Everything Electric Needs Copper. The Mines Are Fifteen Years Behind.

You’ve saved $1.5 million. Now comes the real test.

Can it produce $10,000 a month, or will that pace drain your portfolio?

Most retirees do not get a clear answer until it is too late.

The issue is not just how much you have. It is whether your portfolio was built to pay you, not just grow.

That difference can determine whether your money lasts decades or starts breaking down early.

Sequence of returns, taxes on withdrawals, healthcare costs, and whether the 4% rule still applies all play a role.

Fiduciary advisors created a breakdown showing what drives sustainable income and why the same $1.5M can produce very different outcomes.

If you have $1M or more invested, do not guess.

See How Long My Money Lasts

Learn More

Electrification demand will outrun what the world's mine pipeline can physically deliver by 2040, because the supply response is governed by a 15-to-17-year lead time that no price signal can compress, and the resulting gap is wide enough to survive the failure of any single demand leg.

Copper stopped being a proxy for Chinese construction and became a contested national-security input, and two dates mark the change. On July 30, 2025, a presidential proclamation imposed a 50% Section 232 tariff on semi-finished copper and copper-intensive derivative products, effective August 1. In April 2026 that action was modified to apply the 50% rate to the full customs value of semis, 25% to derivatives, and 10% where the copper content was entirely smelted and cast in the United States.

What the 2025 proclamation deferred is the live question. Refined copper cathode, the traded form that feeds every wire mill and rod plant, was left untariffed, with a phased universal duty of 15% from January 1, 2027, rising to 30% from January 1, 2028, held in reserve. The President tied any decision to a Commerce Department update on domestic refining capacity due by June 30, 2026. That window has closed. A search of Federal Register presidential documents on July 24, 2026 returns no copper action of any kind after Proclamation 11032 of June 1, 2026, which adjusts the threshold at which a product qualifies as made entirely of American metal and never mentions refined copper or cathode. The determination is live, overdue against its own trigger date, and unannounced.

The second reason this is a July 2026 question is that the supply arithmetic was restated in January. S&P Global's study Copper in the Age of AI projects the call on copper growing from 28 million tonnes a year in 2025 to 42 million tonnes by 2040, an increase of 50%. Against that, absent significant investment, global primary supply could produce just 22 million tonnes by 2040, a million tonnes less than today, while total copper production peaks in 2030 at 33 million tonnes and falls away after. The same study finds AI data-center demand and defense demand each roughly tripling by 2040, adding a combined 4 million tonnes. In practical terms, the electrification bill that lands on household energy prices, vehicle prices, and utility rate cases over the next fifteen years is being written now, in the form of mine approvals that either happen or do not.

The tariff decision belongs in this report as a clock and nowhere else. Tested against five Section 232 copper events, the world copper price showed no measurable premium; the tariff moved the gap between US and world prices and left the underlying scarcity question exactly where it was. The structural gap never depended on the tariff and will not be resolved by it.

A copper mine is a fifteen-year construction project that starts with a geologist. S&P Global Market Intelligence's dataset of 127 mines that have come online since 2002 puts the average time from discovery to production at 15.7 years, measured across gold, copper, nickel, silver, and zinc. The most recent cohort took longer still: mines that began operating between 2020 and 2023 averaged 17.9 years. For copper specifically, S&P Global's January 2026 study is blunter, finding that it takes 17 years on average for a new copper mine to go from discovery to production.

The consequence shows up when the pipeline is run forward. In Benjamin Capital Research's 20,000-draw simulation of the announced pipeline, with each project discounted by that lead-time distribution, projects reaching a final investment decision from 2026 onward contribute a median of 0.6 million tonnes to 2040 supply. The 90th-percentile outcome is 1.3 million tonnes. The demand path implies roughly 1 million tonnes a year of net new supply. Everything the industry approves from this point forward covers, at the median, about seven months of that requirement, spread across fifteen years.

