Amanda van Dyke drew an iceberg. Above the water she put 28 million tonnes of copper, the world's use for a year, and five pyramids to show the mass. Under the water she put waste rock, 2.8 to 4.3 billion tonnes of tailings, 190 to 720 million barrels of oil equivalent, and 1.5 billion cubic metres of water. The line at the bottom says the copper we see is only the tip. The footprint is the earth, the energy, and the water. The small print says these are orders of magnitude, for primary production only, and that they vary by mine, ore, process, and region.
Here is the idea, and it is the only one. The quote is the tip. The cost is the rock under it. You do not own a copper price. You own the work of moving that rock, at the grade, the water, and the permit of one mine. A world average on a poster is not that mine. This is not a recommendation to buy or sell any miner.
What the drawing actually says
Read the labels before the metaphor. The tip is split. About 22 million tonnes come from primary mining. About 6 million tonnes come from recycling. The underwater numbers are tied to the 22, not the 28. The footer says primary impacts only. If you hang the tailings math on all 28 million tonnes, you have charged a scrap furnace for a pit it did not dig.
The International Copper Study Group's later table is close, and not the same. For 2025 it shows mine output of 23.2 million tonnes, refined output of 28.7 million tonnes, and refined use of 28.2 million tonnes. Of the refined metal, 18.7 percent was secondary. That is 5.3 million tonnes from scrap, and 23.3 million tonnes of primary refined metal. The drawing rounds use to 28, primary to 22, and recycling to 6. The direction is right. The primary tonne on the drawing is a little low. Use 22 and the footprint is a little small. The Study Group's 23.3 million is about 6 percent larger than 22. The number 23 on its own is closer to 5 percent. Neither change turns billions into millions.
The pyramids are a shade heavy on purpose, or by rounding. The drawing says each Great Pyramid is about 5.8 to 6.0 million tonnes. Five of them are 29 to 30 million tonnes. The copper is 28. "About five" is fair. It is a picture of mass, not a survey.
The arithmetic under the water
The tailings number is the one that can be checked on a napkin. The drawing says 128 to 196 tonnes of tailings per tonne of primary copper. Multiply by 22 million. You get 2.82 billion tonnes at the low end and 4.31 billion at the high end. The label says 2.8 to 4.3 billion. The sum is honest to its own inputs.
The fuel number checks the same way. The drawing uses 51.7 to 200 gigajoules per tonne of copper, and about 6.1 gigajoules per barrel of oil equivalent. Divide, then multiply by 22 million tonnes. You get about 190 million barrels at the bottom and about 720 million at the top. The label matches. What it does not do is tell you which end of that band your mine sits on. A range of nearly four times is a warning, not a budget.
The water number checks too. The drawing uses 70.4 cubic metres per tonne. Times 22 million, that is about 1.55 billion cubic metres. The label says 1.5 billion. Then it says intensity varies widely by mine, process, and region. Believe the second sentence more than the first. An average of 70.4 can hide a plant that recirculates most of its water and a plant that does not.
The waste-rock label is the most careful thing on the page, because it refuses a single figure. It says billions of tonnes. It says the amount varies by mine type and strip ratio, from less than 1 tonne of waste per tonne of copper to more than 10,000. That spread is the investment. Two mines with the same copper tonnes can sit at opposite ends of it. One barely moves barren rock. The other moves a mountain to win a truck of metal. The iceberg drawing makes them look alike. The footnote says they are not.
Where the essay is wider than the picture
Her Substack essay, which the post points to, uses a related set of numbers and a few that do not match the drawing. She cites a USGS-led paper by Nassar and co-authors, written on 2018 mine data. On her reading, it put copper's rock-to-metal ratio near 513 tonnes of rock per tonne of copper, across hundreds of operations, with ratios from about 2 to about 17,000. The 513 is her reading. It was not copied from the paper's table for this piece. About two-thirds of that rock, she says, was waste. About one-third was milled and became tailings. A third of 513 is about 170 tonnes milled. That sits inside the drawing's 128 to 196, if her reading of the paper is right. Treat 513 as an average she assigns to 2018, and the drawing as a band.
They do not live together on water. The essay says milling used about 5.3 billion tonnes of water in 2016, near 260 tonnes of water per tonne of copper, and that some of it is recycled. The drawing says 70.4 cubic metres, and 1.5 billion for the year. Those are different cuts. One is closer to gross water through a mill. One is closer to a footprint after you notice that water goes around again. Chile, in her essay, makes the point. In the Atacama, most of the water in the circuit is recirculated. Mining is also a large share of the country's desalination. Recycled water is not free. It is pumps, power, and pipe. It is also not the same as 260 tonnes of new water taken from a dry place. If you add 70.4 and 260, you have counted the same circuit twice.
