There are a lot of problems that the data centre industry in the UK has to contend with, whether it’s cooling, planning, power, or more recently public image. But if we break down all those elements, there may be a bigger issue at play – impatience. 

The industry is moving fast to capitalise on the hot new commodity of the moment – AI. Everywhere you look there’s a new AI feature being added to popular apps, or new AI companies launching promising to make people’s everyday lives easier. It’s easy to see why there’s this new wave of AI-in-everything, because that’s where the money is. 

Earlier this year, Gartner estimated that AI spending would top $2.5 trillion in 2026, and with that amount of money floating around – it’s bound to attract a whole swathe of people looking to get their payday. It’s also why you’re seeing more data centre developments than ever before, with new facilities being proposed on what feels like a daily basis. After all, how else are we going to enable AI if we don’t invest heavily in the infrastructure running it? 

That rush to deliver the promise of AI before the money runs out, is one of the reasons the industry is in the spotlight. After all, if you’re proposing to build 140 data centres, which is the number cited by Ofgem as currently in the connections queue, people will start to have questions. And those questions are almost certainly going to focus on power, because we all know AI is power hungry, and we also all know we’re currently living through an energy crisis. 

The queue is being mistaken for real demand

The problem is the debate about the industry’s power usage is being heavily distorted. That’s according to John Booth, DCA Advisory Board Member and Managing Director of Carbon3IT Ltd, who insists that we stop treating speculative projects in the grid connections queue as if they were firm future demand – and maybe, just maybe, take a deep breath and actually plan what we need. 

The distortion comes from the way headline figures are being repeated without enough scrutiny over what they actually represent. This is one of the issues I had with Carbon Brief’s headline, which read as a little on the sensationalist side, but it’s also something Booth has noted with the media’s representation of the industry’s power problem. After all, a project in the connections queue is not the same thing as a live facility, a committed build, or even a scheme with a guaranteed occupier. Yet too often, those distinctions are flattened out in public debate, creating the impression that every project is real, imminent, and destined to become a major new source of demand on the grid.

Booth’s argument is that this is where the conversation starts to go wrong. “The UK data centre operators have a very good handle on data centre construction projects and demands from their hyperscale and global clients, and they are progressing at pace to deliver their requirements,” he says. In other words, the established players aren’t rushing, they’re scaling appropriately to meet demand, but with money to be made in the market, new entrants are coming in who may not have as firm of a grip on realistic demand. The problem is that the wider pipeline is now being viewed as though it all carries the same weight and certainty.

That, according to Booth, is simply not true. “The majority of the ‘new’ projects announced over the past 2 years are property plays, i.e. identify a suitable location, obtain planning and power, and then flip to either an end client or an existing operator or hyperscaler.” 

Now, there’s nothing particularly shocking about that as a commercial strategy. Property speculation exists in every hot market. In fact, I grew up well aware of the property boom happening in Spain in the early 2000s, as investors flocked to build new luxury apartments and homes, hoping to make significant returns. Like those investors found out in the 2008 financial crisis, however, while AI is hot right now, there’s no such thing as a guaranteed return on investment – and that’s why there’s a serious problem when speculative activity gets folded into a wider story about what the country needs to power and build.

Risky speculation shouldn’t shape the narrative

Booth is blunt about the risks. “This is a very risky strategy, as power connections are stretching out to 2037 and new planning rules may require substantial re-design for future AI designs. We also have to believe the AI companies that there will be an actual requirement for this amount of computing power in the future, this is by no means a given.” 

Now, there are many people betting against AI. You just have to take to social media, or even the media in general to see people openly talking about the ‘AI bubble’ and if, or even when, it’s going to pop. While Booth is by no means anti-AI, he does raise an important point. 

The current AI boom has created a rush for position, and in a rush for position plenty of people will try to secure land, power and optionality long before they have secured certainty. And doesn’t that speak to why the debate currently feels so overheated? 

There is a tendency to look at a huge queue number, merge it mentally with the excitement around AI, and conclude that the UK is on the brink of an unavoidable power crunch driven entirely by data centres. But that is a lazy reading of a much more complicated picture. Some demand is real. Some demand is strategic. Some demand is speculative. Some projects will progress. Some will stall. Some will be sold. Some will be redesigned. Some will never make it past the stage of being a good idea on paper backed by the hope that someone richer arrives later.

It’s time to take a deep breah

That’s why for Booth, the answer is not to dismiss the issue, but to slow down and get more serious about what is actually needed – especially when it comes to planning something as complex as the energy grid. “The key point is to remove the speculative projects from the connections queue, take a breath, evaluate exactly what is needed from AI data centres, and build accordingly with a spatial strategy in mind.” 

That last point is especially important. If the UK is serious about AI Growth Zones, serious about supporting strategic digital infrastructure, and serious about avoiding the mistakes of fragmented development, then this cannot just be a race to connect everything, everywhere, all at once. It has to be a question of what should be built, where it should go, and what kind of power system and planning framework is needed to support it.

That in turn brings us back to patience. Not inertia, not delay for delay’s sake, but patience in the sense of discipline. The industry has money chasing AI, developers chasing sites, policymakers chasing growth, and the media chasing dramatic numbers. Under those conditions, it becomes very easy for everyone to talk themselves into a future that looks more settled than it really is. And once that happens, policy starts being shaped not by what is likely, but by what is loudest.

Booth argues that this is already happening. “There appears to be a lot of confusion within DSIT, Ofgem, The AI Energy Council and in the media with regards to current and future data centre energy capacity requirements,” he says. He is equally clear on the consequences of that confusion: “The media speculation surrounding data centre energy use and using flawed information does no one any good and we should wait for a concise plan to be developed by all the stakeholders, which is exactly what is happening right now.”

That is probably the most useful intervention here. The point is not that data centres do not need power, or that AI is not going to reshape infrastructure demand. It is that a speculative queue should not be mistaken for a national blueprint. If the UK wants to have a serious conversation about digital infrastructure, energy security and economic growth, it needs to start by separating what is real from what is aspirational, what is strategic from what is opportunistic, and what is genuinely urgent from what is simply being pushed forward by market impatience.

Because impatience is what sits underneath all of this. The impatience to capture the AI boom. The impatience to secure land and grid access before someone else does. The impatience to turn every large number into a headline. The impatience to build a narrative before a proper plan exists.

And that may be the biggest problem of all.