A few weeks ago, I was sitting on a patio with some of my closest friends, deep in conversation about marriage. Most of us wanted the safest, most future-proof path to a life partner. After a few glasses of red wine and no shortage of debate, we agreed that the future may be uncertain, but a partner who shares your core values and has a good heart is a solid foundation for marriage. And yet, a quieter question surfaced. Do we know what we are getting into? Especially if we have to compromise on those values for the marriage itself?
Today’s world suggests the postwar order of open trade and liberal democracy is in retreat. Supply chains are regionalising, and tariffs and trade barriers are becoming routine. Against this backdrop, ‘tech-nationalism’ has taken hold. Advanced AI is controlled by a handful of companies, most based in the US. That dependence extends beyond AI models and data storage. It extends to the energy infrastructure and policy of the countries where these companies operate.
As AI becomes embedded across the mining value chain, can we be confident our sustainability commitments will keep pace? We scrutinise the sustainability of our physical supply chains, but rarely our digital ones. To be clear, I am not an AI sceptic. It offers real opportunities, and it is my job to explore its potential. The engineer in me believes there is a smart way forward. But that question from the patio keeps echoing.
In ‘We Did the Math on AI’s Energy Footprint. Here’s the Story You Haven’t Heard’, MIT Technology Review examined how AI could reshape America’s energy demand, highlighting a Lawrence Berkeley National Laboratory study estimating that by 2028, AI data centres could account for 6 – 12 % of total US electricity consumption. That is enough to power 22% of American households for a year.
A single prompt to an LLM uses less electricity than running a kitchen appliance for a few seconds. But multiplied across daily life, the totals add up. The MIT Technology Review article also found that tech companies are rarely transparent about their energy sources or usage, and since AI is not classified as its own industry in the US, no standardised reporting forces disclosure. That is a strange paradox, given that these data centres directly affect electricity bills, and it makes the question of where that power comes from more urgent.
Consider a national vehicle fleet. The incentive to electrify it shrinks if the grid runs heavily on coal and natural gas, the same fuels partly powering today’s data centres. According to the IEA’s ‘Energy and AI’ report, fossil fuels supply just over 50% of the electricity used by US data centres, largely because these facilities need stable, round-the-clock power. Grid flexibility can smooth demand spikes, but it does not solve the underlying capacity constraints, which are only expected to grow.
There is a caveat, though. The IEA also estimates AI applications in other sectors could cut CO2 emissions by 1.4 gigatonnes by 2035, roughly four times the emissions data centres themselves are projected to produce. But it warns these gains are not guaranteed and could be offset by rebound effects, where efficiency simply drives higher consumption. The point is, it is a complex relationship. Energy and technology in a global economy are becoming increasingly interconnected.
If the mining industry is serious about playing a decisive role in the climate transition, it is time we asked what an expanded reliance on AI means, not just in terms of token costs, third-party data dependence, or a geopolitical shock that redraws the map, but in energy as well. How far does our responsibility extend across that value chain? The same rigour we have applied to physical supply chains can be applied here too. It is a matter of asking the right questions early.
I keep coming back to that night on the patio, talking about marriage with old friends. Do we know what we are getting into?
