As the AI Energy Council gathers, the question hanging in the air is: how do we power the future without blowing the grid?
The massive data centres needed to train and run the latest AI are thirsty for electricity. Data centre power use in the UK is on track to multiply six times over by 2034, at which point it could be sucking up nearly a third of all our nation’s electricity. That’s a colossal strain to put on a system that was built for a completely different world, one with predictable, one-way power flows.
The AI Energy Council – a team-up of tech giants, energy firms, the Ofgem regulator, and the National Energy System Operator – has the critical job of trying to predict just how thirsty this AI beast will become. Their work is happening just as the government is pouring £2 billion into its AI Opportunities Action Plan, a grand vision for weaving AI into our hospitals, classrooms, and businesses.
UK Science and Technology Secretary Peter Kyle said: “Giving our researchers and innovators access to the processing power they need will not only maintain our standing as the world’s third-biggest AI power, but put British expertise at the heart of the AI breakthroughs which will improve our lives, modernise our public services, and spark the economic growth which is the cornerstone of our Plan for Change.
“We are clear-eyed though on the need to make sure we can power this golden era for British AI through responsible, sustainable energy sources. Today’s talks will help us drive forward that mission, delivering AI infrastructure which will benefit communities up and down the country for generations to come without ever compromising on our clean energy superpower ambitions.”
The sheer scale of the energy problem is hard to overstate. Globally, the electricity needed for data centres is expected to double in just five years, eventually demanding three times more power than the entire UK currently uses. And AI is the main culprit.
A single rack of AI servers can demand 120 kW of power, a massive leap from the 5-10 kW a normal rack needs. These aren’t steady sips of power, either. AI workloads can spike unpredictably, creating sudden, massive power surges that threaten the stability of the entire grid.
In response, the UK is planning a monumental overhaul. The centrepiece is the “Great Grid Upgrade,” a £58 billion investment designed to be a “once in a generation expansion” of the electricity network. This includes building a new high-capacity electrical superhighway running from north to south and expanding the offshore grid to bring in vast amounts of new wind power.
Ed Miliband, Secretary for Energy Security and Net Zero, commented: “We are making the UK a clean energy superpower, building the homegrown energy this country needs to get bills down for good and create new jobs as part of our Plan for Change.
“Bringing together the biggest players in AI and energy will help us discuss the role AI can play in building a new era of clean electricity for our country, and meeting the power demands of new technology as we build a clean power system for families and businesses.”
But there’s a huge roadblock. Even if we build the wind farms and solar panels, connecting them to the power grid to address surging AI demand right now is another story. The current process is slow, leaving more than 600 renewable energy projects – worth billions – stuck in a queue. Some have been told they could be waiting for 15 years.
Urgent reforms are being pushed through to try and clear this backlog, a vital step if our AI future is to be powered by green energy. The government is also trying to speed things up by declaring data centres “critical national infrastructure” and setting up “AI Growth Zones” where planning and power connections can be fast-tracked.
The data centre industry is shifting from being just part of the problem to becoming part of the solution. Instead of just being passive power hogs, they are becoming active partners in the energy grid. Many are chasing Net Zero targets, investing in their own on-site renewable power, and taking part in “demand-side response” programs. This means they can intelligently pause non-urgent AI tasks when the grid is under stress and fire them up again when green energy is plentiful, helping to balance the whole system.
AI itself could also help. The same complex algorithms that demand so much power can also be used to make our grid smarter, predicting energy spikes and optimising power flow in real-time.
The way forward is clear, but it won’t be easy. The UK has the right ideas and is putting serious money on the table to address the power grid demands of AI, but everything depends on speed and execution. The grid connection jam must be broken, and the Great Grid Upgrade needs to happen at pace.
