The product is refined copper — the reddish metal that carries electricity better than any other practical material. These companies dig copper ore out of the ground, crush and concentrate it, and sell either copper concentrate or finished copper cathode (cathode = 99.99% pure copper plates) to fabricators who turn it into wire, cable, busbars (the thick metal bars that carry heavy electrical current inside electrical gear), and transformers. It is needed wherever electricity moves: power grids, motors, buildings, electric vehicles, and — the reason it matters for an AGI build-out — data centers and the grid expansion required to power them. Per the scan, a single hyperscale data center uses roughly 4,000–20,000 tonnes of copper, and each gigawatt (GW = a unit of power capacity; one GW is roughly the output of a large power plant) of new capacity needs about 5,000–10,000 tonnes for power distribution alone, with grid upgrades roughly doubling that figure (source: 500-stocks scan, Copper Miners section).
On the numbers in the scan: the forecast additions to copper demand (AI plus grid, on the order of a few million tonnes a year by the end of the decade) arrive on a 2–5 year timeline, while any new mine takes 10–15 years to come online. So the demand side can change faster than the supply side, on these figures. The stated constraints on supply are physical — declining ore grades, water, permitting, and decade-long build times — rather than a shortage of capital or know-how, per the scan. In money terms, the large producers are capital-heavy commodity businesses whose cash flow rises and falls with the copper price, so what an owner pays today for a given amount of production depends heavily on where the copper price sits at purchase. This is a description of the mechanics, not a recommendation.
Copper is sold by weight — the standard unit is the tonne (1,000 kg), and the price is quoted in dollars per tonne (or per pound) on exchanges like the London Metal Exchange and COMEX. As a rough conversion, a price of "$4 per pound" is about $8,800 per tonne. est. The metal is close to a pure commodity: one producer's copper cathode is interchangeable with another's, so no single miner sets the price. The market price is the same for everyone; what differs between companies is their cost to produce a tonne.
That is the entire profit engine. A miner earns cash equal to (copper price − its cost per tonne) × tonnes sold. Because the selling price is set by the world market and is the same for all, the companies that earn the most per tonne are the ones that can dig it cheapest. Cost depends on ore grade (how much copper is in each tonne of rock — higher grade means less rock to move per tonne of metal), energy prices, labour, and how much waste rock has to be stripped away. Many copper mines also produce by-products — gold, silver, molybdenum, zinc — and the cash from selling those by-products is subtracted from the copper cost, which is why some mines report a very low "net" cost per tonne.
To turn this metal into shareholder cash, a producer must first spend heavily up front. Capex (capital expenditure — money spent building and maintaining mines, equipment, and processing plants) for a major new copper mine can run into the billions of dollars and take a decade before the first tonne is sold. After that, a running mine can generate large free cash flow (free cash flow = the cash left over after both operating costs and capex are paid) when prices are high — and very little, or negative, when prices are low.
Source: 500-stocks scan, Copper Miners "How It Works" + general knowledge on commodity pricing (pricing figures approximate, not live-verified).
Global refined copper demand currently runs at roughly 26 million tonnes a year est., driven historically by construction, electrical grids, vehicles, and consumer goods. The buyers are not end-consumers but industrial fabricators and equipment makers: wire and cable mills, transformer and motor manufacturers, electrical equipment firms, automakers, and grid utilities.
The AGI lens (forecast). Given that AI compute demand is rising sharply, the new demand layer is data centers and the electrical grid built to feed them. The scan's figures: a single hyperscale data center uses 4,000–20,000 tonnes of copper; the US alone may need 50–80+ GW of new data center capacity, each GW needing ~5,000–10,000 tonnes for internal power distribution; and grid expansion to deliver that power roughly doubles the copper requirement again. Put together, the scan estimates incremental AI-driven copper demand could reach 2–4 million tonnes per year by 2028–2030 est.. Against a ~26M-tonne base, that arithmetic works out to roughly an 8–15% addition est. layered on top of normal growth (source: 500-stocks scan, Copper Miners "Supply / Demand"). These are forecasts, not contracted orders — copper is sold on the open market at the prevailing daily price, so there is no fixed multi-year backlog of data-center copper the way there is for, say, jet engines.
Why copper specifically: it is the most conductive practical metal, and AGI infrastructure is fundamentally electrical — moving large amounts of power into and around buildings full of chips. More compute does not just need more code; it needs more physical electrical infrastructure, and that is copper-intensive. One factual caveat (covered in Supply, below) is substitution: when copper gets expensive, aluminum replaces it in some power cables and busbars, which limits how high copper demand can run before buyers switch.
✓ VERIFIED — the following figures were confirmed from primary sources after initial publication:
Global mine supply is roughly in line with the ~26M-tonne demand figure est. — the market is large but historically grows only a few percent a year. The binding constraint, per the scan, is physical: a new copper mine takes 10–15 years from discovery to production. You cannot order more copper the way you order more chips; you have to find an orebody (a body of ore-bearing rock), permit it (often against community and environmental opposition), secure water and power, and build it over a decade.
The stated bottlenecks, in plain terms:
Who controls supply. Copper mining is spread across a handful of large producers plus state-owned giants. Among US-listed pure-plays (a pure-play = a company whose business is almost entirely copper), the scan names Freeport-McMoRan (FCX) as the largest US copper pure-play and Southern Copper (SCCO) as the lowest-cost major; the global picture also includes diversified miners (BHP, Rio Tinto, Glencore) for whom copper is one of several metals, and Chile's state-owned Codelco. est. No single Western company holds a dominant share of global output, on these figures.
Source: 500-stocks scan, Copper Miners "Supply / Demand" + Aluminum section; producer landscape from general knowledge (not live-verified).
