The Future Uses of Copper: AI Data Centres, EVs & the Grid

Last updated: July 20, 2026

The Future Uses of Copper

Copper is trading at prices no one has ever paid before. In mid-July 2026 the London Metal Exchange cash price sits around $13,500 per tonne, after blowing through its December 2025 record of $11,771 and setting fresh highs through the first half of the year. The rally is not a speculative fever. The world’s second-largest copper mine, Grasberg in Indonesia, has been under force majeure since a deadly mud rush in September 2025. Chile, source of roughly a quarter of the world’s mined copper, produced less in February 2026 than in any month for nine years. And on the demand side, three of the biggest construction projects in human history — the AI data-centre buildout, the electrification of transport, and an expansion of the electric grid the IEA sizes at 80 million kilometres of new lines — all need the same metal at the same time.

S&P Global’s January 2026 study put the arithmetic plainly: global copper demand is on track to rise about 50%, from roughly 28 million tonnes a year today to around 42 million tonnes by 2040, while primary supply stalls and then shrinks — leaving a projected shortfall of about 10 million tonnes a year by 2040. Daniel Yergin, who chaired the study, calls copper “the great enabler of electrification.” What follows is where all that copper is going, why engineers keep choosing it over every alternative, and why the mines and smelters that produce it cannot simply ramp up to meet the moment.

Why Copper? The Economics of the Second-Best Conductor

Copper conducts electricity at about 59.6 million siemens per metre — roughly 94% as well as silver, the best conductor that exists — for somewhere between one-eightieth and one-hundredth of silver’s price per kilogram. That single ratio explains copper’s grip on the modern world. Silver wins on physics; aluminium wins on price; copper wins on physics-per-dollar, and it wins by such a margin that engineers only leave it under duress. Copper also carries heat at about 401 watts per metre-kelvin, second among engineering metals only to silver, and it can be drawn into wire finer than a human hair without snapping — a ductility aluminium cannot match.

The substitution question therefore runs in two directions, and copper sits at the equilibrium point. Move up to silver and the performance gain is a few percent while the cost rises a hundredfold, which is why silver only replaces copper in places like solar-cell contacts where micrograms matter. Move down to aluminium and the price falls, but conductivity drops to about 61% of copper’s, so every conductor has to get roughly 60% fatter to carry the same current — and aluminium joints creep, oxidise, and loosen over time, which is why aluminium branch wiring installed in North American homes in the 1960s and 70s became a documented fire hazard. Aluminium earns its place in overhead transmission lines, where weight rules. Nearly everywhere else, substitution means accepting a worse conductor, a bulkier cable, or a riskier connection.

Copper in the Electric Grid: The 80-Million-Kilometre Buildout

The electric grid is the largest machine humanity has ever built, and the IEA says the world now has to build it again. Meeting national climate and energy-security targets requires adding or refurbishing 80 million kilometres of power lines by 2040 — the equivalent of the entire existing global grid — and lifting annual grid investment from around $300 billion, where it stagnated for a decade, to more than $600 billion a year by 2030. Goldman Sachs Research expects grid and power infrastructure to drive more than 60% of all copper demand growth through 2030. No future use of copper is bigger, and none is less optional: at least 3,000 gigawatts of renewable projects — five times the solar and wind capacity added in 2022 — are sitting in connection queues today because the wires to carry their power do not exist yet.

The honest nuance is that the famous high-voltage lines you see crossing the countryside are mostly not copper. Overhead transmission uses aluminium conductors on steel cores, because a cable strung between towers has to be light above all else. Copper’s territory is everything the aluminium line plugs into: the transformer windings at either end, the substations, the switchgear, the underground and submarine cables where compactness beats weight, and the distribution network that carries power the last kilometres into homes, factories, and data centres. S&P Global projects transmission and distribution alone will consume about 7.1 million tonnes of copper a year by 2040 — more than the combined output of Chile’s entire mining industry today.