The upshot: the copper that will wire the 2040 grid had to be found before this decade started, and the price signal available today cannot change that.

Escondida in Chile is the largest copper mine in the world, and its own disclosures run one way. BHP reports Escondida's average concentrator grade, the copper content of the rock actually fed into the mill, at 0.90% for the year to June 2026, with the June quarter down at 0.85%. For the year to June 2027 it expects a feed grade of roughly 0.70%, a decline of about 22% in a single year. The pattern is national. Cochilco, Chile's state copper commission, puts the country's weighted-average ore grade at 0.94% in 2006 and 0.62% in 2024. At 0.62%, more than 99% of every tonne of rock moved, crushed, ground, and processed is waste.

Falling grade is a cost story before it is a volume story. Lower grades mean more energy, more water, more haulage, and more capital for each tonne of finished metal, which raises the price a new project needs in order to clear its investment hurdle. This is why the supply curve does not simply shift out when copper rallies: the marginal tonne is being produced from progressively worse rock, and the industry has to run harder every year to stand still.

What this means on the ground: even the copper already permitted and already operating delivers less metal per unit of effort each year, so a flat production number conceals a rising real cost of production.

Demand is arriving from several directions at once, and the vectors are additive. A battery electric vehicle uses about 83 kilograms of copper, roughly 183 pounds, against 23 kilograms in an internal-combustion car, on International Copper Association figures, which is more than three times the metal for the same journey. The International Energy Agency puts annual copper demand for electricity grids at 5 million tonnes in 2020, rising to 7.5 million tonnes by 2040 under stated policies and to nearly 10 million tonnes on its faster decarbonisation path. S&P Global expects AI data-center demand and defense demand each to roughly triple by 2040, adding 4 million tonnes between them. BHP projects global copper demand growing by around 70% to over 50 million tonnes a year by 2050, an average growth rate of 2% a year.

These uses resist substitution for a physical reason. Copper's electrical conductivity is second only to silver among industrial metals, and the applications driving the growth are exactly the ones where conductivity is the specification: motor windings, fine wiring, busbars (the thick copper bars that distribute power inside electrical equipment), transformers, and power distribution. Aluminum wins specific segments where weight and cost dominate. The segments where the metal's job is to carry current without heating up have stayed with copper.

The data-center leg is the loudest number in copper and the worst measured. NVIDIA's technical page on its high-voltage data-center architecture read, in the version archived on January 13, 2026, "The rack busbars alone in a single 1 gigawatt (GW) data center could require up to half a million tons of copper." The version archived on January 14 read "up to 200,000 kg of copper." The page's own modification timestamp records the edit; no correction notice accompanies it. Between those two sentences sits a factor of roughly 2,300. The discrepancy was first flagged in analysis by Thunder Said Energy, reported in the trade press at the time, which treated the original figure as a unit-conversion error. A number that large, in a market that consumes 28 million tonnes a year, is the difference between a rounding error and a crisis. The lesson survives the correction: published estimates of how much copper a data center consumes span two orders of magnitude before anyone disagrees about the forecast, because the scope boundary moves in silence between the racks, the building, and the grid connection that feeds it. No result in this report therefore depends on a copper-per-megawatt figure; every demand input is an aggregate published tonnage path measured in millions of tonnes.

The bottom line for investors: the demand case here rests on tonnage forecasts from S&P Global, the IEA, and BHP, and it would stand unchanged if every per-megawatt intensity estimate in circulation were withdrawn tomorrow.

Mined copper is useless until it is smelted and refined, and that step is concentrated. Refinery-production data in the USGS Mineral Commodity Summaries 2026 puts China at 12,400 of 27,600 thousand tonnes of world refined copper output in 2024, which is 44.9%, rising to an estimated 14,000 of 29,000 in 2025, or 48.3%. That is a move of 3.4 percentage points toward China in a single year. S&P Global's January 2026 study puts China at roughly 40% of global smelting capacity and about 66% of the world's imports of copper concentrate, the semi-processed ore that mines ship to smelters, meaning two-thirds of the concentrate that crosses a border is headed to a Chinese smelter.

