Imogen BennettMicrosoft has pulled out of its agreement to anchor Nscale's planned data centre in Loughton, Essex, the Telegraph reported on 18 October 2026. The move lands just as the British AI infrastructure company prepares an initial public offering that could raise up to $3 billion. For private investors weighing a place in that offering, the story is a live lesson in concentration risk.
What happened and what is confirmed
According to the Telegraph, as relayed by Reuters and Bloomberg, Microsoft recently terminated its contract for the Loughton site. The newspaper gave no reason for the withdrawal. Microsoft had not commented publicly at the time of writing.
The project has faced a well-documented problem: power. Authorities have reportedly told Nscale they do not expect a grid connection to the site until the 2030s. That is a long way from the timetable set out in January 2025, when the plan was announced as part of a £2 billion UK investment.
The original design was ambitious. The site was meant to deploy about 23,000 Nvidia GB300 chips and offer 50 megawatts of capacity at first, scaling to 90 megawatts. Nscale says it remains committed to the location and is exploring alternative power sources. A spokesperson added that its relationship with Microsoft "remains strong and long-term".
The numbers behind the headline
Several figures explain why the news matters more than a single cancelled contract.
- 85%: the share of Nscale's contracts that comes from just two customers, Microsoft and Anthropic, according to its IPO filing.
- $3 billion: the maximum the company is reportedly seeking to raise, as soon as this month.
- 6.4 MW of 30 MW: the capacity currently running at its flagship Norway data centre, which is also delayed.
- 2030s: the earliest date authorities have reportedly suggested for power at Loughton.
When two customers account for roughly 85% of contracts, losing or shrinking one relationship is not a rounding error. It changes the revenue story on which an IPO valuation rests.
Why power, not demand, is the bottleneck
The Loughton episode shows a pattern across the sector. Demand for AI computing is strong, but physical capacity depends on grid connections, planning permission and energy supply. Those move on timescales of years, not quarters.
For a company building data centres, a delayed grid connection means paying for land, chips and staff while earning nothing. Contracts signed before power is secured can be cancelled, as this case appears to show. Investors should therefore read a data centre developer's "pipeline" as a list of intentions, not revenue.
What an investor should ask: the wealth management view
A professional adviser looking at a new listing typically starts with four questions. How concentrated is revenue? How much of the capacity is actually energised? How much debt funds construction? And what happens to the valuation if the largest customer leaves?
None of this means Nscale is a poor business. It means the price paid for a share should reflect the risk of delay. IPOs are also priced to sell, and early trading can be volatile in either direction.
Retail investors often meet IPOs through headlines and platform notifications, not through the prospectus. The risk factors section of the filing, where an 85% customer concentration would be disclosed, is where a decision is usually won or lost.
A worked example: a £20,000 ISA and one hot listing
Take a hypothetical investor, call her Priya, who holds £20,000 in a stocks and shares ISA spread across global index funds. After reading the Nscale headlines, she is tempted to put 25% of the pot, £5,000, into the offering on the first day of trading.
Here is the if/then logic an adviser would walk through with her.
If the shares fall 30% in the weeks after listing, a plausible swing for a newly listed, capital-hungry company, her £5,000 becomes £3,500. That is a £1,500 loss, or 7.5% of her entire ISA, from one decision.
If instead she limits the position to 5%, or £1,000, the same 30% fall costs £300, or 1.5% of the ISA. She keeps exposure to the upside without letting one grid-connection delay define her year.
There is a tax angle too. Losses inside an ISA cannot be offset against gains elsewhere, so a bad bet cannot be used to reduce a tax bill. An adviser would also check that her emergency savings, ideally three to six months of expenses, sit outside the market before any speculative position is added.
The point of the example is not that Priya should avoid the listing. It is that position size, not stock-picking, decides how much a surprise like the Microsoft withdrawal can hurt.
Reading the signals before an IPO
A few practical signals separate a well-prepared listing from a rushed one. Look at how much of the stated capacity is contracted, and with whom. A contract with a creditworthy customer is valuable, but only if the customer cannot walk away cheaply. Termination clauses matter, and prospectuses usually describe them.
Next, check whether the company has secured power for each site, not merely land. A grid connection date in the 2030s, as reported for Loughton, turns a growth story into a waiting game. Finally, look at who is selling. If existing shareholders are cashing out heavily at the offering, ask why.
None of these checks needs specialist software. They need time, a calm head and, for larger sums, a second opinion from someone who reads prospectuses for a living.
How the UK's AI ambitions are affected
The wider story is about Britain's plan to become a hub for AI computing. Government and industry have repeatedly cited large data centre projects as proof of progress. A flagship site waiting until the 2030s for power complicates that narrative.
Nscale says it is looking at other sites and alternative power. Whether those can be built quickly enough to satisfy the remaining anchor customers is the real question for the company's first years as a listed business.
For households, the effect is indirect but real. Pension funds and index trackers hold technology shares, so swings in AI infrastructure sentiment can reach ordinary savers who never buy an IPO directly.
What to do now
If you are considering the offering, or already hold AI-heavy funds, a short checklist helps.
- Read the risk factors in the prospectus, in particular customer concentration and power supply.
- Work out your current exposure to AI and data centre themes across all funds, not just individual shares.
- Decide a maximum position size before the first trade, and write it down.
- Keep emergency savings outside the market.
- Check that you are using regulated firms. The Financial Conduct Authority publishes guidance for consumers on its consumer pages, including how to check whether a firm is authorised.
A qualified wealth manager on Expert Zoom can review your portfolio's overlap, set sensible position limits and explain how the tax wrapper you hold affects your options. A short consultation before a high-profile listing costs far less than a single avoidable loss.
This article is for general information and is not personal financial advice. The value of investments can fall as well as rise, and you may get back less than you invest. Consider speaking to a regulated adviser before making investment decisions.