Grid hardware is also where the buildout physically bottlenecks. Large power transformers — each wound with tonnes of copper magnet wire — have waiting lists measured in years, and the IEA’s 2025 transmission-grid report flags cables and transformers as the supply chains most likely to delay the whole expansion. A transmission line takes five to fifteen years to plan, permit, and build; the solar farm it serves takes one to five. The wires, not the generators, set the pace of the energy transition — and the wires are made of exactly the metal the market is short of.

Copper in AI Data Centres: 27 Tonnes per Megawatt

When Microsoft built a $500 million data centre in Chicago, it went in with about 2,177 tonnes of copper — roughly 27 tonnes for every megawatt of capacity, a figure BHP uses as its benchmark for the industry. Multiply that intensity across the AI boom and the numbers turn geological: S&P Global expects AI and data-centre capacity to reach around 550 gigawatts by 2040, more than five times 2022 levels, and BHP projects data centres’ copper appetite will grow from about half a million tonnes a year today to around 3 million tonnes a year by 2050 — roughly the combined annual output of the world’s four largest copper mines.

The copper starts at the property line and runs all the way into the chip. Medium-voltage feeders bring grid power in; transformers step it down through copper windings; busbars and power-distribution units route it across the halls; uninterruptible power supplies and backup systems hold it steady; and inside every rack, copper carries the last metre of power and much of the data. Liquid-cooling loops — increasingly mandatory as AI chips pass a kilowatt each — add copper cold plates and manifolds on top. A data centre is best understood as a building-sized machine for converting copper-delivered electricity into computation, with a gas turbine’s worth of heat to remove.

Chip-to-chip copper is a story of its own. The industry expected optics to take over inside the rack years ago, but direct-attach copper keeps winning at short range because a passive copper cable needs no laser, no transceiver, and almost no power. The same logic that wired the grid wires the rack: copper is simply the cheapest way to move electrons reliably. Gold still handles the corrosion-proof connector plating — a job covered in our article on the future uses of gold — but the bulk conductor everywhere in the data hall is copper. In the United States, data centres already consume about 5% of electricity, a share S&P Global sees hitting 14% by 2030; every percentage point of that growth is copper-delivered.

Copper in Electric Vehicles & Charging Networks

A battery-electric car contains roughly 83 kilograms of copper — close to four times the 20-odd kilograms in a comparable petrol car, according to teardown analyses by UBS and the International Copper Association. The metal is spread through the machine: more than a mile of hair-fine magnet wire in the motor stator, thick cables linking battery to inverter to wheels, foil current collectors inside the battery cells themselves, and the low-voltage harness that runs every sensor and screen. An articulated electric bus can carry up to about 370 kilograms. With around 22 million EVs sold worldwide in 2025 — each using nearly three times the copper of the car it replaced, on S&P Global’s figures — transport has become one of the fastest-compounding sources of new copper demand.

The vehicles are only half the transport story. Every EV needs somewhere to charge, and charging hardware is dense with copper — from the coiled cable in a home wallbox to the liquid-cooled conductors inside a 350-kilowatt highway fast charger and the beefed-up local grid connections behind them. Analysis for the International Copper Association has projected charging infrastructure alone multiplying its copper draw several times over as networks scale through 2030. The pattern repeats at every level: electrify a thing, and its copper content jumps; then electrify its supporting infrastructure, and the copper jumps again.

Copper in Solar Farms & Wind Turbines

Renewable power plants use several times more copper per megawatt than the fossil plants they replace, because their generation is spread across hundreds of machines and hectares that all have to be cabled together. Navigant Research figures put utility-scale solar at about 5 tonnes of copper per megawatt, onshore wind at about 4.3 tonnes, and offshore wind at about 9.6 tonnes — the offshore premium coming almost entirely from the long submarine export cables that bring the power ashore. Wood Mackenzie has projected the global wind fleet alone consuming more than 5.5 million tonnes of copper over the five years to 2028.