That share has a floor, and it sits well above any realistic diversification target. Running the announced ex-China smelter and refinery pipeline forward, discounted by build lead times and with assumptions set generous to diversification, the probability that China's share falls below 40% by 2035 is 0.2%, and the probability it falls below 33% is zero. The median modeled 2035 share is 47.6%, with a 42.1% to 54.0% range and a 41% chance the share is higher in 2035 than it is today. The arithmetic behind that result is simple. Reaching 40% requires the share to fall 0.83 percentage points a year for a decade, in a series that just moved 3.4 points the other way. Achieving it by construction, in the absence of Chinese closures, requires the rest of the world to expand refining capacity by a factor of 1.4 while China stands still. Columbia University's Center on Global Energy Policy reports that a new copper smelter in the United States could cost up to 5 billion dollars, and that new smelters in China can be built up to five times cheaper, which is why the rest of the world has struggled to close the gap.

The refining concentration supports this thesis at a lower altitude than the geology does, and its limits are worth stating precisely. A 48% share is concentration risk, meaning the market has one dominant processor whose policy choices set terms for everyone else. It sits far below the roughly 91% share China holds in rare-earth refining, where a single decision can halt a supply chain outright. Copper's primary constraint is geological, and the refining layer raises the cost of routing around it.

For the portfolio, the practical consequence is that Western capacity additions are a decade-scale project, so any policy that tries to force refined-copper supply onshore operates against a build clock, and build clocks run in years.

The most common retail-facing copper argument is that visible exchange inventories are low and a price spike is therefore imminent. The relationship runs the other way in Benjamin Capital Research's own testing, a replication of its oil-inventory study run on 967 weeks of London Metal Exchange data from 2008 to 2026. Weeks when LME stocks sat more than one standard deviation below their own trend, which is the statistician's way of saying unusually low for the period, were followed by a 13-week median copper return of negative 1.4%, against positive 1.9% in all other weeks. That difference of 3.3 percentage points carries a p-value of 0.000, meaning the odds of a gap that size arising by chance are below one in a thousand. At 26 weeks the difference widens to 5.5 percentage points. Tighter stocks do not even produce reliably better outcomes in ranked order, with a correlation of just 0.07 between how tight stocks are and what copper does next, where 1.0 would mean a perfect relationship. The result is robust when the test is repeated on COMEX, the US copper futures market, and on stock-draw signals instead of levels.

True squeezes are the exception. Weeks where stocks sat more than two standard deviations below trend, of which there were twelve in eighteen years, did run 13.9 percentage points higher afterward. Those are positioning events of the 2021 type, the year a squeeze on the London market briefly dislocated prices, and they cannot be anticipated from the inventory level: loosening the threshold slightly, to 1.5 standard deviations, collapses the effect to zero with a p-value of 0.97.

The premise also fails on today's data. LME registered stocks stood at 306,500 tonnes on July 10, 2026, which is 1.61 standard deviations above their five-year trend, and COMEX stocks reached a record of roughly 652,000 tonnes on tariff front-running, importers racing metal into the country ahead of a possible duty, reported on June 26. Inventories are abundant on the day the shortage argument is loudest.

The second failing signal is the tariff event itself. Across all five Section 232 copper events, the global copper complex showed no measurable premium: LME cash moved less around the events than in random windows, with a median 21-day absolute move of 3.0% against 3.8% for random dates, and every p-value at 0.16 or above. Copper equities showed no individually meaningful abnormal return at the four clean events. What did move, violently, was the gap between the US COMEX price and the world LME price, which is where the cost of a tariff actually lands. That spread widened 1,147 dollars a tonne after the investigation order was signed, a move with a p-value of 0.008, and then collapsed by 2,679 dollars a tonne in a single session when the proclamation excluded refined cathode from the duty. The event trades the spread; the spread trades the fine print; the fine print whipsaws.