Energy has the same problem of width. The drawing runs from 51.7 to 200 gigajoules. The essay cites Rötzer and Schmidt for about 69 gigajoules per tonne at a 0.7 percent grade, about 89 if the grade falls to 0.5 percent, and about 97 if the pits get deeper and the strip rises. Those three sit inside the drawing's band. They do not justify the top of it. A slide that quotes 720 million barrels is using the wide end. A slide that quotes 69 gigajoules is using one study at one grade. Say which one you mean.
The average is not your ore
This is the step the poster cannot take for you. A ratio of 513, or a band of 128 to 196 tonnes of tailings, is a world sketch. Your shares are a deposit. Grade is the hinge. Her essay says copper grades have fallen hard over a century, from well above 1 percent toward well under 1 percent. At a lower grade, the same tonne of finished metal requires more rock. The tip of the iceberg can stay 28 million tonnes. The mass under the water grows. That is a cost. It arrives as diesel, steel for the mill, a bigger tailings dam, a longer haul, and a permit fight over land and water. It does not arrive as a footnote on the copper price.
Do the small sums she does, and then throw them out if they are not your mine. She says an electric car's 83 kilograms of copper is about 43 tonnes of rock, at the world ratio. A megawatt of solar, at 2.2 tonnes of copper, is more than 1,100 tonnes of rock. A data centre she puts at 50,000 tonnes of copper is about 26 million tonnes of rock, for the copper alone. Those lines are 513 times the metal, rounded. They are true only as a sketch of the average mine in that study. A high-grade underground mine can be a small multiple. A low-grade open pit can be a number that looks like a typo. The drawing's own waste range, from under 1 to over 10,000, is the honest range. The pyramid is the sales picture. The range is the underwriting.
Ask for three numbers before you trust a copper tonne in a model. How many tonnes of rock, ore plus waste, does this mine move per tonne of copper it actually sells. How much of the water is new water, not water going around the circuit again. What is the energy per tonne at the grade in year ten, not the grade in the first lucky bench. If the slide shows only copper tonnes, it is showing you the tip.
The tip is mostly not a gadget
The drawing does not say where the 28 million tonnes go. The essay does. In 2024, on the split she uses, building construction was 26 percent. Consumer goods, cooling, and electronics were 23 percent. Power and telecom were 17 percent. Transport was 13 percent. Industrial equipment was 12 percent. Other uses were 9 percent. Most of the metal is in buildings, wires, and machines. A car and a phone are real. They are not the bulk of the tip.
That matters because the rock follows the use. She cites S&P Global for a rise from about 28 million tonnes toward 42 million by 2040, and for a possible gap near 10 million tonnes if new supply does not arrive. She cites the same work for power networks at 4.3 million tonnes a year now and 7.1 million by 2040, more copper than electric vehicles use today. Treat those as one firm's path, not as fate. The investor point does not need the path to be exact. If more of the tip is grid copper, then more of the underwater mass is tied to power lines, not to a handset cycle. A forecast that counts devices and forgets the system will undercount the rock.
Recycling is the part of the tip that does not ask for a new pit. Six million tonnes on the drawing, or 5.3 million in the Study Group table, is metal that has already been mined once. It still takes power and a plant. It does not take 128 tonnes of fresh tailings. A demand story that treats all 28 million tonnes as if they must be dug this year overstates the quarry. A supply story that treats scrap as if it can replace the pit overstates the furnace. The drawing keeps them apart. Keep them apart in the model.
The footprint has an address
A global iceberg also hides the map. Her essay, citing the USGS summary for 2026, puts mined copper at about 5.3 million tonnes in Chile, 3.2 million in Congo, and 2.7 million in Peru. That is about half the world's mined copper, in three countries. She puts China's refined output near 14 million tonnes out of about 29 million. You can own a rock story and still not own the finished tonne, if the smelter is somewhere else and the permit to move the rock is local.
The disruptions she lists are the same idea in a bad year. She says 2025 took about 1.5 million tonnes of mined copper off the market, more than 6 percent of mined supply. That 1.5 million is her figure. It was not rebuilt mine by mine here. What is on the record is narrower. Benchmark Mineral Intelligence put the Grasberg loss near 590,000 tonnes from September 2025 through the end of 2026, after a mud rush. Kamoa-Kakula's flood was real. It cut that complex's guidance for 2025. Those are not one combined hole of 600,000 tonnes in a single year. The tonnes that failed were not lost because the world ran out of the element. A pit, a cave-in, a flood, or a grade did not deliver the tip on schedule. The underwater mass is not a metaphor when the mill stops. It is a hole that does not ship.
