Scotland’s AI showcase now has a power problem
Lanarkshire was supposed to be one of those rare tech projects that lets politicians, developers and investors all nod in the same direction without squinting too hard. A huge new AI campus on Scottish soil, a clean-energy story baked into the pitch, and the promise that Britain could build serious digital infrastructure without pretending electricity is an afterthought. Neat, tidy, very on-brand for a government that likes to talk up AI as if it were already a public utility.
That was the sales job in January, when CoreWeave and DataVita unveiled plans for a multibillion-pound complex near Motherwell. The pitch was dressed up as jobs, growth and a place for Scotland to muscle in on Britain’s wider AI ambitions. Data centres usually arrive in public with the charisma of a filing cabinet, so the companies went hard on the upside. More investment. More construction work. More technical jobs. More computing power. Fewer excuses for why the UK keeps renting its digital future from somewhere else.
The electricity plan sat right in the middle of that pitch, not off to the side in a footnote nobody read. That matters. A data centre can promise offices, contractors and grand opening photos all it likes, but if the power story doesn’t hold, the rest gets wobbly fast. In this case, the claim was that the site would run on clean energy through a private-wire setup, with onsite or closely linked renewables doing the heavy lifting. For a project of this scale, that was doing a lot of work. It was also the part that made the whole thing sound palatable to ministers who want AI policy without a side order of fossil-fuel awkwardness.
When a project sells itself on green power, the electricity plan is not a nice extra. It’s the whole argument.
Fresh documents now suggest that argument may be thinner than advertised. Public-record checks and internal material point to a gap between the glossy January version and what can actually be shown on the ground. The clean-power pledge is supposed to be in place by the early 2030s. On paper, that gives everyone time. In practice, time is exactly what these projects burn through fastest.
That is where the mood changes from upbeat launch event to slightly embarrassing spreadsheet review. A site can talk about self-supplied renewables all day long, but unless land, planning, grid access and construction are moving in step, the promise starts to look more like a date circled in pencil than a firm delivery plan. Here, the new material raises the awkward question: is the renewable supply genuinely being built, or is it still largely a future tense story?
For tech news readers, this is the sort of detail that decides whether an AI project is real infrastructure or just a very expensive press release with nicer typography. For AI policy watchers, it is even sharper. Scotland has been sold a growth story, but growth stories have to eat electricity somewhere. And for power and politics, the answer to “where does the juice come from?” tends to arrive sooner than the ribbon-cutting speeches would like.
There’s also a broader cultural itch here. Digital culture keeps getting told that every new AI campus is a clean, modern leap forward. Then the power bill shows up, and the room gets quieter. That’s the moment when the slogans stop mattering and the actual machinery, permits and timelines begin running the meeting. The Lanarkshire project was framed as a rare case where all the upbeat parts lined up. The fresh evidence suggests the wiring may not be nearly as settled as the launch suggested.
What happens next will shape the next round of scrutiny. If the site can prove the renewable plan is real, the companies will have a tidy rebuttal. If they can’t, the green pledge starts to look less like an operating model and more like a promise that wandered too far ahead of the hardware.
What developers promised versus what they can show
When ministers named Lanarkshire as an AI growth zone, the pitch sounded neatly packaged for both the tech crowd and the green-energy crowd at once. The government’s own announcement talked up jobs, local spending and a fast track for a major AI build-out in Scotland, while North Lanarkshire Council welcomed plans for what it called one of the world’s most advanced AI sites, all in the same breath as the usual industrial optimism and economic boosterism you’d expect from a shiny project launch. The renewable angle sat right in the middle of that story, not as a decorative extra but as part of the sell: clean power, modern infrastructure, serious scale. The official growth-zone announcement tied the designation to growth and jobs, while local leaders leaned into the same message on North Lanarkshire Council’s page about the plans.
That clean-power promise was described in very specific terms. The site was supposed to run on renewable electricity generated on or near the project itself, with the full build-out ready by the end of the decade. The number attached to that plan was eye-catching: roughly one gigawatt of renewable supply, split across solar and wind. Put plainly, that is a huge amount of power. It would be larger than the UK’s biggest onshore wind project, and it would be enough to supply electricity for a very large chunk of Scottish households if it were flowing into homes rather than into server racks and cooling systems.
Big promises sound tidy on launch day. The only problem is that electricity has to exist somewhere other than the presentation slide.