As of the July 10, 2026 close that spread stood at 448 dollars a tonne, 3.3% of the London price, against roughly 2,011 dollars a tonne implied by the pending 15% duty. The market is pricing a small fraction of a decision it has had a year to think about.

Translated into calendar risk: the tariff headline moves the one variable that has historically reversed inside a single session, while the variable that decides the decade, mine approvals, moves quarterly and in public.

Stage 1: Demand accelerates on several fronts at once. Electric vehicles, grid buildout and replacement, renewable generation, reshored manufacturing, defense procurement, and AI data centers all scale in the same window, and each is policy-supported, which loosens its link to the ordinary business cycle. Aggregate copper demand rises toward roughly 42 million tonnes by 2040 on S&P Global's path, about 50% above today. Because the drivers are legislated and capitalized independently of one another, a slowdown in any one of them leaves the others running.

Stage 2: Supply cannot answer on the same clock. A price signal today cannot conjure metal for fifteen years, because permitting, financing, construction, and ramp each consume years and none of them compress under urgency. The lead-time structure is fixed and lengthening, at 15.7 years on average across 127 mines and about 17 years for copper specifically. Near-term supply is therefore determined by mines already running, whose grades are falling, and the pipeline sanctioned from 2026 onward, the projects whose boards have committed the capital, contributes a median 0.6 million tonnes by 2040.

Stage 3: Grade decline raises the cost floor underneath the whole industry. Existing mines yield less metal per tonne of rock each year, so energy, water, and capital intensity per tonne of copper rise. That lifts the incentive price a new project needs to clear its hurdle rate, which slows the very approvals that would close the gap. The constraint feeds itself: worse rock raises the cost of the answer to the shortage.

Stage 4: The refining layer adds a policy chokepoint on top of the physical one. Concentrate must be smelted and refined, and China holds 48.3% of refined output and takes about 66% of traded concentrate. A buyer who wants refined copper from outside that system pays a premium or waits for Western capacity that takes years to build. If the phased Section 232 duty on refined cathode is imposed, US pricing splits from world pricing by the size of the duty, and the cost lands on American fabricators and their customers.

Stage 5: The gap gets paid through price and through friction. A structural deficit clears through a higher long-run incentive price and, at the margin, through demand rationing and substitution wherever conductivity allows it. The cost surfaces in vehicle prices, transformer lead times, utility capital plans, and data-center construction budgets. The weak link in this chain is timing: a structural gap says the pressure builds and leaves the quarter in which it is repriced undetermined, which is why this report makes no dated price call, and why the tested inventory result above functions as a discipline on the analysis.

The chain holds at every link when stressed in Benjamin Capital Research's testing. Remove the AI demand leg entirely and the gap runs 6.5 million tonnes. Let every supply component deliver at its generous maximum simultaneously, with the existing base holding flat, the full sanctioned wave persisting, the new wave landing at its ceiling, and recycling reaching 10 million tonnes, and the gap still runs 3.0 million tonnes. The probability that the 2040 gap is positive is 100% in every scenario tested.

The closest rhyme is US oil-import dependence building through the 1960s and into the 1970s. Demand outran domestic supply for years while the strategic vulnerability sat unpriced, and when the repricing came it arrived abruptly, through policy, and only once a catalyst forced the issue. The structural similarity is precise: a slow-building physical gap in a no-substitute input that markets treated as a background condition until it became the foreground.

Two of the three differences that matter make copper the harder case.

First, the supply response. Oil could be drilled on a several-year horizon, and US production responded to price within the length of a business cycle. Copper's 15-to-17-year mine lead time is far more rigid, so the interval between recognizing the problem and delivering the metal is roughly three times longer.

