Dear SoTA,
The UK will not have sovereign AI without sovereign energy. Every conversation about domestic compute, open-weight models and AI state capacity sits on top of a simpler question that almost nobody in the AI debate wants to answer: where does the power come from, and what does it cost?
Right now the answer is embarrassing. Electricity in the UK costs roughly $0.35 per kWh. In the US it is $0.16. In China it is $0.08. A frontier GPU cluster is, at its core, a machine for converting electricity into intelligence, and we are asking UK companies to run that machine on power that costs four times what their competitors pay. High energy costs kill computation scale. You can fund all the GPUs you like through the Sovereign AI Fund, but if the electrons feeding them are the most expensive in the developed world, the economics never close.
The demand side is not waiting for us. AI data centre electricity demand rose 17% in 2025, nearly six times faster than overall global electricity growth, and is projected to triple by 2030. China understands this clearly. It added 543 GW of generating capacity in 2025 alone: solar, wind, nuclear, coal, gas, all of it simultaneously, at enormous scale. The US added around 60 GW. The UK added 4 to 5 GW. China builds more electrical capacity in a couple of years than the US and UK grids have accumulated in their entire histories. The countries that win the AI era are the ones with the most power. This is not a metaphor.
The frustrating part is that the UK’s problem is not generation technology. We are good at building wind farms. Roughly 45% of our electricity already comes from renewables. The problem is everything around them, and it comes down to three engineering bottlenecks.
First, the connection queue. The grid was designed in the 1950s to move power from a few large, predictable fossil plants in the Midlands to big southern cities. When renewable applications exploded, grid slots were allocated first come, first served, so developers flooded the queue with speculative applications, and dead projects sat blocking live ones for years. Connecting a new project can take up to 15 years. Meanwhile, clean generation has flipped the grid’s geography: the wind is in Scotland and the North Sea, the demand is in the south, and the cables between them are too thin. The result is genuinely absurd. Last year the UK paid around £1.5 billion to switch off wind farms and burn gas instead, because the power lines couldn’t carry the clean electricity to where it was needed.
Second, the cost stack. The subsidy schemes that kickstarted the transition, the Renewables Obligation and CfDs, cost over £10 billion in 2024, loaded straight onto electricity bills. They did their job, deployment happened at scale, but they made electricity artificially expensive relative to gas, which disincentivises exactly the electrification we need and prices out large industrial loads like data centres.
Third, intermittency. During a winter Dunkelflaute, a long stretch of cold, windless, cloudy weather, renewable output can collapse to near zero just as heating demand spikes, and gas has to surge to supply up to 80% of Britain’s power. Batteries handle hours, not weeks. Until firm clean capacity exists at scale, every AI data centre we site here inherits that volatility and that gas price exposure.
None of these are physics problems. They are build problems. And the UK has mostly responded by rationing, delaying and optimising around scarcity rather than building its way out.
We can’t change policy, we can try to influence it, but that’s not in our control. What is in our control is engineering. And we believe the majority of costs in energy can be taken out with great engineering.
At Fuse we’re building an AI-native, vertically integrated energy company: developing power generation across solar and gas, trading power, installing distributed energy, and selling directly to customers. We’ll own and manage 100MW of operational generation by Q1 2027, with more than 2GW of generation and storage in the pipeline.
The strategy is deliberately unglamorous. Build generation cheaper than anyone else, our EPC cost is 30% below market per Megawatt-peak (MWp) built, our cost to serve is 2-3X lower than incumbents, sell energy cheaper than they do, until energy is genuinely low cost & abundant. It’s working: we’ve lowered total energy costs by 17% to date (compared to price cap).
We start with existing energy technology because it works today and generates cash today. Once free cash flow is strong enough, we’ll start pushing into frontier energy technologies like nuclear fusion and fission. Do existing energy better first.
Because we own assets across the stack, we can deploy AI effectively: from demand & generation forecasting to customer AI. The endgame for AI specifically is co-locating generation with compute, so power reaches clusters without queuing 15 years behind a dead speculative application. That includes owning and optimising the grid itself. Energy and compute are one strategic stack now, and the £500 million Sovereign AI Fund shows the government is starting to see it the same way.
One thing we refuse to do is dress this up as climate tech. Aeroplanes must be safe, but their purpose is to fly. Energy should not damage the environment, but its purpose is to be cheap and abundant. Sustainability is the by-product of building well, not the mission. The mission is terawatt-hours cheap enough that no UK lab, startup or data centre ever has to model electricity cost as an existential risk.
Everything above is an engineering problem, and we need engineers. If you work on power electronics, mechanical engineering, quant research, AI, or you’re a strong generalist software engineer, we’re hiring across the stack in the UK & beyond. Write to me directly, or look at fuseenergy.com.
The UK once pioneered the energy grid and commercial nuclear. It has all the ingredients to be a world leader in energy again.
Yours,
Alan
Author biography:
Alan is a founder at Fuse Energy, a vertically integrated, AI-native energy company. Fuse builds solar and gas generation, trades power, buys and optimises grids, and sells energy directly, on track to own & operate 100MW of generation by Q1 2027 with a 2GW+ generation and storage pipeline. Fuse has raised $210 million at a $5 billion valuation to make energy cheap and abundant.
Write to the Society for Technological Advancement on letters@ilikethefuture.com.
Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.