Inside a wind turbine, copper fills the generator windings, the transformer at the tower base, and the cabling that snakes down the tower and out through the collector network. A solar farm strings copper from every panel junction box through combiner boxes and inverters to the substation. None of this is exotic engineering — it is ordinary electrical construction repeated tens of thousands of times, which is precisely why the tonnage is so large and so hard to thrift. Panel-level silver can be shaved by the milligram; the cables connecting ten thousand panels across a square mile of desert cannot.

Copper Interconnects: The 1997 Chip Breakthrough Still Powering AI

Every advanced processor on Earth — every AI accelerator, every phone chip — is wired internally with copper, and the reason dates to a single announcement. In September 1997, IBM revealed it had solved a problem that had defeated the industry for a decade: how to deposit copper wiring onto silicon chips without the copper atoms poisoning the transistors. Its “damascene” process, which etches trenches and fills them with copper behind a thin barrier layer, replaced the aluminium interconnects chips had used since the 1960s. Copper’s lower resistance let signals move faster with less heat, and the technique scaled so well that it remains the standard interconnect process nearly three decades later.

The tonnage is tiny — interconnect layers are measured in nanometres — but the dependence is absolute, and it deepens as chips shrink. Billions of transistors on a modern GPU are linked by tens of kilometres of copper wiring thinner than a virus, and the industry’s roadmap keeps copper in the role for the foreseeable future because every candidate replacement conducts worse at practical dimensions. The chip that runs the AI model, the rack that houses it, the data centre that powers it, and the grid that feeds it: four nested layers of infrastructure, one metal running through all of them.

Copper in Humanoid Robots

A humanoid robot is a walking collection of electric motors — several dozen actuated joints, each with a stator wound in copper wire, plus a battery, power electronics, and a wiring harness tying it together. S&P Global’s 2026 study ran the scenario: a world with 1 billion humanoid robots in service would add about 1.6 million tonnes to annual copper demand, around 6% of everything currently consumed. Adamas Intelligence, modelling a more aggressive 10-billion-robot future, estimates the fleet would need roughly four times today’s entire annual copper production, out of some 600 million tonnes of total material inputs.

Per-unit copper figures are not publicly disclosed by any robot maker, so treat every number as an analyst estimate — but the direction is not in doubt. Tesla’s Optimus, Unitree’s models, and their dozens of competitors all rely on high-torque electric actuators, and there is no way to wind an electric motor at scale without copper magnet wire. The same rare-earth magnets, silver contacts, and copper windings that fill an EV drivetrain reappear, miniaturised and multiplied, in every robot joint. If humanoids reach even a fraction of the production volumes their manufacturers project, robotics becomes a copper demand category that did not exist in 2024.

Antimicrobial Copper on Hospital Touch Surfaces

Copper kills microbes on contact, and it is the only solid material with U.S. regulatory recognition for doing so. In 2008 the EPA registered copper alloys as antimicrobial, certifying that surfaces of copper, brass, and bronze kill more than 99.9% of certain infection-causing bacteria within two hours — a claim stainless steel and plastic cannot make, since many pathogens survive on those for days. Copper ions rupture microbial membranes and scramble their DNA, a mechanism so physical that bacteria struggle to evolve resistance to it.

Hospitals are the natural market. Trials replacing high-touch surfaces — bed rails, door handles, IV poles, overbed tables — with copper alloys have reported meaningful reductions in surface contamination, and interest revives with every drug-resistant-infection headline and every pandemic. Antimicrobial hardware will never rival the grid as a tonnage driver, but it illustrates the breadth of the metal’s franchise: the same element that carries the current also defends the ward.

Is There Enough Copper for the AI & Energy Buildout?