Second, the shape of the constraint. The 1970s shock was political, delivered by a decision to withhold supply. Copper's primary constraint is geological and has no author: ore grades fall because the good deposits were mined first. The policy layer, China's 48.3% refining share and the Section 232 process, sits on top of the geology and raises the cost of routing around it, which makes it a second-order problem.

Third, the precedent for how fast concentration unwinds. The most determined diversification effort on record followed China's 2010 rare-earth supply interruption against Japan: in Benjamin Capital Research's measurement, Japan's dependence on Chinese supply fell from roughly 90% to about 58% over some seven years, then plateaued, never approaching zero. Copper's refining share is currently moving in the opposite direction. The same pattern appears in our modeling of manufacturing relocation, where even at the fastest migration pace anyone has achieved, China still assembles more than half the world's smartphones until around 2032. Concentrated supply chains take a decade to move when everyone involved is trying.

The implication the precedent supports concerns mechanism, and it is silent on dates. Slow physical gaps in essential inputs get priced late, in one move, when a catalyst arrives, and the policy response that follows arrives faster than the tonnage does. Mobilization is announced in weeks. Metal shows up over a decade.

Everything in this section describes how asset classes have historically behaved in comparable macro environments. None of it is a recommendation, and no specific securities are named.

Equities. In past periods when a physically constrained industrial input repriced upward on a multi-year horizon, the value of low-cost reserves already in the ground has historically re-rated ahead of the physical deficit. The companies that consume the input have meanwhile absorbed margin pressure until they could pass costs through. Copper-leveraged miners and diversified majors with large copper reserves sit on one side of that split; smelting and refining businesses, cable and transformer manufacturers, and data-center electrical suppliers sit on the other. The relevant data point is BHP's projection of copper demand growing by around 70% to over 50 million tonnes a year by 2050. Equity markets have typically priced reserve value on a multi-year lead relative to the deficit itself.

Rates and fixed income. A sustained copper-cost passthrough functions as a mild, persistent goods-inflation impulse, since copper feeds construction, capital goods, and electrification capital expenditure, which is business investment in equipment and infrastructure. In similar cost-push environments, the historical pattern has been a firmer term premium, meaning the extra yield investors demand for holding longer-dated bonds, and a slower disinflation path than a services-led model would imply. The IEA's grid path, from 5 million tonnes of copper a year in 2020 to 7.5 million by 2040 under stated policies, is a durable capital-cost channel running over 12 to 36 months at a time.

Credit. Commodity-price strength has historically widened producer cash-flow cushions while compressing the margins of fabricators with limited passthrough power, and credit spreads in the sector have tended to price that divergence before the equity market does. The near-term marker cuts against the structural view and deserves stating: the International Copper Study Group's April 2026 forecast round projects a refined-copper surplus of about 96,000 tonnes in 2026 and 377,000 tonnes in 2027, having forecast a 150,000-tonne deficit for 2026 as recently as October 2025. The physical market is comfortable today.

FX and emerging markets. Copper-exporter terms of trade, what a country's exports buy relative to its imports, improve when the metal's trend price rises. The currencies and fiscal positions of Chile, Peru, the Democratic Republic of Congo, and Zambia have historically gained a structural tailwind in those periods, offset by resource-nationalism risk and by the same grade decline that constrains supply globally. Chile alone accounts for about 24% of world mine supply, 5,510 thousand tonnes of a world total near 23 million tonnes in 2024, and holds the largest reserves at about 180 million tonnes on USGS figures.

Commodities. Metals bound by long lead times have historically carried a structural scarcity premium relative to metals whose supply can respond within a cycle, expressed as a higher trend price and higher volatility around it. Benjamin Capital Research publishes no price target for copper and makes no call on the timing of any repricing. The measured basis for the structural view is the modeled 2040 gap: a median of 8.5 million tonnes with a 95% confidence interval of 3.8 to 13.2 million tonnes, and a 93% probability the gap exceeds 5 million tonnes.

