibly different proposition. You’ve got very high data velocity. That means the creation of data is prolific, the movement of it even more so,” he says. “We’re pushing the envelope. Curation of efficient, high-capacity routes is becoming harder — and it’s only going to get harder.”
Analysts from TechUK and Frontier Economics have made similar points, noting that as AI models grow in complexity, so does the need for secure, high-capacity connections between data centres, compute hubs and edge locations.
At the same time, energy availability – particularly grid access for large-scale data centre builds – remains a sticking point.
National Grid’s own modelling suggests future demand from digital infrastructure may significantly outpace existing plans unless additional investment is unlocked.
Project reach
One example of how infrastructure might be delivered more efficiently is Project Reach, a £300m joint venture between Neos Networks and Network Rail.
The project will see 1,000 kilometres of high-capacity fibre laid along the UK’s rail corridors – a move that also aims to eliminate mobile blackspots in tunnels and improve resilience across the network.
“Project Reach will support the upgrade of the UK’s connectivity infrastructure, creating new data superhighways that will drive the UK’s digital ambitions forward”, said Myall in a statement at the time.
By using pre-permitted rail routes rather than digging up roads, the project sidesteps many of the planning and delivery delays typically associated with large fibre deployments.
Officials at DSIT have pointed to this model as a potential blueprint for scaling infrastructure across utilities and transport corridors.
The challenge of balancing ambition and delivery
While recent government action has been welcomed by industry, particularly the commitment to AI compute capacity, the challenge now lies in coordination and speed of delivery.
Myall warns that private sector investment is already moving, and that policy must now match the pace of innovation.
“We’ve got good statements of intent. But we need more. The private money is already moving”, he told City AM.
The government’s forthcoming AI Infrastructure Roadmap, expected later this year, will address planning reform, energy strategy and fibre deployment in more detail.
The hope among stakeholders is that it will clarify where public investment will focus – and where the private sector is expected to lead.
Other experts have agreed that timing is critical. A recent report by the Tony Blair Institute said the UK must “accelerate delivery of fibre, grid connectivity and compute hubs simultaneously”, warning that the AI race will be won not just on models, but on logistics.
The UK has staked a claim on AI innovation. Its commitment to safety leadership, compute investment and growth zone development has positioned it as a serious player on the world stage.
But with AI models becoming more resource-intensive and data flows increasing exponentially, the hidden enabler of Britain’s AI ambition may prove to be less glamorous – fibre in the ground, power in the grid, and policies that help move things faster.
But without them, the risk isn’t just falling behind, it’s watching the benefits of AI scale elsewhere.
