Britain’s grid queue has a fake-demand problem
The awkward truth at the center of Britain’s power planning is simple: the country’s trying to prepare for an AI buildout, but a chunk of the projects asking for grid access may never become real data centers at all.
These are the phantom schemes. A developer files for a connection, secures a place in the queue, plus presents a project that sounds plausible on paper. The site might be vague. The financing might be thin. The customer demand may be theoretical. In some cases, the whole thing appears to exist mainly as a set of slides, a land option and an application number. That’s enough to put pressure on the system, even if the data hall never gets built.
A paper project can still take a real place in the queue.
This is where the problem stops being a paperwork headache and starts affecting the electricity system itself. Grid operators need to know where new demand’s coming from so they can plan cables, substations, and reinforcement work. If the queue is stuffed with speculative requests, those forecasts get noisy fast. The result is a messier picture of future load, and a slower path for the projects that actually have equipment to install, land under control and financing in hand.
That said, the timing could hardly be worse for anyone trying to do competent power planning. Britain wants to stay in the race for compute, and that means making room for the data centers behind AI services, cloud workloads and the wider tech stack that keeps digital culture humming. The pressure’s real. So is the temptation to file early, claim a promising site, and worry about the rest later. In a market where everyone fears being left behind, a grid application can start to look like a reservation ticket.
That’s part of why the issue has spread across power, land, capital and AI ambition all at once. A suitable plot’s one hurdle, and a grid connection’s another. Money’s another again, and the cost of tying all three together has gone up. A speculative application can become a form of use, when those pieces are scarce. It gives the applicant a story to tell investors, a reason to hold land and a spot in line while the market decides whether the scheme’s serious.
The trouble’s that the queue doesn’t know the difference between a genuine build and a hopeful sketch until much later. By then, grid planners may already have spent time on connection studies, policy teams may have shaped expectations around phantom demand, and real operators may have been told to wait.
That’s the contradiction Britain’s living with right now. AI policy is pushing for faster infrastructure. Developers want into the race for compute. The grid, meanwhile, is being asked to plan around projects that may be little more than wishful thinking with a planning reference number. And the next question is why the line filled up so fast in the first place.
How a cheap queue turned into a speculative stampede
For a while, getting a place in the UK power grid queue looked less like a serious engineering decision and more like a cheap bet. A half-formed plan, or simply a hunch that artificial intelligence was about to soak up every spare megawatt in sight, you could file an application and get in line without putting much money at risk, if you were a developer with a plot of land. Compared with the cost of land, planning lawyers, cabling, substations and the rest of the machinery needed to build a real data centre, the entry fee was small enough to encourage a bit of lottery-ticket thinking.
That structure mattered. When the cost of joining the queue is low and the upside is huge, people do what people always do: they crowd in. By late 2024, the backlog had already started to swell. Over the next several months it ballooned so quickly that total requested capacity moved from roughly 40 gigawatts to well beyond 100 gigawatts. A large slice of that jump came from data centre applications, which alone account for more than 70 gigawatts of requested power, or about one and a half times Britain’s peak electricity demand in 2025.
A cheap queue does exactly what you’d expect: it fills up with hopefuls, opportunists, and anyone who thinks they can sort the details later.
That last part’s where the trouble really starts. Plenty of these applications weren’t clean, shovel-ready projects with land secured and financing lined up. Some were speculative filings made to reserve a spot before anyone else could get there. Worth noting. Others came from firms that had no immediate plan to build but wanted to control a connection point while the AI money was flowing. In practice, the queue became a holding pen for lookalike proposals, many of them dressed in the language of digital ambition but backed by little more than optimism and a spreadsheet.
Then the middlemen arrived. Once grid access itself became a scarce asset, a new kind of trading began around it. Some players filed for power, accumulated a site or a connection chance and then tried to sell the package on at a markup. The mechanics were simple enough: grab the land, secure a queue position, wait for demand to stay hot, and flip the lot to someone richer or more desperate. That’s not development. It’s resale dressed up as infrastructure.
The broader AI boom made the whole thing worse. In 2025, every large technology company seemed to be talking about inference, training clusters, sovereign compute, or some other phrase that translated, on the ground, into a hungry new load on the system. Britain wanted a piece of that action. Fair enough. So did investors, landowners and a long tail of companies that suddenly discovered they were in the data centre business. The UK power grid became a magnet for applications partly because everyone believed someone else would pay for the build later.