High prices ration demand and accelerate the search for alternatives, and the alternatives exist. Aluminum already substitutes for copper in overhead transmission lines, in some building wire, and in parts of grid infrastructure where weight and cost outweigh conductivity. Manufacturers thin copper content per unit whenever the price justifies the engineering, and secondary supply from scrap scales with the price of the primary metal. A market that reprices upward for a decade would trigger all three responses at once, and the deficit that motivated them would shrink accordingly.

The bound on this argument is physical. Copper's conductivity advantage is a property of the element, and the applications carrying the demand growth are the ones where conductivity is the binding specification. Aluminum substitution has historically captured specific segments and stalled at the boundary of those segments. Recycling scales, but it is constrained by collection capacity and by the size of the available scrap pool, which is itself a function of copper installed decades ago. Prior copper scarcity scares in the 2000s and mid-2010s were partly resolved by price-induced supply and thrifting, which is the base rate this probability reflects, and it is why the reading here is partial mitigation of the gap.

Estimated probability counter-argument is correct: 35%

The 42-million-tonne demand path for 2040 assumes electric vehicle adoption continues, announced grid spending converts into delivered capital, and the AI data-center buildout is a durable capital cycle. Each is contestable. EV adoption curves have flattened in several markets, grid capital plans slip routinely against regulatory and supply-chain friction, and the AI buildout carries the signature of a capital-expenditure boom that could correct sharply. If the demand curve flattens, the deficit is pushed out by years and the urgency dissolves.

The AI and data-center leg is the largest single swing factor in Benjamin Capital Research's model, moving roughly 2.0 million tonnes of the 8.5-million-tonne base gap, with a 1.2 to 2.8 million tonne range depending on how the demand is split. It also decides the extreme tail: the probability of a gap above 10 million tonnes falls to 7% under a full AI reversal and rises to 50% under an AI boom. Whether a gap exists at all is settled independently of the AI leg. Freezing data-center copper demand at today's level leaves a median gap of 6.5 million tonnes with a 74% probability of exceeding 5 million tonnes, because electric vehicles, grid replacement, and reshored manufacturing carry the deficit without help. The probability below therefore prices the size and timing risk, since the existence half of this argument has been tested and refuted on the model's own data.

Estimated probability counter-argument is correct: 30%

The lead-time argument is built on greenfield projects, entirely new mines built where nothing exists today, and greenfield is the slowest way to add copper. Brownfield expansions, which enlarge mines that already operate, reuse permits, roads, power, and processing capacity, and they deliver in years instead of decades. Restarts of large idled operations can return meaningful tonnage on an even shorter clock, ramping production in the Democratic Republic of Congo and Indonesia has repeatedly beaten forecasts, and improved recovery technology raises output from ore already being mined. The near-term evidence sits with this argument. In April 2026 the International Copper Study Group revised its 2026 view from a 150,000-tonne deficit to a 96,000-tonne surplus and put 2027 at a 377,000-tonne surplus, which is what a faster-than-expected supply response looks like while it is happening.

This argument gets its best case in the everything-delivers scenario: the existing production base holds flat with no depletion, the full sanctioned project wave persists, the new wave lands at its maximum, and recycling reaches 10 million tonnes. That combination still leaves a median gap of 3.0 million tonnes, with a 16% probability the gap exceeds 5 million tonnes. Brownfield tonnage is finite and grade-constrained like everything else, and the 2040 wall is a greenfield problem: the argument mitigates the near-term balance while leaving the structural gap standing. What this probability prices is the chance that every generous supply assumption delivers simultaneously.

