Power play: Can the grid cope with AI’s insatiable appetite
As Britain sprints towards a position of global AI superpower, a new crisis is looming, quietly, not in the lab, but in the wires beneath our feet.
Because underneath all the buzz of chatbots, generative automation, and AI lies a simple yet profound question: where will all the electricity come from?
This is the focus of the newly formed AI Energy Council, a high-level partnership comprising tech behemoths, energy players, and regulators, hosted by the UK’s National Energy System Operator. Their remit: solve the energy squeeze that could limit the country’s AI ambitions before they even leave the server.
AI is Hungry—and the grid is not ready
Right now, Britain’s data centres already take a sizeable share of national electricity generation, but this is just the tip of the iceberg. If internal estimates are anywhere close to being accurate, the power demand (services only) of UK data centres could grow by a factor of six by 2034, and they could account for nearly a third of all electricity usage.
The numbers worldwide are even more dire. Over the next five years, AI is projected to increase global data centre electricity demand by a factor of three, which is more than some countries use in their entirety.
At the centre of the issue are the AI servers, which consume electricity at rates never seen previously. A normal rack of standard servers will take 5-10kW of power, and a rack of AI servers? Up to 120kW, and that demand is anything but constant. AI jobs, by nature, cause unpredictable spikes and stresses on grids that are designed for stability, not volatility.
Welcome the AI Energy Council—and a £ 2 Billion Tech Bet
This energy crisis comes just as the UK government is initiating the £2 billion AI Opportunities Action plan—a major program to embed AI in schools, hospitals, public services, and industry.
Speaking at the summit of the AI Energy Council, Secretary of Science and Technology Peter Kyle made it clear:
“We’re powering a golden era for British AI. But if we can’t solve the energy piece, it could all come to an end.”
Kyle gives an assurance that the UK will meet the energy needs of AI without compromising targets for clean energy. The work of the council, he says, will establish an “AI infrastructure that powers innovation and protects our climate for generations to come.”
But talk is not going to power the future of Britain. Electricity will.
The Great Grid Upgrade: Britain’s Biggest Energy Overhaul in a Generation
The centerpiece of the UK’s response is the £58 billion “Great Grid Upgrade,” a national program to create a modern energy backbone of the AI age. This will mean:
- High-capacity electric superhighway from north to south.
- Exponentially increased offshore wind infrastructure.
- New grid connections to Scotland’s environmentally abundant Highlands.
- Streamlined grid connection paths to our new mega solar projects.
This is the most significant grid overhaul in modern UK history.
Ed Miliband, Secretary for Energy Security and Net Zero, said:
“The UK is becoming a clean energy superpower. AI is not a problem; it is part of the solution. But we need speed. We need scale. We need all the watts we can muster – and we need them quickly.”
The Backload Bottleneck: 600 Projects Stuck
But ambition alone is not enough to address a more troubling problem: upwards of 600 renewable energy projects are awaiting connection to the national grid in limbo.
Some solar and wind project developers have been told they will wait up to 15 years to connect.
The backlog poses a risk to the energy needed to power the AI revolution. Urgent regulatory reforms at the provincial and federal levels are coming to reduce red tape and speed up planning and clean energy connection timelines.
The government plans to speed up this process by:
- Declaring data centres as “critical national infrastructure” to immediately give them priority access.
From Problem to Partner: The New Role of Data Centres
The data centre sector is changing. Once viewed solely as energy-hungry customers, data centres are beginning to act as grid partners, by doing things that would have been considered impossible only a few years ago:
- Demand-side response programs: AI workloads that are not real-time (for example, model training) may now be paused in times of grid stress and can be restarted when wind or solar are plentiful.
- On-site clean power: Several of the largest data center providers in Canada are investing in private solar fields, hydrogen fuel cells, and installed battery storage to reduce dependence on the national grid.
- AI energy management: The algorithms that are creating energy surges are also being used to assist in balancing the system, predicting spikes, and smoothing demand in real-time. This is important to develop a smart and sustainable grid.
- Creating “AI Growth Zones”, special investments are made to ensure that decisions on power grid expansions and approvals are accelerated.
Critics caution that the provinces and federal government will need to undertake even more ambitious reforms, or the demand for electricity from AI could outpace the capacity of the grid, resulting in businesses, researchers, and hospitals having to keep servers for which they cannot source power.
Grid-Intelligent AI: A Vision for a More Sustainable Future
Ironically, AI may contribute to solving the crisis it is simultaneously creating.
Advanced AI systems are now being trained to do the following:
- Predict with a surprising level of accuracy when short-term spikes in energy demand will occur.
- Control the flow of power between renewable sources and demand hotspots.
- Create a better discharge situation for energy storage by optimizing battery discharge timings.
- Play out various scenarios on the entire power grid, and identify weak points before they lead to blackouts.
If deployed properly, AI could help make the grid more resilient, agile, and capable of enabling the next generation of technology, from autonomous vehicles to sophisticated hospitals.
The Verdict: AI Needs the Grid, But the Grid Needs AI Too
The challenge of the AI Energy Council is unequivocal: transform a looming energy crisis into a transformation opportunity. The UK government is wagering billions that it can deliver a world-leading AI sector along with a cleaner and smarter grid. Will it succeed? This depends on the speed, execution, and coordination of the two different worlds of tech and energy.
The decisions made in this period now will affect not only how powerful Britain’s AI models become, but if we can keep the lights on while we are building them.
We can say with conviction, the AI revolution will not be streamed, coded, or written… unless we plug it in first.
