Ted Newson is a Researcher at the Global Warming Policy Foundation
In September, Andy Burnham spoke to the UN General Assembly about capitalising on the rise of AI, with his visit intended to “drive forward global work” on all matters pertaining to artificial intelligence. His speech was full of hope, but the reality is starkly different. On October 18th, Microsoft pulled out of a £2bn data centre in Essex that had been agreed last year. Politicians talk about Britain embracing AI, but without tackling high electricity prices and a grid that can’t connect new demand, Britain will fall behind.
The company behind the development, Nscale, described by the Telegraph as “Britain’s sovereign data centre provider”, was set to be connected to the grid at the end of this year. However, on account if grid capacity concerns, it may not receive power until the 2030s. The binding constraint is power, not construction. It’s Britain not having the necessary power generation and grid capacity to keep homes, businesses, and data centres running smoothly.
Most worryingly, the company called grid connectivity a “London-wide grid issue”, suggesting that we could see similar issues across the capital. Delaying this project by years will dissuade businesses from investing in AI here and encourage them to set up in places with cheap, abundant energy like the US. If we can’t support one 50MW data centre now, how will Britain find the grid capacity to support an AI boom?
When announcing Britain’s AI strategy last year, then-Prime Minister Keir Starmer praised the Loughton-based project as the “largest UK sovereign AI data centre“. Just a year on, the project is delayed until sometime in the next parliament. Ministers have committed to decarbonising the power sector by 2030, phasing out reliable gas plants in favour of intermittent wind and solar. The Government’s 2030 target also raises questions about whether its plans for firm generation and grid infrastructure will keep pace with demand.
Data centre electricity demand is soaring worldwide. Energy analyst Kathryn Porter expects data centres to add an extra 6GW to national electricity demand, over 13 per cent of the yearly high of 45GW demand we had on February 11th. Britain should be looking at how much power it can generate reliably and at a low cost. In the AI era, electricity is becoming what railway access was in the nineteenth century and good broadband was in the twenty-first: a prerequisite for economic growth.
Britain’s electricity prices are roughly two to three times as high as the US and three to five times as high as in China. With data centres consuming so much electricity, politicians like Zack Polanski argue that we should reject them entirely as they put strains on the power supply, but this will only diminish Britain’s global influence. Instead, we must build a power system that supports new industry not just through new nuclear but also new gas. I’d love nuclear to be built quickly, but in modern Britain it isn’t. Hinkley Point C is now delayed as late as 2031 and has gone well over budget at a cost exceeding £48bn.
Britain’s grid is increasingly powered by intermittent sources. That means wind and solar, which only generate power when weather conditions are right. As a result, to keep supply stable we pay for a parallel ‘backup’ fleet that runs when the wind drops. Last year capacity market costs reached almost £2bn. By 2030, these are expected to reach as high as £8.3bn.
The recent Prosperity Institute report highlighting rising costs also put overall UK balancing, backup, and integration costs at over £40bn a year by 2030/31. These costs are driven by the intermittent capacity we keep adding. Lowering bills, for households and data centres alike, means paying for one reliable system rather than two. Gearing our power sector towards reliable generation wouldn’t eliminate the need for a capacity market altogether, but it would dramatically reduce the cost of it on bills.
Grid investment is essential, but it cannot solve every problem on its own. Locating generation close to major centres of demand could reduce transmission requirements and ease local constraints. New power sources tend to be intermittent which creates further supply issues when the wind drops.
NESO issued an “emergency assistance” request for more power on October 14th and increasingly we are dependent on power imports to keep the lights on. We might like to treat imports as guaranteed, but a recent NESO report warned that neighbouring countries can restrict power exports, particularly when they need electricity to balance their own grids. France did exactly that in June. The most reliable power is dispatchable and close to demand: not wind in Scotland, and not nuclear from France.
If Britain’s grid is going to be trusted by big business, it must treat power as integral to national security and economic growth. The AI boom can’t wait until the 2030s in Britain and connecting data centres faster only works if generation is built alongside them. What we need is cheap, reliable power in abundance.
Building gas close to demand while fast-tracking nuclear is how we get supply up and cut the hidden balancing, backup, and integration costs of remote intermittent power. Until Britain can power them, the next Loughton will go somewhere else.
