The UK wants to be an AI superpower, but last week it was reminded that AI runs on electricity. Microsoft has pulled out of Nscale’s planned flagship AI data centre in Essex after it emerged the site would not get the grid power it needs for years. Here is what happened, why it matters, and what UK businesses should do about it.

What happened

Microsoft has ended its agreement to take computing capacity at Nscale’s planned AI data centre in Loughton, Essex, which was set to be the largest in the UK. Reuters reported the termination on 18 October 2026. According to reporting by the Guardian and others, the local network operator, UK Power Networks, indicated that adequate grid power would not reach the site until the early to mid-2030s, long after its planned opening in early 2027.

Nscale announced the project in 2025 as a 50MW campus, expandable to 90MW, with 23,040 Nvidia GB300 chips earmarked for Microsoft’s Azure cloud. It was part of Microsoft’s £22 billion UK investment and was hailed at the time as a boost to Britain’s AI ambitions.

Both companies have stressed that the wider relationship is intact. Microsoft described the project as a relatively small piece of its portfolio and said it remained on track to meet its UK commitments. Nscale said it remained committed to the site and the UK, and is exploring alternative power sources.

The real story: why the UK cannot power every AI data centre

Loughton is not a one-off AI data centre problem. It is the most visible example of a problem the energy regulator has been warning about for more than a year. In July 2026, Ofgem reported that contracted demand connection offers had jumped from 41GW to 125GW between November 2024 and June 2025, driven largely by data centre projects. Around 73GW of data centre demand was referenced in its consultation on reform.

Much of that queue is speculative. Projects can secure a place in line at very little cost, which blocks capacity for schemes that are actually ready to build. Ofgem has proposed a Data Centre Commitment Fee and new progress milestones that developers would have to meet to keep their place. Its consultation closed in September and a decision is awaited.

The government is also trying to steer AI data centre capacity towards designated AI Growth Zones, with plans to reserve and reallocate grid capacity for strategically important projects and a Connections Accelerator Service to help developers. The Loughton decision shows how much is riding on those reforms working, and working quickly.

Why this matters to businesses that will never build an AI data centre

It is tempting to see this as a story for energy companies and hyperscalers. It is not. Every UK organisation planning to use AI, or simply to keep growing its cloud footprint, depends on the same infrastructure.

• Capacity may lag demand. If new UK AI data centre capacity arrives later than planned, demand for the most powerful AI services in UK regions could outstrip supply, with knock-on effects for availability and price.
• Data residency choices matter more. Organisations that need to keep data in the UK for regulatory or contractual reasons have fewer options than those that can use EU or global regions.
• Energy costs are already a board-level concern. ONS data shows 72% of UK businesses reported concern about energy prices in late September 2026, the highest level since the question was introduced. Energy-hungry AI workloads will feel that pressure too.
• Resilience and capacity are linked. Last week’s Google Cloud outage in London showed the risk of relying on one region. Constrained UK capacity makes planning for alternatives harder, and more important.

5 practical steps for your AI and cloud strategy

1. Right-size your AI ambitions

Not every task needs the largest model or the most powerful chips. Smaller, efficient models can handle many business tasks such as summarising, classifying and drafting at a fraction of the compute cost. Match the tool to the job.

2. Decide where your data must live

Be clear about which data genuinely has to stay in the UK and which can be processed elsewhere. That decision shapes which cloud regions and AI services you can use, and how much flexibility you keep if UK capacity is tight.

3. Keep a hybrid option open

For predictable or sensitive workloads, on-premise or private infrastructure can offer cost certainty and control, while the public cloud handles peaks and experimentation. A hybrid design also gives you somewhere to run critical services if a single cloud region fails.

4. Ask your providers about capacity, not just features

When you sign or renew cloud and AI contracts, ask which AI data centre or region behind your service physically sits in, what happens if demand outstrips supply, and what service levels apply. Contracted capacity is not the same as capacity that is powered up and running.

5. Measure the energy and cost of your workloads

Track how much compute your AI and cloud services consume, and what they cost. Efficiency is no longer only a sustainability goal. In a power-constrained market, it is a competitive advantage.

The bottom line

The UK’s AI ambitions now depend as much on cables, substations and connection queues as on chips and talent. Businesses cannot fix the grid, but they can plan around it. The organisations that design efficient, flexible and resilient AI and cloud strategies today will be best placed to grow, whatever happens to the next headline AI data centre.

Build an AI and cloud strategy that does not depend on one headline project

Cloud Technology Hub designs hybrid and multi-cloud infrastructure across Azure, AWS and on-premise environments, with on-premise options through our partner HPE. Email info@technohub.cloud to talk through a strategy that fits your workloads, budget and data residency needs.

Speak to our team at info@technohub.cloud. TALK

Leave a comment and Read More Here