Estimated probability counter-argument is correct: 30%

Copper is running out of physics inside the machine. As signalling rates climb, loss and crosstalk rise with frequency, the reach of a passive copper cable shortens, and the power burned in the chips that compensate returns as heat; SemiEngineering's survey of the field puts it plainly, that in scale-up networks copper has growing reliability issues with higher frequencies. Light has no such limit. Co-packaged optics, optical links built directly into the switch silicon, reached commercial availability during 2026 in scale-out switching, the network layer that connects racks to each other. NVIDIA has said the same technology arrives in its scale-up switches, the layer inside the rack, in 2027 and 2028. The strongest version of this argument comes from a source with every incentive to defend copper demand. CRU Group, a metals consultancy, describes a structural and accelerating shift in the connectivity stack from copper toward fibre, running through two channels at once: new AI buildouts increasingly being designed around optics-based architectures from the outset, and retrofits in which copper interconnects within the scale-up layer are being displaced by optical alternatives. CRU expects copper twinax, the short copper cabling that links racks and switches, to grow through the end of this decade, then slow in the early 2030s, becoming increasingly confined to shorter-reach, lower-speed, or cost-constrained segments. If the AI data center is one of the legs holding up the copper thesis and it is replacing its copper with glass, the thesis has a hole in it.

The hole is bounded by physics on the other side. A data center uses copper for two unrelated jobs: moving information between chips, racks, and switches, and moving electricity through the substation and grid connection, transformers, switchgear, busbars, power distribution units, rack power cabling, and cooling equipment. Optical fiber competes for the first job only. No photonic interconnect delivers a hundred kilowatts to a rack down a strand of glass, and rack power densities climbing from single-digit kilowatts toward fifty, a hundred, and beyond land entirely on the power side of that ledger. CRU makes the same point against its own substitution finding, concluding that aggregate copper demand is likely to prove more resilient than a data-interconnect-only view would suggest, because power cables and grid infrastructure keep growing alongside the buildout.

The damage has been modeled with every parameter set generous to the argument. The interconnect share of data-center copper was allowed to run as high as 40%, optical displacement of that share as high as 90% by 2040, and the offsetting growth in power and grid copper that CRU expects was deliberately not credited, making the result an upper bound.

Optical substitution removes a median of 0.47 million tonnes from 2040 demand, with a 90% range of 0.22 to 0.87 million tonnes. Against a median gap of 8.5 million tonnes that is about 5%. The AI leg's own capital-expenditure reversal is the larger of the two channels at 2.0 million tonnes, and firing both at once still leaves a 6.4-million-tonne median deficit with a 100% probability the gap is positive. The argument is correct about the technology and bounded on magnitude: light is taking the data cables, and the gap is made of power cables, grid connections, electric vehicles, and everything else that has to be wired. For this to become the deciding factor, optical substitution would have to escape the data layer, which current physics forbids, or the AI leg would have to be a far larger share of 2040 demand than S&P models. The version of the worry worth watching is a sourced tonnage split showing the interconnect share of data-center copper materially above 40%, or evidence that high-voltage direct-current distribution is cutting power-side copper intensity fast enough to offset rising rack density.

Estimated probability counter-argument is correct: 15%

If mine approvals stay below roughly 1 million tonnes a year of new capacity through the next four quarters of company guidance, the thesis accelerates. If a wave of major greenfield approvals lands alongside a sourced tonnage split showing data-center interconnect copper materially above 40% of that leg, it is time to reassess.

United States Geological Survey, "Copper," Mineral Commodity Summaries 2026, National Minerals Information Center (refinery production, mine production, and reserves tables).

S&P Global, "Copper in the Age of AI: Challenges of Electrification," special report, January 2026.

S&P Global, press release, "'Substantial Shortfall' in Copper Supply Widens as the Race for AI and Growing Defense Spending Add to Accelerating Demand," January 8, 2026.

S&P Global, "The Future of Copper: Will the looming supply gap short-circuit the energy transition?", 2022.

S&P Global Market Intelligence, "Discovery to production averages 15.7 years for 127 mines."

S&P Global Market Intelligence, "From 6 years to 18 years: the increasing trend of mine lead times."

International Energy Agency, "The Role of Critical Minerals in Clean Energy Transitions," mineral requirements chapter.