This is where power and politics start to tangle. A queue that looks busy can be mistaken for a queue full of genuine demand. A pile of applications can be read as proof that the market’s alive and well. In reality, some of that volume was noise and noisy demand is a bad thing for planners. It pushes network operators to reserve capacity, redraw timelines, and spend time on projects that may never break ground. Real builders then sit behind the paper projects, waiting while the line’s clogged by people who were never serious in the first place.
The government and the regulator have both had to respond to that mess. The state opened a consultation on accelerating electricity network connections for strategic demand, while Ofgem later set out a push to free grid capacity by tackling speculative data centre projects. That reaction says a lot about how far the queue drifted from its original purpose. It had stopped being a simple waiting list and turned into a market of its own, with plenty of people trying to game the rules before anyone bothered to tighten them.
Ofgem’s plan to scare off the ghosts
Once the queue filled up with projects that looked real on paper and a bit less real in the real world, Ofgem stopped pretending the system could sort itself out. In July, the regulator opened a consultation on proposed data centre connection reforms, and it kept the door open for industry comments through September before settling the rules. That timing matters. The UK is still trying to attract AI infrastructure, but it also has to stop every hopeful spreadsheet from acting like a power station waiting to happen.
The basic idea is simple enough, which is probably why it took so long to become policy. If a developer wants a place in the data center queue, it should not be able to grab it with a polite email, a small fee, and a good story about demand. Under the proposed Ofgem data center reform, applicants would have to put real money down upfront, and that money would not come back if the project never reached the finish line. For the biggest schemes, the deposit could climb into the hundreds of millions of pounds. That is a far cry from the old “throw your hat in the ring and see what happens” approach.
A queue that costs almost nothing to join will always attract people who never planned to build anything.
Ofgem’s thinking’s that a serious operator can swallow the risk. A phantom data center can’t. A deposit’s annoying but manageable, if a project really’s customers lined up and a financing plan that can survive contact with a bank. A slide deck and a dream of flipping grid access later, the deposit suddenly looks less like a formality and more like a wall, if it’s just a parcel of land. That’s the point. The regulator wants to sort the builders from the opportunists before the queue eats another year of planning.
The proposal also asks developers to prove demand in advance. “ It also means showing the money’s there to build the thing. That finance test matters because the grid doesn’t need more projects that can win an application and then go hunting for capital after the fact. Britain has already seen enough of that game. The new rules are meant to make sure that a site with a connection request is tied to an actual business plan, not a speculative auction of electricity access.
This isn’t the first time the UK has tried to stiffen the system. In 2024, the rules were tightened so applicants had to show more basic readiness before they got far in the process, including land rights, planning progress and similar milestones. That earlier step was about reducing the number of obviously unserious filings. The current round goes further. It treats the queue less like a public waiting room and more like a gate with someone checking tickets, which, frankly, is overdue.
The logic behind all this is not hard to follow if you look at the numbers. Government data on data centre electricity consumption in Great Britain, 2020 to 2024 shows demand was already moving in a direction that will make planners sweat. When consumption rises and the application pile keeps growing even faster, the grid can end up planning for buildings that may never appear. That is how phantom data centers become a policy headache rather than a private embarrassment.
For now, the regulator’s betting that a higher price of entry will clear out the ghosts without scaring off the builders who can actually pour concrete, order kit, and plug in customers. Whether that balance holds is another question entirely, and the answer will shape what happens next in the queue.
Why the cure could hurt the patient
The logic behind Ofgem’s fix is tidy enough. If a developer can sit in the grid queue on a shoestring, then the queue fills with projects that may never get beyond PowerPoint. So the regulator wants to make waiting expensive. That should flush out the paper data centers, the land-flipping middlemen, and the applicants who only wanted a place in line.
The snag is that a fee high enough to scare off ghosts can also make real builders hesitate. At some point, the charge stops acting like a filter and starts acting like a toll booth. If the deposit gets large enough, a serious project has to tie up a lot of cash before it’s planning permission, a signed customer, or even a final site package. That’s not a trivial ask, especially in a market where financing costs are already touchy and timelines can stretch for years.
A deposit that scares off spreadsheet projects can also scare off the people who actually own the shovels.