Time sits under the water too. She says the average path from discovery to first production is about 17 years. S&P Global has put a recent average near 18 years for mines that started in 2020 to 2023. That sample is several metals, not copper alone. A data centre that wants copper in three years is not going to summon a new iceberg. It will bid for the tip that already exists, or for the mines that are already moving rock. That bid is a price. The rock those mines still have to move is a cost. The spread between them is the investment. If the cost of moving the rock rises faster than the price of the tip, the copper quote can look strong while the mine does not.
What you can underwrite
You can underwrite a mine whose rock-to-metal ratio you have read, at the grade it will mine, with a tailings plan and a water plan that match the tonnes. High grade is not a slogan here. It is fewer tonnes of rock per tonne of metal, which is a smaller iceberg. Low grade can still work, if the strip, the power, the water, and the dam are in the cost, and the permit is real. The drawing's range exists so you will not skip that page.
You can underwrite the primary tonne and the scrap tonne as different claims. The 22 million, or the Study Group's 23.3 million of primary refined metal, is a claim on pits. The 5 to 6 million from scrap is a claim on collection, furnaces, and policy. A company that does one is not a company that does the other.
You can underwrite energy and water as line items that get worse when grade falls. Her cited path, from 69 gigajoules toward 90 and higher, is a direction, not a contract. The drawing's top end, 200 gigajoules, is a stress case unless the mine study gets there on its own numbers. Use the mine's number. Use the world band only to see whether the mine's number is even on the page.
What would make this reading wrong
The reading is wrong if your mine is at the low end of the range and you price it as if it were the average. A ratio near 2 and a ratio near 10,000 should not share a multiple. The poster is a world sketch from Critical Minerals Hub. It says so. An investor who buys "copper" because the iceberg looks scary has bought a drawing.
The reading is wrong if recycling grows enough to shrink the primary 22 million tonnes. Then the underwater mass shrinks with it, even if total use holds at 28. Watch the Study Group's secondary share. It was 18.7 percent of refined output in 2025, up from 16.4 percent in 2022. That is a real drift. It is not yet a replacement for the pit.
The reading is wrong if you treat 1.5 billion cubic metres as new water taken from rivers. The drawing does not say that. Her essay's 260 tonnes includes water that plants reuse. A desert mine that desalts and recirculates is a different risk from a mine that does not. The risk is still real. It is pipes, power, and politics. It is not a single global bucket.
The reading is wrong if the grade fall stops, or reverses, because the next mines are richer than the ones closing. Then the iceberg stops growing. She argues the long trend has gone the other way. One rich mine does not end a trend. It does mean you should underwrite the deposit in front of you, not the century.
The reading is also wrong if you use it as a reason to ignore price. The tip is what clears in the market. A shortage of finished metal can pay for a lot of waste rock. The error is to stop at the tip. The correction is not to ignore it.
The idea, once
The drawing puts 28 million tonnes of copper in the air and the quarry under the water. About 22 million of those tonnes, on the picture, are primary. They come with 2.8 to 4.3 billion tonnes of tailings, a waste pile the authors will not pin to one number, a fuel bill somewhere between 190 and 720 million barrels of oil equivalent, and about 1.5 billion cubic metres of water on one definition of water. The Study Group's own table says use was 28.2 million tonnes and primary refined metal was 23.3 million. The averages hide mines that barely move waste and mines that move 10,000 tonnes of it. The quote is the tip. The cost is the rock under it. Buy the work of moving that rock. Do not buy the pyramid.
A note on sources and limits
The iceberg labels are from the Critical Minerals Hub graphic posted by Amanda van Dyke with her note on 28 million tonnes of copper. Tailings of 2.8 to 4.3 billion tonnes follow from 128 to 196 tonnes of tailings per primary tonne and 22 million primary tonnes. The fuel band follows from 51.7 to 200 gigajoules per tonne and 6.1 gigajoules per barrel. The water total follows from 70.4 cubic metres per tonne. The Study Group table shows 2025 mine output of 23.2 million tonnes, refined output of 28.7 million, refined use of 28.2 million, primary refined of 23.3 million, and secondary refined of 5.3 million, which is 18.7 percent, up from 16.4 percent in 2022. Chile 5.3 million, Congo 3.2 million, and Peru 2.7 million are the USGS 2026 summary for 2025 mine output. The Grasberg loss near 590,000 tonnes, from September 2025 through 2026, is Benchmark Mineral Intelligence, as reported by Reuters. The lead time near 18 years is S&P Global, for mines started in 2020 to 2023 across several metals. The rock-to-metal figure near 513, the end-use split, the grade history, the 69 to 97 gigajoule estimates, the 260-tonne water figure, China's refined share, the 1.5 million tonnes of 2025 disruptions, and the S&P path to 42 million tonnes are from her essay "The Mineral Footprint of Copper." The 513 was not copied from the Nassar table for this piece. Nothing here is investment advice. The footnote on the picture is the point. Estimates vary by operation.