International Copper Study Group, copper market forecast press releases, October 2025 and April 23, 2026 forecast rounds.

Proclamation 10962, "Adjusting Imports of Copper into the United States," signed July 30, 2025, published August 5, 2025, 90 FR 37727.

Proclamation 11021, "Strengthening Actions Taken To Adjust Imports of Aluminum, Steel, and Copper Into the United States," signed April 2, 2026, effective April 6, published April 9, 2026, 91 FR 18201.

Proclamation 11032, signed June 1, 2026, published June 4, 2026, 91 FR 34085.

United States International Trade Commission, Harmonized Tariff Schedule, Chapter 99 Subchapter III, U.S. Note 16 (the operative scope list for the Section 232 copper headings).

Congressional Research Service, "Section 232 National Security Tariffs on Copper Imports," IN12614.

Columbia University Center on Global Energy Policy, "Protecting Existing US and Allied Copper Smelting Capacity."

Cochilco (Comision Chilena del Cobre), Anuario de Estadisticas del Cobre y Otros Minerales, ore-grade tables.

BHP, Operational Review for the year ended 30 June 2026, released July 16, 2026 (Escondida concentrator grade and FY2027 guidance).

BHP Insights, "How copper will shape our future," September 30, 2024.

International Copper Association, E-Mobility factsheet (copper content of battery electric and internal-combustion vehicles), 2017.

International Institute for Sustainable Development, "Commodity Profile: Copper," February 2026.

London Metal Exchange copper cash and three-month settlement series and registered stock series, 2008 to 2026.

CRU Group, "The outlook of copper-based data cables in AI data centres," 2026.

NVIDIA, high-voltage direct-current data-center architecture developer page, versions archived January 13 and January 14, 2026 (Internet Archive); Thunder Said Energy analysis of data-center copper intensity, as reported January 13, 2026.

SemiEngineering, "All AI Data Center Interconnects Will Be Optical Within 5 Years."

Methodology. The 2040 gap figures come from a Monte Carlo simulation of the announced copper supply pipeline run across 20,000 draws, calibrated against S&P Global's published supply curve, with each project discounted by a lead-time slippage distribution fitted to the 127-mine dataset and existing-mine output carried net of grade-decline depletion. The simulation is deliberately generous to supply at every parameter, which is why its median gap of 8.5 million tonnes sits below the 9.9 million tonne shortfall S&P Global's 2022 study modelled for 2035 under its Rocky Road scenario. The refining-share work applies displacement arithmetic in share space and then simulates the announced ex-China smelter pipeline discounted by build lead times. The inventory result is a replication of Benjamin Capital Research's oil-inventory study on 967 weeks of LME data, comparing forward returns after weeks with detrended stock levels more than one standard deviation below trend against all other weeks, tested at 13 and 26 weeks and repeated on COMEX prices and on stock-draw signals. The tariff work is an event study of five Section 232 copper dates against random-date windows, measuring the London price, copper equities, and the COMEX-to-LME spread separately.

No result in this report uses a copper-per-megawatt data-center intensity figure. Published intensity estimates span two orders of magnitude on undisclosed scope alone, and the number space contains a documented unit-conversion error of roughly 2,300-fold that circulated through industry publications before correction. All demand inputs are aggregate published tonnage paths.

Where a figure carries a date, it is the date of the underlying observation, which in most cases precedes the date of this report. Section 232 status was verified against Federal Register presidential documents and White House presidential actions on July 24, 2026 and should be re-checked before any decision that depends on it. Whether the Commerce Department delivered its June 30 update is not determinable from public records, since the report goes to the President and carries no publication requirement.

This report is for informational and educational purposes only. It does not constitute investment advice, a recommendation, or a solicitation to buy or sell any security. All asset class commentary reflects historical patterns and educational analysis, not personal investment advice. Past performance does not guarantee future results. Readers should consult a qualified financial advisor before making investment decisions.