Everything About AI
NextFin News - BHP's finance chief is making a simple but consequential argument: the world's largest mining company has quietly stopped being an iron-ore proxy and become a bet on electrification, and the artificial-intelligence buildout is now one of the clearest reasons why. Vandita Pant, the chief financial officer of the Australian resources group, said in an interview this week that copper sits at the intersection of three durable demand streams — the energy transition, urbanization and AI data centres — while China's steel consumption pivots away from a slumping property market toward advanced manufacturing. The timing matters. Copper was trading near record levels at about US$6.68 a pound on October 28, up roughly 29% from a year earlier, and BHP's own copper business just delivered a record US$18.2 billion of underlying EBITDA in fiscal 2026 — more than half the group total for the first time in the company's history.
The central question is not whether copper demand will grow. BHP's own forecast puts global demand rising from about 34 million tonnes in 2026 to more than 50 million tonnes by 2050, an increase of close to 70%. The question is whether the supply side can respond fast enough to stop a structural deficit from forming — and whether a miner that has spent a decade pruning its portfolio can execute the expansion quickly enough to capture it. BHP's answer, backed by a US$10 billion annual capital plan and a balance sheet with net debt below US$9 billion, is that it can. The market is being asked to believe that mining's next era is won not by the biggest dealmaker, but by the fastest builder.
The Situation: A Copper Company That Still Mines Iron Ore
BHP's fiscal 2026 results, released in August, show how far the transformation has gone. Group underlying EBITDA reached US$32.9 billion, up 27%, on revenue of US$58.8 billion. Free cash flow jumped 83% to US$9.8 billion, underlying attributable profit rose 30% to US$13 billion, and return on capital employed came in at 26.1%. The board returned a dividend of 172 US cents per share — US$8.7 billion in total, the largest distribution in four years. Those are the numbers of a company operating at the top of a cycle.
But the mix shift is the real story. The copper division generated US$18.2 billion of EBITDA at a 70% margin, contributing 54% of group earnings. By-product credits — gold, silver and uranium recovered alongside copper — added US$4.5 billion of revenue, up 45% from the prior year, effectively subsidizing the cost of copper production at Escondida and the South Australian operations. Iron ore, long the dominant profit engine and the metric by which investors have valued BHP for decades, is no longer the whole story.
Production discipline underpins the earnings quality. BHP produced about 2 million tonnes of copper in FY2026 for the second consecutive year, with Escondida — the world's largest copper mine — contributing 1,261 kilotonnes. Yet the company is pointedly not assuming today's price lasts. Its medium-term financial planning uses a copper price assumption of US$5.40 a pound for fiscal 2027 through 2031, roughly 19% below the spot level, and US$4.76 a pound beyond that. In other words, BHP is underwriting a multi-billion-dollar growth plan with a copper price well under the current market, not above it. That gap between the planning price and the spot price is the margin of safety management is buying into its own thesis.
The pivot is not just commodity selection; it is geography. Pant framed the demand story as Asia-centric: China's property-led steel consumption is fading, but advanced manufacturing is picking up the slack, and India's infrastructure runway offers a structural growth path that could outlast the Chinese cycle. For a miner whose project pipeline spans decades, that geographic rotation is the difference between a cyclical rebound and a multi-decade demand regime.
Why AI Is a Copper Story Before It Is a Chip Story
The market's first-order read of the AI boom runs through semiconductors, memory and utilities — the visible layers of the stack. BHP's argument pushes one layer deeper: every watt of AI compute needs copper, and the metal arrives late to the party but stays longer. The company's own outlook puts data-centre-associated copper demand at roughly 3 million tonnes a year by 2050, up about sixfold from 2024. That single increment would absorb a meaningful share of the roughly 17 million tonnes of incremental global demand BHP expects over the same period.
The mechanism is physical, not financial. AI training data centres are far more power-dense than conventional cloud facilities, and power density means transformers, switchgear, busbars and cabling — all copper-intensive. Standard data centres used for cloud storage require about two tonnes of copper for every megawatt of power capacity; AI training facilities push that intensity materially higher because of liquid cooling and high-performance computing loads. The transmission chain runs: AI capital expenditure to data-centre power buildout to grid and substation investment to copper demand. That chain is longer than the chip-order cycle, which is why the copper demand arrives with a lag and persists after the initial server spend has been digested.