Ted Newson is a Researcher at the Global Warming Policy Foundation
In September, Andy Burnham spoke to the UN General Assembly about capitalising on the rise of AI, with his visit intended to “drive forward global work” on all matters pertaining to artificial intelligence. His speech was full of hope, but the reality is starkly different. On October 18th, Microsoft pulled out of a £2bn data centre in Essex that had been agreed last year. Politicians talk about Britain embracing AI, but without tackling high electricity prices and a grid that can’t connect new demand, Britain will fall behind.
The company behind the development, Nscale, described by the Telegraph as “Britain’s sovereign data centre provider”, was set to be connected to the grid at the end of this year. However, on account if grid capacity concerns, it may not receive power until the 2030s. The binding constraint is power, not construction. It’s Britain not having the necessary power generation and grid capacity to keep homes, businesses, and data centres running smoothly.
Most worryingly, the company called grid connectivity a “London-wide grid issue”, suggesting that we could see similar issues across the capital. Delaying this project by years will dissuade businesses from investing in AI here and encourage them to set up in places with cheap, abundant energy like the US. If we can’t support one 50MW data centre now, how will Britain find the grid capacity to support an AI boom?
When announcing Britain’s AI strategy last year, then-Prime Minister Keir Starmer praised the Loughton-based project as the “largest UK sovereign AI data centre“. Just a year on, the project is delayed until sometime in the next parliament. Ministers have committed to decarbonising the power sector by 2030, phasing out reliable gas plants in favour of intermittent wind and solar. The Government’s 2030 target also raises questions about whether its plans for firm generation and grid infrastructure will keep pace with demand.
Data centre electricity demand is soaring worldwide. Energy analyst Kathryn Porter expects data centres to add an extra 6GW to national electricity demand, over 13 per cent of the yearly high of 45GW demand we had on February 11th. Britain should be looking at how much power it can generate reliably and at a low cost. In the AI era, electricity is becoming what railway access was in the nineteenth century and good broadband was in the twenty-first: a prerequisite for economic growth.
Britain’s electricity prices are roughly two to three times as high as the US and three to five times as high as in China. With data centres consuming so much electricity, politicians like Zack Polanski argue that we should reject them entirely as they put strains on the power supply, but this will only diminish Britain’s global influence. Instead, we must build a power system that supports new industry not just through new nuclear but also new gas. I’d love nuclear to be built quickly, but in modern Britain it isn’t. Hinkley Point C is now delayed as late as 2031 and has gone well over budget at a cost exceeding £48bn.
Britain’s grid is increasingly powered by intermittent sources. That means wind and solar, which only generate power when weather conditions are right. As a result, to keep supply stable we pay for a parallel ‘backup’ fleet that runs when the wind drops. Last year capacity market costs reached almost £2bn. By 2030, these are expected to reach as high as £8.3bn.
The recent Prosperity Institute report highlighting rising costs also put overall UK balancing, backup, and integration costs at over £40bn a year by 2030/31. These costs are driven by the intermittent capacity we keep adding. Lowering bills, for households and data centres alike, means paying for one reliable system rather than two. Gearing our power sector towards reliable generation wouldn’t eliminate the need for a capacity market altogether, but it would dramatically reduce the cost of it on bills.
Grid investment is essential, but it cannot solve every problem on its own. Locating generation close to major centres of demand could reduce transmission requirements and ease local constraints. New power sources tend to be intermittent which creates further supply issues when the wind drops.
NESO issued an “emergency assistance” request for more power on October 14th and increasingly we are dependent on power imports to keep the lights on. We might like to treat imports as guaranteed, but a recent NESO report warned that neighbouring countries can restrict power exports, particularly when they need electricity to balance their own grids. France did exactly that in June. The most reliable power is dispatchable and close to demand: not wind in Scotland, and not nuclear from France.
If Britain’s grid is going to be trusted by big business, it must treat power as integral to national security and economic growth. The AI boom can’t wait until the 2030s in Britain and connecting data centres faster only works if generation is built alongside them. What we need is cheap, reliable power in abundance.
Building gas close to demand while fast-tracking nuclear is how we get supply up and cut the hidden balancing, backup, and integration costs of remote intermittent power. Until Britain can power them, the next Loughton will go somewhere else.