The scale matters because the proposal was not framed as a vague future aspiration. It was sold as a working energy model for the AI datacentre Scotland wants to host, with a private-wire setup doing the heavy lifting. That means power would be generated and delivered directly to the site, rather than simply bought from the national grid in the ordinary way. In theory, that arrangement can make a data centre look much cleaner on paper. In practice, it depends on the turbines, panels, cables and land actually being there.
That is where the public record gets awkward. The project company’s live footprint is tiny by comparison. DataVita currently operates just two existing data centres, and those facilities draw a relatively modest amount from the grid. That is a normal enough starting point for a datacentre operator. It is not, though, the same thing as having a gigawatt-scale private renewable system already built, commissioned and feeding a giant new AI campus in Lanarkshire. The leap between those two states is vast. So far, the public can see the small footprint. The large-energy backbone remains mostly a promise.
The gap between claim and visible reality is made sharper by the kind of language used around the project. In January, the development was presented as a multibillion-pound venture led by CoreWeave and DataVita, with the sort of confident tone that tends to accompany big-tech arrivals. That sort of announcement naturally invites the usual shiny headlines. It also creates a fair expectation that the supporting infrastructure is at least under construction, if not already partly in place. Yet there is no public evidence that the full private-wire renewable system is already operating, or even close enough to be treated as a done deal.
That matters because the renewable energy pledge was doing a lot of work from the start. In the Scottish policy world, clean power is not a side quest. It has been central to how ministers talk about digital investment, industrial renewal and net zero planning, and the Scottish government has repeatedly framed the country’s low-carbon economy as a place where large projects can grow without leaning on dirty electricity. The government’s net zero economy material sits in that same policy lane. So when a major AI site arrives with its own green-energy story attached, the claim lands with extra force. It needs more than a nice brochure and a few happy quotes.
What makes this case a little slippery is that the public rhetoric and the operational facts seem to live in different rooms. On one side, there is the pitch for a near-future campus powered by renewable generation built specially for the site. On the other, there are two existing data centres and a project footprint that, at least in public view, has not yet produced evidence of the vast solar-and-wind system it promised. That does not prove failure by itself. Projects can move in stages, and large infrastructure is rarely built in a neat straight line. But it does mean the burden of proof has shifted. If the energy system is real, people should be able to see something more concrete than broad assurances.
For now, the headline promise remains that this Lanarkshire AI growth zone will run on its own green electricity and do so at a scale most sites could only dream about. The trouble is that the dream has to survive contact with planning papers, cables, land, turbines and time. And time, as the next part of this story shows, is where the arithmetic starts to bite.
Why the maths does not work
Once you stop reading the pitch deck and start counting acres, the clean-power story gets awkward fast. A datacentre cluster of this size would need a genuinely large renewable build-out, the sort that eats up tens of square kilometres when you add solar arrays, wind turbines, cabling, access roads and the rest of the unglamorous bits. The filed planning material, by contrast, appears to cover only a sliver of that footprint. That gap matters. A lot. It is the difference between a plausible supply plan and a nice idea with a hard hat on.
DataVita’s own Lanarkshire AI growth zone page keeps the emphasis on ambition, but ambition doesn’t pour concrete or secure land rights. The same goes for CoreWeave’s part in the Scottish AI project: the public story is about speed and scale, while the paperwork tells a much narrower tale. One related parent-company proposal seems to deal with only a handful of turbines. Helpful? Sure. Enough to cover a build like this? Not remotely, unless the electricity demand turns out to be doing some very creative yoga.
Clean power plans fail first in the spreadsheet, long before anyone notices the missing turbines.
That is where the land question becomes unforgiving. A handful of turbines can feed a fraction of the load, but the original promise sounded much bigger, closer to a private grid built around the site rather than a few token assets tucked somewhere nearby. If the renewable element is meant to cover most, or all, of the campus, then the acreage has to be there. So does the planning consent. So does the time. And time, annoyingly for everyone involved, is not something you can fast-track with a press release.
The grid problem is even less forgiving. Britain’s connection queues are already long, and Scotland’s network is congested enough that new supply can sit in limbo for years. That is fine if you are planning a modest industrial unit or a warehouse. It is a much bigger headache when the entire selling point is datacentre power at a scale that depends on fresh infrastructure arriving on cue. The Scottish government’s own AI strategy says growth-zone applicants were supposed to show a realistic route to power, either through an existing grid connection or an independent energy solution. That sounds sensible because it is. Electricity has a habit of making fools of anyone who treats it like a branding exercise.
