Over the next decade, trillions of dollars will be invested globally in the infrastructure needed for the AI era.
Currently, Britain is unlikely to attract even a small share of this investment. It is a choice that will determine our strategic relevance. Without significant changes, Britain could jeopardise its long-term economic stability and national security, becoming reliant on the whims of foreign servers.
As the future of AI will ultimately be shaped by countries that can build and control its infrastructure, countries that do not build will quickly find themselves dependent on those that can. Today, Britain accounts for just under 3% of the world’s total compute, much of which isn’t even optimised for AI. As the other countries pour investment into AI infrastructure development, and the UK stands still, its share of global compute could drop to less than 1% within the next five years.
The reality today is that Britain is not going to compete with the US, China or the Middle East in the race to develop frontier AI models. The country lacks the companies, capital, land and cheap energy to do so. However, we do have the opportunity to become the best place in the world for deploying AI at scale.
In many places, the UK is uniquely positioned to take advantage of the productivity gains that AI can offer: we have the NHS, a globally competitive secondary and tertiary education sector, world-class natural and life sciences labs, we are at the forefront of the arms industry up and down the value chain and then there is the vast back-office machinery of the government itself. Deployment, not training, is where the real productivity revolution and the true national security advantage lie.
But deployment still needs infrastructure. Every ChatGPT prompt, every cancer screening and every AI tutor will require computational power. Without AI infrastructure on UK soil, personal and sensitive data will instead be sent to servers worldwide. Our country needs to have servers on its own soil, lest it become a vassal to those that do.
The government is aware of this and the recently released compute roadmap predicts the need for 6GW of compute capacity by 2030. But as Britain’s creaking infrastructure — its roads, water, rail and runways — continues to deteriorate, each requiring enormous investment, spending money on data centres might not be regarded as a vital investment.
The good news is that the country wouldn’t need to foot the bill itself: global capital is begging to spend money on this.
So what is the problem for Britain?
Investors read spreadsheets, not speeches. They observe the UK’s decade-long grid connection queue, soaring industrial power tariffs and a planning system that struggles to distinguish a cutting-edge data centre from a brownfield warehouse. All of this is causing the UK to become one of the world’s most vulnerable AI ecosystems, with little ability to guarantee its resilience.
A way forward
Fixing this first begins with energy.
A nation that wants to run AI all day, every day needs clean baseload power and fast. This means accelerating construction of the Sizewell C nuclear power station. It means approving small modular reactors at speed, all over the country. It requires a grid operator that treats AI demand as a strategic priority, not an afterthought. Data centre builders need clear timelines, flexible connection agreements, and the right to co-locate private generation. Do that and capital will flow.
Second, cut through the red tape. The government’s proposed AI Growth Zones should operate like digital enterprise zones on steroids. One minister approves land use, environmental hurdles and transmission lines in a single move; decisions are made in weeks, not years; local councils retain a share of the business rates. The aim isn’t to scatter vanity projects across the map; it’s to forge dense corridors of AI capacity where power, energy, and talent compound.
Third, treat compute as a public good. Ringfence a sovereign pool of UK-owned GPUs for emergencies, sensitive data and safety research. Then auction bulk advance purchase commitments, think vaccine task force for teraflops, to crowd in private builds. Let universities, startups and the MoD bid for cycles on equal terms and publish the queue in real-time. Nothing concentrates investor minds like a guaranteed demand signal.
Finally, get serious about delivery. The AI Opportunities Action Plan is a promising proposition, but Whitehall is still playing pass-the-parcel between government departments. Create an AI Infrastructure Delivery Group reporting directly to No10, staffed by grid engineers, planners and hardnosed project managers with authority to bulldoze bureaucratic blockers.
At a time of limited fiscal space, the argument against is that this is costly, while others will say it’s too carbon intensive. AI will consume energy. In the short term, it is likely to be carbon-intensive. But this is only temporary.
The massive expansion of AI infrastructure is driving unprecedented investment in clean and renewable energy, including investments in nuclear and fusion energy. Investing in AI will not only help unleash abundant clean energy but it will also play a crucial role in fostering the innovations needed for climate adaptation and mitigation.
History won’t remember how many strategic plans or consultation papers we published; it will remember whether we acted to build the infrastructure that will determine who succeeds and who falls in the years ahead. We don’t need to win the frontier model arms race. But we do need our own AI arsenal – or settle for buying compute by the hour from those who did.
Choose quickly. The servers are starting up.
The authors are senior employees at the Tony Blair Institute for Global Change.
Read the orginal article: https://sifted.eu/articles/uk-ai-tony-blair-institute-policy-opinion/
