AI is supporting growth. Miners still have to pay for energy.

The International Monetary Fund expects global growth of 3.1 per cent in 2026 and 3.4 per cent in 2027, according to its new outlook briefing released on 13 October. It also expects 4.7 per cent inflation this year, largely reflecting higher energy and food prices. The figures are forecasts for years still to be completed. IMF briefing.

The useful detail lies beneath the average. The IMF describes a divide between economies benefiting from AI investment or favourable export prices and energy-importing economies facing greater pressure. It flags downside risks from conflict, sovereign yields and trade tensions, alongside the possibility of faster AI productivity gains.

For miners, that makes a single bullish commodity call rather ambitious. A customer building a data centre and a customer trying to keep a factory profitable may face quite different conditions. Both might buy metal. Only one may be expanding its budget.

Copper demand and electricity costs can rise together

AI infrastructure offers a plausible demand case for copper used in electrical equipment and connections. Our economic interpretation is that sustained construction spending could support suppliers even if household consumption or other industrial spending weakens. The IMF’s GDP forecast alone cannot quantify the copper tonnage involved, let alone establish a shortage.

The energy requirement is better documented. In its 2025 Energy and AI report, the International Energy Agency projected data-centre electricity consumption of around 945 terawatt-hours in 2030, more than twice its estimated 2024 level. It also put data centres at roughly one-tenth of global electricity-demand growth to 2030. The dated projections cover data centres generally, including other digital services. AI is an important driver, but the report does not attribute all the demand to AI. IEA executive summary.

The share of global demand deserves attention. A copper thesis relying entirely on AI leaves out other sources of electricity demand. Conversely, a forecast of greater electricity use tells an investor little about a particular mine’s margin until the investor checks its power contract, production costs and selling terms.

Consider a scenario in which data-centre and grid construction continues while energy remains expensive. A copper producer could receive stronger customer demand and still pay more for diesel, electricity or transport. An aluminium producer could encounter a similar tension, with its outcome depending heavily on the price and reliability of power. An energy producer might benefit from higher selling prices, although its own costs, contracts and taxes would still matter.

We would examine those exposures separately. A general forecast for world growth cannot substitute for the calculations a mine manager has to make.

Higher yields need a diagnosis

The Reserve Bank of Australia’s September meeting minutes, released on 13 October, add a useful complication. Members linked rising long-term real yields to some combination of a higher expected neutral interest rate and higher term premia. They said the split was difficult to estimate. A neutral rate is the rate consistent with stable economic conditions; a term premium compensates investors for holding longer-term debt.

Members identified government and AI-related corporate debt issuance among possible contributors to term premia. They also observed that corporate funding remained readily available at the time of their 28–29 September meeting. Their observations describe that meeting’s conditions; borrowers’ positions could have changed since then. RBA minutes.

The distinction affects our interpretation. If the neutral rate has risen, a given short-term policy rate may restrain spending less than previously assumed. If investors instead require more compensation to hold long-term debt, financing can become more expensive without a corresponding improvement in a project’s earnings prospects.

For a mine developer, either way, the actual lending terms matter. We would check when the loan reprices, whether construction needs another funding round and how much of the projected return arrives years after the initial spend. An established operator funding investment from current cash could have more room to wait than a developer whose financing depends on a valuation its lenders no longer accept.

What would reach the wider economy?

Our downside scenario is a combination of persistent energy costs and weaker investment. If AI spending disappoints, suppliers could lose expected orders while households and factories continue paying more for energy. Governments would face pressure to provide support, and central banks could face an awkward choice between weaker activity and inflation that remains too high. The timing and severity would depend on each country’s exposure; the IMF’s briefing does not establish that this outcome will occur.

Australia could experience both sides. Stronger export receipts could support income and public revenue while higher domestic input costs reduce what businesses and households can buy. Even a favourable commodity price would leave a high-cost operation vulnerable if its expenses rose faster.

The strongest counterargument is that AI improves productivity sooner than expected and energy supply expands sufficiently to contain costs. Under that scenario, businesses could grow without the same inflation pressure. The IMF explicitly recognises that upside possibility. We would watch delivered infrastructure, customer orders, energy contracts and financing terms before deciding which scenario a particular miner can withstand.

The investment case becomes more convincing when the company can explain how it will earn a return under several outcomes. A reference to AI in the presentation is considerably easier to arrange.