At London Tech Week recently, I was lucky enough to spend some time with Simon McIntosh-Smith, Director of the Bristol Centre for Supercomputing (BriCS), to discuss Isambard-AI, the most powerful supercomputer in the UK and ranked 11th fastest globally.

It was a fascinating conversation that’s really stuck with me because beneath the technical detail, several themes kept surfacing including access, sovereignty and the UK’s long-term position in the global AI race.

For all the hype surrounding artificial intelligence, much of the public conversation still focuses on applications, smarter tools, faster insights, transformative business models.
For me, beneath all that sits a far more fundamental question, one that the UK is only just beginning to grapple with; where does the infrastructure powering AI actually live, and who controls it?

The answer increasingly matters and initiatives like Bristol’s Isambard supercomputing programme are showing us why.

From experimentation to national asset

Built out of years of research at the University of Bristol, Isambard didn’t start as a flagship national asset. It began as a relatively modest experiment, an attempt to explore alternatives to the dominant computing architectures underpinning most systems.

Early iterations focused on ARM-based processors, a departure from the traditional Intel and AMD ecosystem. At the time, it was a niche bet.

That bet has paid off.

From its first iteration in 2016 through to its current form, Isambard-AI, the programme has evolved into a serious piece of national infrastructure. What makes the latest system notable isn’t just its sheer computing power, but its purpose. It has been explicitly designed to support AI workloads at scale, while remaining accessible to UK researchers, startups and organisations that would otherwise struggle to compete for resource.

The compute gap holding innovation back

Demand for large-scale AI compute has surged globally, to the point where even well-funded organisations struggle to secure capacity. Running models across hundreds or thousands of GPUs isn’t simply expensive, it’s often impractical in the commercial cloud, where availability is constrained and the pricing fluctuates.

By contrast, systems like Isambard offer a different model. Through government-backed allocation, companies, including early-stage startups, can access significant compute resource quickly, in some cases moving from application to live experimentation in a matter of days. For fast-moving businesses, that kind of turnaround can be the difference between proving a concept and falling behind.

This is where the economic argument becomes interesting. The majority of commercial users tapping into systems like Isambard are not large enterprises, they are SMEs and startups. These are precisely the organisations most likely to innovate rapidly, but also those least able to compete in a purely commercial infrastructure market.
Providing them with access to high-performance computing isn’t just a technical decision, it’s an industrial strategy.

Why sovereignty has moved to centre stage

Beyond economics though, there’s a more pressing issue emerging, one that has shifted dramatically in the past two years, and that’s sovereignty.

AI sovereignty and data sovereignty have gone from niche policy concerns to central strategic priorities. Organisations are increasingly asking not just whether their data is secure, but where it physically resides, who has jurisdiction over it and what level of control they truly have.

Public cloud has delivered extraordinary innovation, but it has also introduced new layers of opaqueness. Data may be stored across multiple regions, subject to varying legal frameworks, and critically potentially accessible under foreign government mandates. Even with strong contractual protections, there is a growing recognition that ultimate control may sit elsewhere. That uncertainty is driving renewed interest in domestically hosted infrastructure.

A case for sovereign AI capability

The case for sovereign AI capability is straightforward. If systems are physically located within the UK and controlled locally, organisations gain a far higher degree of confidence over how their data is handled. They know where it is.

They know who can access it and crucially, they can apply governance frameworks that align with national regulation and strategic priorities.

The reality is that AI will require a hybrid infrastructure model, one that combines the flexibility of cloud with the assurance of sovereign capability. The risk for the UK is that, without investment in the latter, it becomes entirely dependent on external providers.

That dependency brings both economic and strategic downsides. It limits the ability of domestic companies to compete at scale, it exposes sensitive datasets to cross-border complexity and it risks placing key innovation pathways outside national control.

Building beyond London

Regional centres have a critical role to play when it comes to deep tech, infrastructure and research. Bristol is certainly not London, yet it has become home to one of the UK’s most significant computing assets, underpinned by a highly specialised team and a strong track record in advanced computing.

From curiosity to capability

And perhaps what’s most exciting, is the level of access available to those who need it.

Researchers, startups and businesses can apply for compute time through government-backed routes, often receiving decisions in days or weeks rather than months. Once approved, users can be up and running quickly, experimenting, building and testing without many of the typical barriers often associated with high-performance computing.

Those who are simply curious, whether they be organisations exploring AI, or individuals considering a career in the field, can follow Isambard’s developments,
explore use cases and connect with the ecosystem through lots of different channels, from online updates to short-form project showcases.

You could tell that level of openness was important to Simon when I spoke to him.
He understands if the UK is serious about building a competitive AI economy, access cannot be limited to a narrow group of insiders. It needs to be visible, discoverable and actively inviting participation.

Momentum now matters

Systems like Isambard-AI are still in their relative infancy. The technology continues to evolve, with ongoing investment in software stacks, storage capacity and user
accessibility.

The ambition is obvious; to support a growing and increasingly diverse set of users, from academic researchers to commercial innovators. But maintaining that trajectory will require sustained commitment from UK government, from industry, and from the broader ecosystem.

The critical importance of AI infrastructure isn’t up for debate. The real question is whether the UK can truly own it.
With the right support, Isambard AI is living proof it can.

Written by Greg Halse
Associate Director, Ambitious PR

Greg Halse