The pressing question, as the AI Energy Council assembles, is how do we power the future without crashing the grid?
The data centers that will be needed both to train and to run the latest AI models are voracious consumers of electricity. The electricity used by data centers in the UK is set to grow six-fold by 2034 – enough to power a third of all British homes. That’s a powerful load on a system that was built for an entirely earlier age, when power flowed predictably and largely in one direction.
It falls to the AI Energy Council — which includes tech Goliaths, energy companies, the Ofgem regulator and the national grid operator — to try to predict how power-hungry this AI revolution is likely to be. Their efforts come as the government announces a £2 billion investment in its AI Opportunity Plan — an ambitious quest to embed AI in hospitals, schools and businesses up and down the country.
Peter Kyle, UK Secretary of State for Science, Innovation and Technology, said:
“By making this computerscape available to our researchers and innovators, we can ensure that the UK leads the world in AI and stands ready to make use of its benefits,” Gove said, last week, adding:
“Ensuring access to processing power is also central to my aims to make sure that the UK remains at the forefront of AI breakthroughs that save lives and improve lives, modernise public services and develop the high-wage, high-skill jobs that transform societies for the better as we recover from the pandemic.
But we’re absolutely clear that we need to power this golden age of British AI with responsible, sustainable energy.”
Today’s conversations will be part of advancing that mission — ensuring AI infrastructure works for communities in every corner of the country for generations to come, while upholding our commitment to becoming a clean energy superpower.
The Scale of the Energy Challenge
The scale of the energy problem is difficult to exaggerate. Worldwide, the computing power used by data centers is expected to double in just five years. Their total energy consumption could eventually be three times more than the UK’s current use — with AI to blame.
A single rack of AI servers can require as much as 120 kW of power, in contrast to the 5–10 kW that is used by traditional configurations. And the demand isn’t constant — artificial intelligence workloads can spike unpredictably, meaning there can be sudden spikes in demand potentially threatening grid stability.
The UK plans to do the same but on a vast scale. Central to this is the “Great Grid Upgrade” — a £58 billion investment billed as a once-in-a-generation enlargement of the electricity network. The infrastructure would involve constructions such as a superhighway to carry electricity south and north, the offshore grid that would have us draw on vast amounts of wind.
The Secretary of State for Energy Security and Net Zero, Ed Miliband, said:
“We’re turning the UK into a clean energy superpower – driving over quarter of a million jobs as we back not wind, but our businesses, industry, economy and our country to lead,” said Alok Sharma, the business secretary.
“Bringing the world’s best AI and energy companies together helps us have a discussion on how AI can help cut emissions as we clean up our energy system, meeting the demands of new tech as we build a greener grid with record amounts of renewables feeding into it for families and businesses.”
Connection Gridlock and Renewable Bottlenecks
But there’s a significant hurdle. Even if we do develop wind farms and solar arrays, hooking them up to the grid as fast as AI’s demand for power is increasing is quite another story. The present system of hooking power plants to the grid is slow and clogged, with more than 600 renewable energy projects, which represent billions in potential investment, waiting in line. Some developers have been told it will take as long as 15 years.
There are urgent reforms under way to drain this backlog — essential if we want the future of AI to be fueled by clean energy. The government is also attempting to hasten progress by designating data centers as literally being “critical national infrastructure,” and creating “AI Growth Zones” where the planning and electricity connection processes are expedited.
From Power Consumers to Grid Participants
The data center industry is moving from being part of the problem to being part of the solution. Instead of simply sucking down power, many data centers are starting to play in the energy game. They are purchasing on-site renewable energy and participating in “demand-side response” programs. That’s because they can prioritize halting nonessential AI tasks when the grid is under strain, and quickly restart them when an abundance of green energy becomes available — assisting in the overall balancing of the system.
AI itself can help too. The complex algorithms that consume all that electricity are the same ones that can help make the grid smarter — anticipating energy spikes and acting on that information in real time.
The Way Is Clear, but Not Easy
The UK has the right ideas and is backing them with seriously big money to meet AI’s needs on the grid. But it all depends on speed and execution. The nasty logjam of grid connections needs to be smashed, and the Great Grid Upgrade needs to be urgently pursued.
