A Scottish AI megasite promised green energy. The reality looks very different.
An £8.2 billion AI data centre complex in rural Scotland overstated its renewable energy commitments. Finding that out required journalists to leave their desks and travel there.
Key points
- An £8.2 billion AI data centre complex planned for rural Scotland misrepresented its renewable energy commitments, according to a Guardian investigation published this week.
- Reporters physically travelled to Scotland to verify claims the site could not support with paperwork alone.
- The story signals a wider shift: AI infrastructure now has a physical footprint, and checking it requires boots on the ground.
- Guardian journalist Aisha Down said physical constraints, not software limits, will decide whether the AI boom succeeds or collapses.
For most of the past decade, covering the tech industry meant reading press releases and watching livestreams. The hardware was tucked away somewhere nobody needed to go.
That's changed.
AI runs on data centres: enormous warehouse-style buildings packed with specialised computers that consume staggering amounts of electricity and water. Those buildings need land, grid connections, planning permission, and a believable story about where the energy comes from.
That last part is where a Scottish project ran into trouble.
The Guardian's global tech team found that an £8.2 billion AI infrastructure complex planned for a rural Scottish site had misrepresented its intentions to run entirely on renewable energy generated on-site. The gap only became clear once reporters travelled there in person.
"Our reporting is showing that you can't simply wave a magic wand and have a datacentre appear," said Aisha Down, who covers AI for the Guardian and made the trip to Scotland. "There are a lot of huge physical constraints and reality checks. These physical, tangible things are what makes or sinks the AI boom."
This matters well beyond Scotland. AI2Day first covered green energy claims attached to AI infrastructure on 13 July 2026, in our report on Sunrun's plan to turn domestic solar homes into distributed computing nodes. The Scottish case is a harder version of the same problem: corporate promises that can't be verified from a desk.
Why should ordinary people care about this?
Because the cost of AI infrastructure lands somewhere, and right now that somewhere is often near the site itself: local electricity grids, planning systems, and the environment around these facilities.
If a data centre claims green energy but draws from the national grid like any large industrial site, that distorts emissions figures and undermines every corporate climate pledge that follows the same template.
Down's phrase is worth holding: physical constraints make or sink the AI boom. The billions flowing into AI become steel, concrete, cooling systems, power cables. Those things take time. They face planning objections. Grid capacity is finite, shared with hospitals and homes.
Should you worry?
Not panic, but pay attention. When a new data centre announcement arrives with a clean-energy headline, ask who verified it and whether anyone visited the site. The skill that once sent reporters to council meetings now sends them to server farms in remote countryside. That shoe-leather instinct is the only reliable check on promises that look perfect on a slide deck.
