Pulsant's Mark Lewis on Data Sovereignty & AI Infrastructure

According to Pulsant's CMO Mark Lewis, much of the UK's enterprise AI future will be built in regional data centres.

Speaking at Data Centre LIVE, he argued that data sovereignty and low-latency connectivity are reshaping where organisations choose to host their infrastructure.

After surveying UK IT leaders, Pulsant found that while cybersecurity remains organisations' biggest concern, data sovereignty has quickly become another major priority.

"They wanted to know where their data is. What infrastructure is their data on? How can I make sure I know where it is, and that it's not just disappearing onto the internet?" said Mark Lewis, Chief Marketing Officer at Pulsant.

And how can I make sure that the right people have access to it at the right time, and not anybody else?" 

Mark believes that businesses should pay greater attention to the infrastructure beneath them.

He explained that organisations often overlook where workloads physically run and whether that infrastructure itself needs to meet sovereignty requirements.

Looking beyond London

While London is the UK's dominant data centre market, Mark said many businesses are better served elsewhere.

Around 80% of UK data centres are concentrated in Slough and London's Docklands, yet enterprises across cities such as Edinburgh, Newcastle, Birmingham and Manchester want their infrastructure closer to home.

Some organisations also require geographic separation for resilience and regulatory reasons, with disaster recovery sites deliberately located away from London.

For customers using Pulsant's private network, regional deployments also provide flexibility to expand nationally without being tied to a single location.

AI beyond the gigafactory

The rapid growth of AI has fuelled demand for large, power-hungry facilities, but Mark believes that enterprises shouldn't assume they need hyperscale infrastructure.

"Not all data centres are the same," he said.

While enormous AI campuses are essential for training frontier models, enterprise customers typically have different requirements.

"Enterprises don't necessarily need that density of compute, that amount of power. In fact, in most cases, we find enterprises will happily take 40kW or below for a rack."

Mark added that organisations can already deploy AI hardware in conventional enterprise facilities.

"You can still run NVIDIA H100s in that environment, which means you can add AI to your existing infrastructure, your existing estate, and do all of that in a regular data centre. You don't need a massive gigafactory to do that."

Instead, Pulsant sees growing demand for inference, where trained models are put into practical use through enterprise applications.

"We mostly work on inference," Mark explained. "It does need to be close to populations, so you need low latency to populations and you need the ability to expand."

Critical infrastructure, critical decisions

Mark also welcomed the UK's decision to classify data centres as critical national infrastructure, saying it reflects the sector's central role in everyday life and the wider economy.

However, he believes the designation also reinforces the importance of risk management when designing digital infrastructure strategies.

Looking ahead, Pulsant expects inference workloads to become a far larger market than foundation model development itself.

Using a railway analogy, Mark described large language models as "the locomotives at the front of the train", while inference represents the many carriages behind them.

As businesses build AI-enabled applications, Pulsant is positioning its regional data centre network to support organisations as those workloads continue to grow over the coming decades.