That’s why Britain isn’t starting from a position of strength here. Power is pricier than in much of the US, land is tighter and the patchwork of available sites is nowhere near as roomy as what you can find in parts of mainland Europe. That matters a lot in AI infrastructure UK planning, because data centers aren’t fancy software. They need acres, substations, cooling, fibre and a utility relationship that doesn’t look as if it were assembled in a hurry. Even when a site exists on paper, the practical problems can be enough to make investors look elsewhere.
The government knows speculative demand’s become a problem, which is why it’s already moved to tighten grid access requests. Its own announcement about tackling speculative demand in the connection queue lays out the basic case for action, but the politics are awkward. It has to look serious without looking hostile, if Britain wants to keep attracting compute. That’s a narrow path to walk, and it gets narrower when every new rule is read by investors as another cost, another delay, another lawyer’s bill.
There’s also a size issue. Heavy upfront deposits may suit giant operators with deep balance sheets, internal legal teams, and enough land already under option to absorb a miss. Smaller entrants, especially AI-focused firms that are still building their first large UK campuses, may not have that luxury. They can have decent financing, good customers and real engineering plans, yet still struggle to post a giant sum before they know whether the grid will cooperate. In practice, that can tilt the field toward the incumbents.
That risk isn’t theoretical. Newer names such as Nscale have said they want to spend heavily in the UK, and that kind of commitment’s exactly what ministers and grid planners say they want. But a tougher regime could ask these firms to front-load too much uncertainty. If the rules become too unforgiving, the result might be fewer speculative applicants, yes, but also fewer plausible newcomers willing to test the market. And once that happens, the queue gets cleaner while the actual pipeline thins.
The numbers matter here in a very mundane way. The ONS has already had to explain how data centres fit into the UK national accounts, which tells you this is no longer a niche corner of the economy but a live line item that needs proper handling. The official note on datacentres and the UK national accounts is a reminder that policy choices here don’t just affect one sector’s mood. They affect investment, output, land use, and the broader case for building in Britain at all.
So the regulator’s job is less about being tough for its own sake and more about not overcorrecting. The queue stays clogged, if the deposit’s too low. The UK becomes the place where everyone admires the ambition and then files somewhere cheaper, if it’s too high. That’s the awkward bit of grid capacity planning and AI policy in the same room. The country needs fewer phantom projects, but it can’t afford to tax the real ones out of existence.
A cleaner queue won’t fix everything overnight
If the UK can strip out the phantom projects, grid planners will finally get something they’ve been missing: a queue that looks like demand, not wishful thinking. That sounds dull on paper, and in practice, it changes the maths. When applications are inflated by sites that may never be built, network operators can end up planning upgrades for power that never turns up, while the places with real demand wait in line behind a pile of slide-deck fantasy.
A cleaner queue won’t create new electricity overnight, but it can stop the grid from budgeting for buildings that exist mainly in pitch decks.
That matters because grid reinforcement isn’t a casual weekend job. New substations, upgraded lines, transformer work and land acquisition all take time, money and a fair amount of bureaucracy. If planners know which data centre proposals have actual customers, financing and land rights behind them, they can put resources where they’re needed instead of spreading them thin across speculative requests. In plain English: fewer blind bets, fewer expensive detours.
The benefit isn’t just cleaner paperwork. It also changes where the money goes. A queue full of fake demand can push operators to widen capacity in the wrong places, or to accelerate schemes that end up serving nobody. Once that noise is cut away, the remaining requests should give a much better read on where Britain needs stronger connections, where a substation needs expansion and where a local network can probably cope for now. That kind of sorting doesn’t make the grid magically roomy, but it does make planning a lot less bonkers.
The reform also works on a longer clock than the current AI boom cycle. Britain isn’t going to wake up one Monday morning with a surplus of spare capacity just because Ofgem tightened the rules. These changes are about the next several years, not the next several months. If the data center deposits and other upfront checks scare off paper projects, the payoff comes later, when the queue is no longer jammed with names that never intended to build at all.
That’s the slightly unglamorous part of the story. The UK can still compete for compute, and it probably has to if it wants a meaningful slice of the AI buildout. But it won’t get there by reserving scarce grid space for sites that only exist on diagrams. Britain’s best shot is a queue that reflects real finance, real land and real demand, so the grid can plan for what’s coming instead of what somebody hoped to sell.
