Scotland's AI Fairy Dust

Remember the Lanarkshire AI datacentre? The Guardian has obtained documents showing that the project has no prospect of meeting its renewables promise. The Scottish government privately acknowledged an 'issue' with 'power provision'. That is a polite way of saying the whole thing was a mirage.

The project was announced with the usual fanfare: a massive datacentre, thousands of jobs, a beacon of green AI. The promise was that it would draw power from on-site renewables, avoiding the national grid. This was crucial because AI datacentres are energy hogs—a single training run can consume as much electricity as a small town. The industry has been scrambling to appear responsible, and Lanarkshire was supposed to be the poster child. Except the posters lied.

The Guardian's investigation reveals that the plans for on-site generation were never viable. The government's internal admission of a 'power provision issue' means the datacentre will inevitably draw from the grid, which in Scotland still relies on gas and nuclear. The entire green premise collapses. But the announcement had already served its purpose: it generated headlines, attracted investment, and allowed politicians to pose as climate champions. The truth, buried in internal memos, was inconvenient but not damaging—because nobody checks.

This is not an isolated case. Across the UK and beyond, AI infrastructure is being sold with green credentials that evaporate on inspection. The industry knows that 'sustainable AI' is an oxymoron at current scales, but admitting that would kill the narrative. So they sprinkle fairy dust—announcements of solar farms that never get built, carbon offsets that don't offset, and 'renewable energy certificates' that are just accounting tricks. The Lanarkshire project is merely the one that got caught.

The Scottish government's role is particularly shameful. Rather than scrutinising the claims, they amplified them. They knew the 'issue' and said nothing. This is not a failure of due diligence; it is a choice to prioritise the appearance of progress over actual progress. The same pattern appears in every sector where AI promises to transform: healthcare, education, transport. The gap between the sales pitch and the reality is vast, and governments are eager accomplices.

What makes this especially galling is the genuine need for responsible AI infrastructure. If we are to build AI systems, we must power them somehow. But the solution is not to lie about it. It is to be honest about the costs, invest in genuine renewable capacity, and perhaps—heresy—ask whether every AI application is worth the energy. Instead, we get datacentre projects that will emit more carbon than promised, while executives take victory laps on Twitter.

The Lanarkshire debacle should be a warning. But it won't be. The next datacentre will be announced with similar fanfare, similar promises, and similar hidden truths. Because the AI industry runs on hype, and hype runs on fairy dust. And as long as politicians and journalists refuse to check the ingredients, there will always be another project willing to sell it.