Putting the two sides together: on the scan's numbers, the forecast demand additions arrive on a 2–5 year timeline while any new supply takes 10–15 years, so demand can grow faster than supply can respond over the AGI build-out period. The evidence the scan points to is qualitative — long mine lead times, declining grades, no quick supply response — rather than a single hard number like a backlog. Because copper sells at the prevailing daily market price, the clearest live signal an owner can watch is the copper price itself and producer profit margins: a rising, sustained price combined with few new-mine starts is the market pricing tightness.
| Factor | Demand side | Supply side |
| Current size est. | ~26M t/yr | ~26M t/yr (roughly balanced today) |
| Forward change | +2–4M t/yr AI/grid by 2028–30 (forecast), plus normal growth | Slow; new mines need 10–15 yrs |
| Speed of response | Fast (2–5 yrs to build data centers/grid) | Very slow (decade+) |
| Factual limit | Aluminum substitution at high prices | Ore grade, water, permitting, capital |
| Net direction (per scan) | Demand can rise faster than supply can match on these figures | |
When could it flip to oversupply? Two things would have to happen, and neither is fast. First, the wave of mines being permitted and built now would have to come online (the decade-long lead time cuts both ways — today's prices are incentivizing projects that arrive in the 2030s). Second, demand growth would have to slow — for example, if the AI build-out plateaus, or if high copper prices trigger heavy aluminum substitution. Separately, a demand shock (a recession) can pull the price down quickly even while the structural picture stays tight, because copper is highly cyclical. So "supply slow to respond over the decade" and "the price can still fall hard in any given year" are both consistent with the mechanics described. This is a neutral description of how the cycle works, not a forecast of timing.
Source: 500-stocks scan, Copper Miners section; cyclicality and substitution dynamics from general knowledge (not live-verified).
The four key US-listed tickers in this group, plus the diversified miners and ETF the scan flags for context. "Copper exposure" means how much of the company's value rides on the copper price specifically. (Note: the 500-stocks scan lists this ticker as "EROS"; the matching NYSE-listed producer is Ero Copper, ticker ERO, used here.)
| Ticker | What it makes | Copper exposure | Rough size est. | Position / characteristics (per scan) |
| FCX (Freeport-McMoRan) | Copper (+ gold, molybdenum by-products) | High — largest US copper pure-play | Large-cap (tens of $B) | Biggest US-listed copper exposure; major mines in US, Indonesia (Grasberg), South America |
| SCCO (Southern Copper) | Copper (+ silver, molybdenum, zinc) | High — copper pure-play | Large-cap (tens of $B) | Flagged as lowest-cost major (per scan); long-life reserves in Peru/Mexico; majority-owned by Grupo México |
| TECK (Teck Resources) | Copper + zinc (exited steelmaking coal) | Medium-high — copper-focused but diversified | Mid/large-cap | Shifting portfolio toward copper; Canadian/South American assets |
| ERO (Ero Copper) | Copper (+ gold by-product) | High — small pure-play | Small-cap | Brazil-focused producer; smaller and more leveraged to the copper price than the majors |
| HBM (Hudbay), TGB (Taseko) | Copper + other base metals | Medium-high | Small-cap | Smaller North/South American producers (scan-listed) |
| BHP, RIO, GLNCY | Diversified (copper is one of many metals) | Low-medium — copper is a slice | Mega-cap | Broad metals exposure; BHP/RIO noted as tilting toward copper (per scan) |
| CPER (ETF) | Copper futures, not a company | Direct price exposure, no operations | Fund | Tracks the copper price directly; no mining, no by-product cash |
Source: 500-stocks scan, Copper Miners "Key US-Listed Companies" + Diversified Miners section. Sizes are rough order-of-magnitude from general knowledge; confirm live market caps before acting.
Because live quotes were unavailable, this section gives the shape of what an owner is buying rather than exact current multiples — confirm the live numbers before acting. Two facts about the money-in/money-out structure dominate:
Money-in / money-out summary. Money goes in as large, lumpy capex spent years ahead of any return. Money comes out as free cash flow that is feast-or-famine with the copper price, historically returned partly as dividends and buybacks by the larger, lower-cost names (FCX, SCCO) and reinvested by those still building (TECK, ERO). The diversified majors (BHP, RIO) carry other commodities, so their cash is steadier but their copper exposure is a smaller share of the whole. These are leveraged plays on the copper price filtered through each company's cost structure. This is neutral arithmetic about structure, not a verdict on value, and no current multiples are stated because they were not live-verifiable.
Source: structure from general knowledge of mining economics + 500-stocks scan cost notes; specific multiples deliberately omitted because they were not live-verifiable.
Where a company-level deep-dive would be most informative, stated factually:
Source: 500-stocks scan, Copper Miners section. This is a map of where to look, not a recommendation.
What was used:
/Users/ravf/projects/work/.claude/worktrees/sector-hub/research/investments/500-stocks/07-mining-materials.html) — the company list (FCX, SCCO, TECK, HBM, EROS/ERO, TGB, CPER), the qualitative supply/demand notes, the per-data-center and per-GW copper intensity figures, the 2–4M-tonne AI demand forecast, the ~26M-tonne market size, the 10–15-year mine lead time, the ~40% Chile+Peru share, and the cost-leadership/pure-play characterizations of SCCO and FCX. Also the adjacent Aluminum and Diversified Miners sections for substitution and the BHP/RIO/GLNCY context.Hard vs approximate:
Confidence: The direction of the demand-vs-supply mechanics described here (demand able to rise faster than slow-responding supply) follows from the scan's figures and basic mining economics. Exact magnitudes and all current financial figures should be re-pulled from the latest filings and a live quote feed before any decision. No price target, no buy/sell, and no value judgment is offered here.