This is the second-order point the market is still pricing imperfectly. Investors have repriced copper miners on the energy transition — renewables and electric vehicles — but AI is a new, incremental demand leg that barely featured in models two years ago. It also lands at the right time for producers and the wrong time for consumers: a supply side that cannot respond quickly. Greenfield mines take a decade or more from discovery to first production, ore grades are declining at mature assets, and permitting risk has risen across most major jurisdictions. When demand grows on three independent vectors — electrification, urbanization and digitalization — and supply responds on a decade-long clock, the resulting imbalance is structural rather than cyclical.
There is also a cross-asset implication that copper bulls have not fully worked through. If data-centre power demand forces utilities to build transmission faster than historical capex cycles, copper demand becomes less correlated with the traditional industrial cycle and more tied to regulated rate-base growth — a demand profile that is stickier and less sensitive to GDP. That would re-rate the entire copper complex, not just the miners.
BHP's copper business generated a record US$18 billion of EBITDA in FY2026, more than half the Group total for the first time, supported by a 70% margin and a US$4.5 billion contribution from by-products.
The Discipline Test: Organic Growth Over M&A
Pant has repeatedly framed BHP's capital allocation around organic growth rather than deals, and the numbers explain why. In an August interview, she described the pipeline as "organic growth options... across commodities, we have high conviction in, which is potash and copper and some smaller projects in iron ore." The company's medium-term capital-expenditure guidance sits at about US$10 billion a year, roughly 40% of it growth capex — a level that funds expansion without stretching the balance sheet that management says ends the year below its target net debt.
The logic is financial, not ideological. Organic projects in existing jurisdictions — expanding Escondida, developing Spence and the Olympic Dam path in South Australia, adding potash capacity at Jansen — carry known geology and known permitting paths. M&A, by contrast, forces a buyer to pay for someone else's optionality at a premium. BHP's aborted approach to Anglo American in 2024 showed the political and execution risk of mega-deals; the smaller Filo acquisition completed with Lundin Mining, at C$4 billion, fits a different profile — bolt-on exposure rather than balance-sheet-transforming risk.
The company has also been willing to monetize non-core assets to fund the pivot. In December 2025, BHP agreed to sell a stake in the inland power network serving its Western Australian iron-ore operations to BlackRock's Global Infrastructure Partners for US$2 billion, freeing capital for copper and potash investment while retaining operational control. That is the shape of the new BHP: keep the operating asset, sell the infrastructure, recycle the proceeds into the growth commodity.
The test ahead is execution speed. Pant has said the company is working to move projects to final investment decision "incredibly quickly" so the underlying copper-equivalent growth can be delivered. That matters because a copper deficit, if real, is a time-bound opportunity: the producer that brings tonnes online first captures the highest prices in the cycle. The risk is that speed and discipline conflict. BHP's FY2027 copper production guidance — 1,650 to 1,800 kilotonnes, down from 1,952.8 kilotonnes in FY2026 — is a reminder that near-term output can step down even as the long-term thesis strengthens. Escondida's medium-term guidance of 900 to 1,000 kilotonnes a year for fiscal 2028 through 2031 sits well below its FY2026 output of 1,261 kilotonnes, reflecting grade decline and the timing of expansion capital. Growth is a promise, not a run-rate.
The Asia Pivot: China's Steel, India's Runway
The Asia-centric frame is where the cyclical and structural threads separate, and it is the part of the thesis that investors have found hardest to underwrite. China's property sector, which once absorbed a disproportionate share of global steel, is in a multi-year contraction. That is cyclical pressure on iron ore and metallurgical coal, and it is why BHP's iron-ore earnings have been the weak link in an otherwise strong year. But management's argument is that Chinese demand is rotating into advanced manufacturing — grid equipment, machinery, defence, electric vehicles — which is more copper-intensive per unit of GDP than property construction.
India is the structural counterweight. Its infrastructure runway — railways, power grids, urban transit, housing — maps almost one-for-one onto BHP's commodity basket. Unlike China's property-driven cycle, India's buildout is earlier in the adoption curve and less levered to a single asset class. That is why the pitch is not "China recovers" but "Asia diversifies." For a company with a 150-year asset life, a diversified Asian demand base is worth more than a concentrated one, even if the near-term tonnage is smaller.
