The AI revolution needs power. Can the UK keep up?

Can you remember when you first became aware of AI? Not as a science fiction concept, but as something real, tangible and unfolding in the world around you.  

For me, that moment came when I first used ChatGPT, OpenAI’s immensely popular chatbot, shortly after its launch. I was blown away by how quickly it could respond while maintaining logic and coherence, no matter what problem I threw at it. Once the initial shock wore off – that such a piece of software could even exist – I was left with a deep sense of apprehension. Whether for good or ill, AI was going to be a paradigm shift in the scale of the internet and smartphones.  

Last month, the UK government announced the AI Opportunities Action Plan, a set of proposals aimed at ensuring the country takes ownership of this revolutionary technology and integrates it into every level of public infrastructure.  

The government has described its recommendations as “unapologetic in their ambition”, proposing a 20-fold increase in public computing power. By curating a National Data Library and using it to train multiple British-developed AI models per year, the plan aims to establish the nation as a major player in the AI development race, while boosting productivity across both the public and private sectors.  

But with such sweeping ambitions comes an inevitable question: how will all of this be powered?  

AI’s growing energy appetite 

Data centres are the backbone of the AI revolution. They train and deploy AI models by processing vast amounts of data, but their power consumption is immense. The International Energy Agency predicts global data centre demand could exceed 1,000 TWh in 2026 – the equivalent of Japan’s total annual electricity consumption.

To support the UK’s AI expansion, a substantial data centre buildout is needed. As part of the action plan, the government has committed GBP 14bn to infrastructure, including fibre optic cabling and new data centre construction.  

But this rapid expansion could directly conflict with another government initiative, its Clean Power 2030 Action Plan.

A clash of priorities? 

Just one month before unveiling the AI Opportunities Action Plan, the government published Clean Power 2030, a strategy for achieving a 95% renewable electricity system by 2030. Delivering such a transformation in just five years is an extraordinary challenge.  

The plan outlines ambitious targets to:

• Double onshore wind capacity  

• Triple solar capacity  

• Quadruple offshore wind capacity  

• Increase battery storage capacity five-fold 

On top of this, grid infrastructure must expand at an unprecedented rate. To support the growing demand, five times as many pylons and underground lines must be built in the next five years as had been installed in the past three decades.  

Politicians and industry leaders have recognised the monumental scale of this challenge. UK energy secretary Ed Miliband called it “the most ambitious reform to our energy system in generations”, while Fintan Slye, head of National Energy System Operator (Neso), said it is “at the limit of what is achievable”.

So, can the UK meet these energy targets while building world-class AI infrastructure? Or will the power-hungry data centres required under one plan force greater reliance on fossil fuels, undermining the clean energy ambitions of the other?  

Opposing forces? 

At first glance, these two plans seem to be at odds, but they are not entirely incompatible. The AI plan proposes the foundation of an AI Energy Council, led by Miliband and science secretary Peter Kyle, to balance AI expansion with energy supply security. This council may explore alternative power sources, such as small modular nuclear reactors (SMRs), as localised energy solutions.  

Nuclear-powered data centres are already being explored by key AI companies. Google and Microsoft have signed agreements with third-party contractors to secure nuclear energy for their US data centre fleets. In the UK, Great British Nuclear has begun shortlisting bidders for its SMR programme, which could provide a vital energy source for the AI boom.  

Efficiency wildcard 

However, concerns over AI’s power demands were recently upended by the launch of DeepSeek R1, a Chinese AI model designed to rival OpenAI and Google. According to its developers, it was trained and operates at a fraction of the cost and energy consumption of US models.

The market reaction was swift: shares in Nvidia – a key chip supplier for AI – plummeted by hundreds of billions of dollars, while stocks in US power companies like Constellation Energy, Talen Energy and GE Vernova also fell. Investors began questioning whether AI infrastructure really needed the colossal power buildout previously assumed.  

Yet, some argue that greater efficiency won’t reduce overall energy demand. If AI can be made 10 times more efficient, why not do 10 times more for the same energy cost? Instead of limiting power use, efficiency gains could lead to exponential growth in AI applications, reinforcing the need for massive data centre infrastructure.  

The road ahead 

The world of AI is evolving at a staggering pace. Soon enough, our first experiences with AI will feel as outdated as dial-up internet. But the infrastructure required to support this transformation presents one of the most complex challenges of our time – one that must be tackled alongside the already daunting demands of the energy transition.  

The UK is at a crossroads. If it succeeds, it could cement itself as a global AI powerhouse while delivering on its clean energy promises. If it fails, one ambition could end up sabotaging the other. The coming years will reveal whether the government can walk this tightrope – or if one vision will come at the cost of the other.

This article originally appeared as a column on montelnews.com