No — not on current plans, and the people who mine it say so themselves. S&P Global’s January 2026 analysis projects demand reaching about 42 million tonnes a year by 2040 against a supply base that peaks around 2030 near 33 million tonnes and then declines, leaving a shortfall of roughly 10 million tonnes a year — a quarter of projected demand — unless a wave of new mines, expansions, and recycling closes the gap. Industry analysts cited alongside Chile’s 2026 production slump put the potential gap at 7.7 million tonnes a year by 2034. Forecasts differ on timing: the International Copper Study Group expects the refined market to tip into a 150,000-tonne deficit in 2026, Benchmark Mineral Intelligence sees a steep 2026 deficit after the Grasberg disaster, while Goldman Sachs still models a small surplus before structural deficits set in from 2029. Nobody’s model shows comfortable abundance. The sections below trace the demand, the squeeze, and why supply is so slow to answer.

How Much Copper Each Industry Will Need

Copper’s demand base splits into an old economy and a new one, and the new one is doing all the growing. S&P Global’s sector-by-sector projection for 2040 shows traditional “Dr. Copper” demand — construction, machinery, appliances — grinding higher with GDP, while energy-transition and technology demand nearly doubles. Figures below are S&P Global’s, in million tonnes of copper per year.

Two features of that table deserve emphasis. First, the declining categories are small: internal-combustion vehicles and fossil power plants together release barely a million tonnes of annual demand by 2040 — no cavalry there. Second, the growth is not speculative tech demand that could evaporate in a downturn; it is wiring, transformers, and cables ordered by utilities under government mandates, the stickiest demand in the industrial economy.

The Copper Supply Squeeze

Supply is faltering at exactly the wrong moment. In September 2025 a mud rush killed workers and buried part of the underground operation at Grasberg in Indonesia — the world’s second-largest copper mine — sending roughly 800,000 tonnes of wet material into the workings. Freeport-McMoRan declared force majeure, and Benchmark estimates the disaster will remove about 591,000 tonnes of copper from the market between September 2025 and the end of 2026. Chile, the world’s largest producer, delivered its worst monthly output in nine years in February 2026 — 378,554 tonnes, its seventh consecutive year-on-year monthly decline — as ore grades sag across its ageing mines. Cobre Panamá, which supplied about 1.5% of the world’s mined copper, has sat idle since a court ruling shut it in November 2023.

Prices have responded the way a tightening market should. LME copper set a record $11,771 a tonne in December 2025, kept climbing through the first half of 2026, and traded around $13,500 in mid-July — with the US COMEX contract peaking near $6.67 a pound in June. The 2025 average COMEX price of about $4.80 a pound was already the highest on record, according to the USGS, before the 2026 leg higher.

Why Copper Mining Can’t Simply Ramp Up

The average copper mine that entered production in recent years was discovered about 17 years earlier. Between discovery and first cathode lie resource drilling, feasibility studies, permitting battles, financing, construction, and commissioning — a gauntlet that has lengthened, not shortened, as deposits get more remote and approvals more contested. A price signal sent today produces new mine supply in the 2040s. That lag, more than any geological shortage, is why every serious forecast shows a widening gap: the demand is arriving on a five-year schedule and the supply responds on a twenty-year one.

Grade decline compounds the lag. The copper content of mined ore has fallen for a century as the richest deposits were worked out; Chile’s flagship mines now move more rock, burn more energy, and consume more water per tonne of metal every year, and Glencore’s Collahuasi — one of the world’s great orebodies — lost 68,100 tonnes of output in 2025 largely to declining ore quality. Mining companies must run faster just to stand still: Codelco is spending more than $40 billion on structural projects mainly to hold output flat. Meanwhile the six countries responsible for two-thirds of mine production carry their own risks, from Chilean water politics to Congolese infrastructure to Indonesian geotechnics — as Grasberg just demonstrated.

