Will AI rewrite the rules for mining in the UK?

Data centres, the fundamental infrastructure essential to AI, now consume more electricity than entire nations, according to Bloomberg. Yet the concern around energy usage belies skyrocketing demand for other commodities: water for cooling, land for servers and minerals for infrastructure. As AI improves, will that demand go away, or grow greater still?

Better with time

It was assumed that building the best AI models required paying for the best computer chips. Nvidia, an American technology company, the near-exclusive provider of those chips, came to boast a huge market capitalisation of $3.5 trillion. Then DeepSeek arrived.

DeepSeek, a rival Chinese AI company, trained its large language model for a paltry $6 million, far less than the $100 million cost for OpenAI's GPT-4 in 2023. It also required far less computing power to deliver only a marginally lower standard of responses. On the day of DeepSeek’s release, $600 million was wiped off Nvidia’s value as the world realised there was another way to deploy AI.

DeepSeek shows us there is still much to discover about AI, and much to improve, too. As the computing power needed to train and operate AI models plummets, you might imagine the consumption of other resources might fall too. A tonne of copper, a litre of water, a megawatt of electricity; with better understanding, we will be able to do more with less. Maybe, but that’s not the point.

Jevons paradox

The Jevons paradox posits that increased efficiency leads to increased, rather than decreased, overall consumption of a resource. As cars became more fuel-efficient, did we consume less fuel? Or did we travel further than ever before? As our homes become more energy-efficient, are we content with lower bills? Or do we nudge the thermostat up?

The same is plainly true of AI. ChatGPT had an estimated 400 million active users in April 2025, up from 300 million in December 2024 and 100 million in November 2023. As AI becomes cheaper, as well as better, it is finding more applications and demanding more resources. 

So AI is here to stay and so is its appetite. We know that’s going to mean more energy: the current total power consumed by the global data centre industry is around 55 GW; by 2030, that’s expected to be 122 GW, according to Goldman Sachs. It’s also going to mean a lot more minerals: cobalt, nickel, zinc, a whole selection of rare earth elements, but none more so than copper. Each GW of applied power requires up to 65,000 tonnes of copper alone, according to Schneider Electric.

Wasting away

What does that mean for the UK, a nation without any active copper mines? Firstly, a reliance on imports, most of which, as can be seen from figure one, come from the EU. This is a relatively balanced arrangement, with raw materials coming in and some more refined products going back. The arrangement with Asia and Oceania, on the other hand, sees more than 200,000 tonnes of copper exports in 2024 alone, and less than 5,000 tonnes of imports.

Digging deeper, figures two and three show us that much of that 200,000 tonnes is waste and scrap in the form of copper alloys and refined copper. Figure four shows us that much of it is headed to China (45%) and India (22%).

The question that naturally arises is: can the UK continue to leak so much copper? After all, the government wants growth and the AI Opportunities Action Plan is one policy designed to achieve that. The very first pillar of that is to lay the foundations for AI to flourish in the UK by facilitating the “rapid build-out of data centres”, which can only happen with the correct materials.

Unearthing new possibilities

AI can demand vast quantities of energy and resources because of the value we expect it to deliver in the long run. Take mining itself. Fortescue, an Australian mining giant, has saved $500 million by using AI to optimise operations. For example, if rain is coming, energy-intensive tasks are brought forward to make use of abundant solar energy. 

Is it possible that the energy cost of copper recycling could become bearable in the UK, if only to keep our AI ambitions on track? Anglesey Mining is seeking to reopen the Parys Mountain copper mine, once the largest copper mine in the world. If tests confirm substantial deposits of copper, zinc, lead, silver, and gold, up to $1 billion worth of value could be extracted – 40% of revenue would come from copper, and 33% from zinc, both of which are essential to AI and our future view of the world. AI is coming and shaking long-held beliefs through the value it offers. Sectors must reevaluate their assumptions to see where opportunity and risk might lie.

Further information

Contact Joe Lloyd or Stuart Jeffries

What are critical minerals and why are they important?