Britain Could Be Sleepwalking Into AI Blackouts – And 2028 Is the Date People Whisper

The skyline appears to have been erased on a gloomy winter’s afternoon in London. While the city continues to bustle with the sounds of delivery vans nosing into side streets, buses braking, and kettles clicking in kitchens, the quiet drama is taking place outside the ring roads, where the UK’s “AI ambition” is being transformed into reality.

As you drive along the M25 corridor, you begin to notice the signs: new substations popping up like punctuation, new fencing, and anonymous industrial plots. From the outside, data centers appear to be large boxes with few windows and a shrug of architectural style. Nevertheless, Ofgem has now revealed that 140 planned data center projects are requesting grid connections totaling about 50 gigawatts, which is just a little bit more than Britain’s current peak demand of about 45GW.

Even seasoned energy people become reticent when they see that kind of figure.

According to the political line, Britain wants to grow. This includes artificial intelligence (AI), which is now viewed more like infrastructure—the kind you fight over—than software. Energy Secretary Ed Miliband explained in a speech late last year that electrification and “new industries like AI” are contributing factors to the possibility that the demand for electricity will at least double by 2050.

AI as national destiny is a clever rhetorical device, but the grid isn’t powered by rhetoric. It is powered by copper, planning permission, and timely projects—not exactly Britain’s favorite pastime.

A part of the 2028 blackout discussion seems to be a ghost story meant to keep ministers awake. Reports from 2024 revealed that National Grid executives had privately warned that unless the South-East paid more for electricity, it might experience blackouts by 2028. This was a disturbing suggestion disguised as regional pricing logic.

However, by pointing out that some public claims misinterpret what grid studies actually said, the nation’s system operators and official analyses have also pushed back against the way “blackouts” are tossed around. Therefore, it is possible that the “by 2028” claim is exaggerated. Nevertheless, it continues because it feels real that the direction of travel—more load, faster than wires—is real.

The tone of Ofgem’s own description is that of a regulator gazing at an inbox that keeps filling up. According to the watchdog, the number of connection requests has surpassed even the most ambitious projections, and it has publicly expressed concern that many of the schemes in line might be speculative or financially unviable, thereby blocking the way for projects that are truly important.

The subtle nightmare is that there are not only shortages but also paralysis due to an excessive number of applications, a lack of certainty, and the grid turning into a bottleneck that cannot be easily “innovated” out.

Furthermore, it’s not as though the British system has cheap slack. According to Reuters, the UK already has some of the most expensive electricity in the world due to high wholesale prices and a number of structural limitations, including expensive new construction, sluggish infrastructure improvements, and the uncomfortable reality that “always-on” power is more difficult to achieve than politicians portray.

It’s unclear who will bear the cost of strengthening the system and how quickly the nation can expand without making every pylon into a local conflict, but investors appear to think that demand for AI is unavoidable.

At this point, the framing of “AI-induced blackout” becomes both helpful and a little deceptive. AI does not eliminate power. Too much demand at the wrong time, too little stable supply, transmission limitations, postponed projects, and human systems making brittle decisions are some of the traditional causes of grid failure. Simply put, AI comes at the worst possible moment, just as Britain is already working to electrify transportation and heating, decarbonize, and keep bills politically viable. It’s similar to putting another bulky suitcase on a trolley with a swaying wheel.

Additionally, the cultural attitude toward this is evolving. Data centers used to be a specialized planning topic that was easily overlooked until a few years ago. They are now regarded as strategic assets, even “growth zones,” and there is a sense of urgency to that rebranding. It’s difficult to ignore how swiftly the topic shifts from “cool tech jobs” to “who gets electricity first” when you watch the conversation change. That won’t be an issue in the future. That is currently a queue-management issue.

So, by 2028, will AI cause blackouts in Britain? It’s possible, but not in the way the phrase implies—a cinematic, national collapse. Local strain, increased expenses, more difficult trade-offs, and the kind of emergency measures that governments detest acknowledging until they are already in use are the more likely risks.

The most obvious thing is the disparity in pace: data centers can be constructed in a few years, while grid reinforcements and new generation typically take longer, arguing with all stakeholders along the way. This is especially evident when you’re standing close to one of those new substations humming next to a muddy construction site.

AI is not “too powerful,” if there is a lesson to be learned from the growing panic. The problem is that Britain is attempting to manage two major national projects simultaneously—the energy transition and the AI rollout—on an electrical grid that was never intended to operate at this rate. It’s possible that the blackouts won’t arrive on time. There is already tension.