04 July 2025

Peter Wilcock, VP, Latos Data Centres

At Nvidia’s GTC developer conference in March, CEO Jensen Huang heralded a trillion-dollar boom in data centres, to underpin ever-more sophisticated AIs. He declared that “the data centre is no longer a warehouse for computing, it is the engine of AI.”

This transformation is already underway. Since 2021, sales of conventional CPUs have slumped by 80%. By contrast, demand for the GPUs needed for AI is accelerating, and expected to grow by almost 30% every year.

As AI moves from experimental to essential, the traditional architecture of our data centres — concentrated, centralised, and physically distant — is no longer fit for purpose.

For the UK to reap the full benefits of AI, we need to rethink the country’s data centre map.

The real-time AI revolution

Up to now, hyperscale data centres are designed to retrieve data – files, software processes, and so on. The “engines of AI” Jensen Huang envisions are designed from the ground up to deliver the processing power to interpret, learn from, and respond to new information on the fly – and at massive scale.

This revolution in real-time AI is set to transform how we live and work. But sheer computing heft isn’t the whole answer. Where that computing is located is going to be increasingly important.

Take the metaverse, gaining ground in home entertainment and industrial training environments alike. Latency is a make-or-break consideration: delays longer than 20 milliseconds or so can cause disorientation and motion sickness for users. That is the maximum delay tolerable for AI-powered content, facial tracking, or environment rendering.

Transportation is another example. Autonomous vehicles depend on decisions made in microseconds to ensure passenger and pedestrian safety. A car that needs to ping a distant data centre won’t respond quickly enough to a sudden hazard. The slightest delay could be a matter of life and death.

In predictive healthcare, AI systems analyse patient data to provide early warnings and enable tailored interventions. There have been many encouraging advances in the use of AI to identify the early stages of cancer, Alzheimer’s disease, COPD, and other conditions. Not only must predictive healthcare AIs be clinically accurate, they must also deal with a wide variety of data – images, videos, and written records – with the utmost security.

In other words, extent to which an AI can be valuable, responsive, and safe, will be increasingly measured in kilometres – from the end-user to the data centre.
Delivering on its promise depends on complementing existing hyperscale infrastructure with new facilities, designed for AI, located much closer to end users.

A new generation of edge

A new generation of smaller scale ‘neural edge’ data centres are coming onto the market, focused on meeting precisely this need.

What sets these facilities apart is their energy-dense design – needed to support the most demanding AI training and inference workloads.

But performance specs are only part of the story. Neural edge data centres can be built quickly and integrated unobtrusively into urban environments. Modular building techniques and sustainable power and cooling mean facilities can start small and grow in line with demand, without compromising their overall performance.

By placing compute power close to end users, in a tech cluster, a manufacturing hub, or even a residential area, neural edge facilities can dramatically reduce AI latency while improving overall resilience. And they are likely to prove more energy efficient in the long run.

The UK’s AI moment

The UK’s ambition to lead in AI is not just about innovation policy. It’s about building the physical foundations for an AI-driven economy – expected to add almost £50 billion to the economy every year. The government’s promotion of AI Growth Zones — regions earmarked for accelerated investment in AI infrastructure — reflects an understanding that AI must be supported regionally, not just centrally.

If that adoption is to be inclusive, scalable, and sustainable, it must be backed by infrastructure that ensures real-time AI is available where it’s needed.
It’s easy to get swept up in the promise of AI — the breakthroughs in healthcare, the efficiencies in transport, the transformations in everyday life. But without the right infrastructure, those breakthroughs risk stalling. Neural edge data centres offer a practical, scalable way to deliver real-time AI where businesses, public services, and homes need it.

If AI is the brain of the digital economy, the neural edge is its nervous system. Fast, distributed, and always on, neural edge data centres will enable the next decade of UK innovation — not in the cloud, but from the ground up.