The evidence floor for calling this structural rather than cyclical is met on three counts. First, the demand drivers are rule- and technology-based: decarbonization policy, EV mandates and data-centre power requirements are not inventory cycles that mean-revert. Second, the supply response is structurally slow: permitting, capital and declining ore grades cannot be fixed by a rate cut or a stimulus package. Third, the demand base is broadening geographically rather than concentrating, which reduces single-economy cycle risk. A cyclical call would require three historical cycles of mean reversion; the current setup has no clean historical analog because no prior cycle combined electrification, digitalization and emerging-market urbanization simultaneously.
The Counter-Thesis: Where the Copper Bull Case Breaks
The strongest argument against BHP's thesis is that AI demand is being over-modeled. Data-centre operators have strong incentives to reduce copper intensity: aluminium substitution in some cabling, higher-voltage architectures that use less conductor per watt, and chip efficiency gains that slow power growth. If AI data-centre copper intensity falls materially below the roughly two-tonnes-per-megawatt baseline for standard facilities, the incremental demand leg shrinks faster than current forecasts allow.
Second, China's advanced-manufacturing pivot may not fully offset the property drag. Steel consumption is still heavily weighted to construction, and stimulus has so far targeted existing inventory rather than new building. If Chinese copper demand growth by 2050 comes in below roughly 40% — versus BHP's implied near-70% global growth — the deficit narrative weakens materially, and the miners repricing on structural scarcity would be exposed as a cyclical trade.
Third, the price itself is a risk to demand. At US$6.68 a pound, copper is near its all-time high of US$6.85 set in September 2026. Sustained prices at that level destroy demand — customers substitute, delay projects, or redesign products. BHP's own planning assumption of US$5.40 acknowledges this: the company is effectively saying the market's current price embeds a premium it is not willing to underwrite. That is prudent capital allocation, but it also caps the upside case that equity investors are paying for.
The falsifying signal is specific and observable. If global copper demand in 2050 lands below roughly 45 million tonnes — less than a third above today's ~34 million — or if data-centre copper demand fails to reach the ~3 million-tonne mark by mid-century, the structural-deficit thesis is wrong and today's miner valuations are pricing a regime shift that never arrives. Watch those two numbers; they are the load-bearing walls of the bull case.
Outlook: Three Horizons, Three Trades
The near-term, medium-term and long-term readings point in different directions, and that tension is the investable insight.
Short term, over the next six to twelve months, copper is vulnerable to a pullback. Prices near record highs, a strong US dollar and any disappointment in Chinese stimulus can trigger profit-taking. BHP's FY2027 production step-down is a reminder that operational execution can disappoint even in a bull market. Beneficiaries in this window are limited; cash-rich producers with low-cost assets can weather volatility, but the multiple is already stretched.
Medium term, over two to five years, the deficit should widen. Supply additions are limited, and demand from grid investment and data centres compounds. Beneficiaries are producers with low-cost, long-life assets and expansion optionality — BHP's Escondida, Spence and Olympic Dam cluster fits that profile, as do peers with similar asset quality. Exposed are consumers without pricing power: cable makers, appliance manufacturers and utilities facing capex inflation that regulators may not fully reimburse.
Long term, over five to twenty-five years, this is a structural regime shift if, and only if, the three demand vectors hold. The beneficiaries extend beyond miners to the entire electrification supply chain — grid equipment, power electronics and the engineers who build it. The exposed are commodity producers locked into single-commodity, single-market models: thermal coal is the obvious casualty, and pure-play iron-ore names face a slower growth path unless they add copper or energy-transition metals.
Three scenarios frame the path:
- Base case: copper averages US$5.50 to US$6.50 a pound through 2030; BHP delivers roughly 40% growth in attributable copper production; the deficit remains manageable but persistent, and the company's US$5.40 planning price proves conservative.
- Upside case: AI data-centre buildout accelerates beyond current forecasts and grid investment surges; copper tests US$7 and above, and BHP's disciplined planning assumptions leave significant money on the table.
- Downside case: AI capex slows, China stimulus fails to transmit, and aluminium substitution gains share; copper falls back toward US$4.50, miner margins compress and the organic-growth plan is delayed rather than accelerated.
BHP is not betting that copper stays at record prices. It is betting that the world is rewiring itself — and that rewiring, whether for a heat pump, an electric vehicle or a data centre, uses copper whether or not the cycle is friendly. The company that can build the tonnes fastest, at a planning price the market considers conservative, wins the era. The one that waits for the cycle to confirm it is already late.
Explore more exclusive insights at nextfin.ai.