The Copper Smelter Bottleneck

Even mined copper must pass through a second choke point: smelting. China built so much smelting capacity — around 40% of the world total — that smelters now outbid each other for scarce concentrate, driving spot treatment and refining charges (the fees smelters earn for processing ore) below zero in 2025. A negative TC/RC means smelters effectively pay miners for the privilege of processing their rock. The economics broke so badly that China’s smelter purchasing team agreed to cut production by more than 10% in 2026, and Beijing has halted roughly 2 million tonnes of planned new smelting capacity. The refining step, long treated as infinitely elastic, has become a bottleneck of its own — one reason refined-output growth collapses from 3.4% in 2025 to a forecast 0.9% in 2026 (ICSG).

Copper’s path from rock to wire shows where the system can seize:

1. Mining: ore averaging well under 1% copper is blasted and hauled — grades falling, costs rising.
2. Concentration: the ore is crushed and floated into ~25–30% copper concentrate, consuming enormous energy and water.
3. Smelting: concentrate becomes blister copper — the stage now starved, with negative processing fees and Chinese output cuts.
4. Refining: electrolysis yields 99.99% pure cathode, the form traded on the LME and COMEX.
5. Fabrication: cathode is drawn into wire rod, cable, tube, and foil — the products the 50% US tariff actually targets.

Recycling is the one supply lever that responds quickly, and it is already large: scrap already supplies about 30% of US copper consumption, and S&P Global projects recycled supply growing from around 4 million tonnes to 10 million tonnes a year by 2040. But recycling can only return copper the economy installed decades ago — and a world that is doubling its installed copper base cannot recycle its way out of mining.

Investor Demand & the Copper Bull Market

Traders call copper “Dr. Copper” — the metal with a PhD in economics — because its price historically diagnosed the health of world industry. The 2025–26 rally says something new: copper is decoupling from the old business cycle and repricing as the binding constraint on electrification. The clearest evidence came from Washington. In July 2025 the United States imposed a 50% Section 232 tariff on copper imports — and when the final order exempted refined cathode, covering only semi-finished products like pipe, wire, and rod, COMEX futures crashed nearly 20% in a single day, the sharpest one-day fall on record. Traders had shipped so much metal to America ahead of the deadline that US warehouses swelled to record inventories of roughly 650,000 tonnes, at one point pricing US copper 28% above London. A tariff designed for national security had, for a few months, rearranged the world’s copper map.

Investor access to copper looks different from precious metals. Nobody stacks copper bars in a home safe — at around $13,500 a tonne, even a modest position weighs more than a car — so exposure runs through markets instead. The main routes, in rough order of directness:

The structural case that draws capital is the same one running through this whole article: demand mandated by law and technology, supply gated by 17-year lead times, falling grades, smelter bottlenecks, and a market where Goldman Sachs projects prices reaching $15,000 a tonne by 2035 in nominal terms. The bear case deserves equal print: near-term forecasts disagree — Goldman sees surplus into 2026 where ICSG sees deficit — Chinese demand growth is decelerating, substitution and thrifting accelerate at high prices, and copper has punished extrapolation before. High prices are the market’s tool for closing gaps, and they work in both directions.

The Outlook for Copper Demand

Copper’s next fifteen years are unusually legible, because the demand is already ordered. The grid expansions are in national plans; the data centres are in hyperscaler capex budgets; the EV factories are built. S&P Global’s 42-million-tonne demand projection for 2040 requires no technological surprise — it is mostly wire, transformers, and cable for machines the world has already decided to build. The supply side is equally legible and far less comforting: production from existing mines peaks around 2030, the replacement pipeline is 17 years long, and the smelting system between mine and market has started rationing itself.

The wild cards sit at the edges. Humanoid robots could add a demand category the size of a major industry; aluminium substitution and miniaturisation could quietly shave points off demand growth, as they have before; a deep Chinese construction slump could mute the old economy’s pull for years. Recycling is the one certainty — a doubling of secondary supply by 2040 is baked into every forecast, and still isn’t enough. The most likely future is the one the price is already signalling: not a world without copper, but a world where copper is permanently more valuable than the one we grew up in.

A Short History of Copper: The First Metal’s Fourth Act

Future Uses of Copper FAQ

Related Articles