The private record makes the timing problem look worse. Officials appear to have known there was a power-provision issue before, or around, designation of the site. That is the kind of detail that changes the mood in a room. If the state is already aware that the energy plan is shaky, then the project is no longer a straightforward green industrial story. It becomes a question of whether people are hoping the system will magically catch up later. Hope is not a grid strategy, though it does seem to turn up in meetings quite often.
There was also a brief flirtation with a gas-powered fallback in internal discussions. That bit is messy for obvious reasons. The developer says fossil fuels will not be used, which is a cleaner line to give and, frankly, a much less embarrassing one to have on a website. Still, the mere fact that gas entered the conversation tells you something about how tight the power arithmetic had become. When the first plan is so difficult that someone starts reaching for a burner, the original promise is already wobbling.
Scotland’s own green data centres campaign makes a decent backdrop to all this because it treats low-carbon server capacity as a power-supply problem first and a glossy project second. That is the right order. The cables matter. The substations matter. The turbines matter. Without them, “green” is just a word floating above the car park.
What seems to be happening now is more cautious, and more convenient. Some participants are apparently betting that a grid connection might arrive in the latter part of the decade, with cleaner power following after that. That is a very different proposition from the original pitch. The original pitch implied a self-contained renewable setup that could support the campus by the early 2030s. The newer version sounds more like: build first, wait for the network to catch up, then declare victory once the electrons finally behave.
That distinction may sound nerdy, but it is the whole story. A site can call itself low-carbon in a broad sense and still fail the basic test of whether its power plan exists in anything like the necessary form. Land has to be secured. Planning has to be approved. The grid has to have space. And if the fallback is a future connection rather than an actual operating renewable system, the clean-power pledge starts looking less like a deliverable and more like a promise to keep trying later. In datacentre power, later is usually where optimism goes to die.
What Lanarkshire means for Britain’s AI growth-zone gamble
Because Lanarkshire sits inside the UK government’s AI growth zone programme, the project has become more than a local planning story. It is now a test of whether Whitehall will stick to its own rules when the pressure is on to announce big-ticket AI investment. The Lanarkshire site was sold as part of Britain’s push to build out AI infrastructure fast, but that pitch only works if the electricity problem can be solved in the real world, not just in a briefing note.
When the growth zones were set up, applicants were meant to show a believable way to power themselves. That could mean an existing grid route, if the connection was actually available, or an independent energy solution that could carry the load on its own. The idea was plain enough. If a data centre campus wants to eat a mountain of electricity, it should arrive with a plan for where that power comes from.
That standard matters because the whole point of a growth zone is speed without fantasy. A site that is still hunting for a workable power route should raise a flag, not get waved through because the ministerial mood music is upbeat. Internal correspondence suggests officials knew the Lanarkshire proposal was not yet meeting that bar, yet the designation moved ahead anyway. That leaves an awkward question hanging over the process. Was the site approved because it had a credible energy plan, or because the label looked good on an AI map?
The awkward part here is not the ambition. It’s the paperwork that seems to have arrived before the electricity.
The politics here are easy to read, even if no minister is likely to put it that bluntly. The government wants to tell a story about Britain as a place where AI can scale quickly, attract capital and generate jobs without months of delay. Fair enough. But electricity supply and planning consent do not bend much for speeches. A grid connection in Scotland can take years, and that’s before anyone starts arguing over turbines, land use, cables, substations or who gets to say yes first. The calendar is rude in that way. It keeps its own schedule.
That tension puts Lanarkshire in a rather embarrassing position for a flagship zone. If the site was accepted while its power plan still looked shaky, then the growth-zone process may have become more permissive than the published criteria suggested. If that happened here, what does it mean for the next wave of AI sites hoping for the same treatment? The danger is obvious. Once one project gets flexibility, the exception can start looking like the rule.
There’s a broader policy problem too. Ministers can chase AI headlines, or they can admit that the physical constraints are slower and less forgiving than the political timetable. They probably can’t have both. The more these projects depend on future grid upgrades, future planning approvals and future commercial deals, the more they drift from the tidy story of instant digital growth. The infrastructure may still get built, but it may arrive later, cost more, and use more conventional power than the glossy launch material implied.
So Lanarkshire matters because it shows where the UK’s AI buildout may be heading. If this is how a flagship growth zone looks at the point of designation, then the rest of the programme may face the same awkward arithmetic. The rhetoric says speed. The land, the cables and the permits say something slower, messier and a good deal less green.
