Copper at a Crossroads

Introduction

A Full-Chain Review of the Global Copper Market and a 3-12 Month Trading Framework

Core Judgment

Long-term supply constraints continue to support a structural copper bull case. China determines whether traditional demand is destroyed in the current cycle, while US AI infrastructure and global grid capital expenditure determine the medium- to long-term demand slope. At record prices, with crowded managed-money net length and divergent regional curves, the base strategy remains to avoid chasing outright longs and instead wait for structural confirmation after a pullback, renewed curve support, or low-cost upstream equities with reliable volume delivery.

Data cut-offs: market prices 2026-08-14; CFTC 2026-08-11; macro and company data through 2026-08-14; event updates through 2026-08-15

Research classification: institutional research and scenario analysis; not personalized investment advice

Method: primary official sources take priority; facts, inferences and scenario assumptions are labeled separately

Investment Conclusion and Reading Guide

The report is not designed merely to prove that copper faces a long-term shortfall. It addresses more tradable questions: how much scarcity premium is already embedded in price; which evidence could extend the move; what could allow short-cycle demand weakness to overwhelm the structural thesis; and whether the best expression lies in outright physical exposure, the forward curve, regional spreads or listed equities.

Note: LME figures are official bid-ask mid-points as of 2026-08-14. The supply gap is the 2035 project-pipeline scenario in the IEA's 2026 outlook, not a forecast for the next 12 months. [1][4]

Executive Call

Suggested Reading Order

1. Market Baseline and What Has Changed

Fact | As of 2026-08-14, the official LME cash bid-ask mid-price was approximately USD 14,424/t and the 3-month mid-price was approximately USD 14,216/t, implying cash backwardation of about USD 208/t. COMEX August-to-December 2026 contracts observed on the same date rose from approximately USD 6.589/lb to USD 6.710/lb, equivalent to contango of about USD 267/t. [1][2]

Sources: LME and CME Group; compiled by Mortise Capital. LME values are bid-ask mid-points. COMEX prices are converted at 1 tonne = 2,204.62262 lb. Differences in exchange, deliverable grade and maturity prevent direct arbitrage comparisons.

Inference | LME cash backwardation indicates a time premium for immediately deliverable metal, while COMEX contango reflects US inventory, financing and policy uncertainty. Their coexistence means that an aggregate global shortage is not the only explanation. A more precise description is a regional and temporal mismatch in deliverable metal.

Fact | CFTC COMEX copper-futures data for 2026-08-11 show managed-money longs of 93,911 contracts and shorts of 14,884, for a net long of 79,027 contracts. This represents approximately 896 kt of notional copper and about 26.6% of total open interest. [3]

Source: CFTC, Disaggregated Commitments of Traders - Futures Only, 2026-08-11.

2. The Full Value Chain: Where Is the True Bottleneck?

Source: Mortise Capital framework. TC/RC denotes concentrate treatment and refining charges; green arrows represent scrap-copper feedback.

The most common error in copper research is to jump directly from mine disruption to a shortage of refined copper. The correct sequence is to verify each stage: mine output -> concentrate availability -> smelter utilization and treatment charges -> refined output -> semifabrication and orders -> end demand -> scrap feedback. Every stage contains inventory, grade, geographic and timing lags.

Why Negative TC/RC Does Not Equal a Refined-Copper Shortage

Fact | The IEA reports that the annual copper-concentrate treatment-charge benchmark fell to USD 0/t in January 2026 and that spot charges have been negative since 2024. China also accounted for roughly half of global refined-copper supply in 2025. [5]

Inference | Negative TC/RC first indicates that smelting capacity exceeds concentrate availability and that mines have gained bargaining power over smelters. As long as smelters continue operating because of local policy, by-product revenue or fixed-cost considerations, refined output can still rise. TC/RC must therefore be assessed together with maintenance schedules, refined production and cathode inventories.

3. Supply: Structural Constraints Are Real, but Near-Term Supply Is Not Contracting in a Straight Line

3.1 Structural Supply Constraints

Fact | The IEA's Global Critical Minerals Outlook 2026 estimates that, under the existing project pipeline, the potential copper supply gap in 2035 is approximately 25% of demand. This is narrower than the previous estimate of roughly 30%, partly because projects in the Democratic Republic of the Congo and Zambia have advanced. [4]

Fact | The IEA also reports that average global copper ore grades have fallen by approximately 40% since 1991, large greenfield projects take an average of roughly 17 years from discovery to production, and only about 5% of large deposits were discovered during the past decade. These indicators explain why supply cannot respond rapidly to price. [5]

Inference | Structural scarcity does not mean that the Earth's crust lacks copper. It means there is insufficient copper that can be delivered on time while meeting reasonable return requirements, permitting timelines and social constraints. Higher prices can improve project economics, but they cannot compress the physical time required for geology, infrastructure and approvals.

3.2 Operating Evidence in 2026 Is Mixed

Sources: BHP FY2026 Operational Review; Freeport Grasberg restart update; Ivanhoe Mines Q2 2026 results. [12][13][14]

Conclusion | A single incident cannot establish a decline in global mine output. A more useful 2026 framework is higher disruption frequency, inconsistent delivery from new projects and continued strong execution at some large assets. This supports volatility and a risk premium, but does not guarantee an absolute deficit in every quarter.

3.3 Refined and Recycled Supply: High Prices Activate Buffers

Scrap is an endogenous, reflexive response to high prices. As prices rise, recycling, dismantling, inventory release and substitution away from refined copper become more economical. However, scrap supply is constrained by policy, grade, collection cycles and regional flows, and therefore cannot fully replicate the immediate deliverability of cathode copper.

The observation sequence should be refined-to-scrap spreads -> scrap imports and domestic collection -> feedstock mix at copper-rod plants -> apparent refined-copper consumption. If the refined-to-scrap spread widens and scrap availability rises materially as price increases, the estimate of marginal refined-copper tightness should be reduced.

4. Global Demand: China Determines the Current Cycle; AI and Grids Determine the Slope

4.1 Regional Demand Matrix: Global Copper Demand Cannot Be Reduced to a Single China Factor

Fact | China remains the world's largest center for refined-copper use and processing. Chinese spot premiums or discounts, copper-semifabrication operating rates, the import-arbitrage window and inventories therefore remain high-frequency anchors for demand over the next 0-3 months. Incremental demand, however, is becoming global: US AI data centers and grids, European grid renewal, Japanese and Korean semiconductors and sovereign AI, Southeast Asian data centers, and industrialization and power investment in India and the Middle East jointly shape the medium- to long-term demand curve. [9][17]

Inference | China functions more like copper's short-cycle demand thermometer, while US AI and global grid spending are incremental sources of the demand slope. The two are not substitutes: if Chinese demand is destroyed, AI cannot fully offset the loss within a single quarter; if AI and grid projects continue to move into execution, they can raise the copper price floor and slow inventory rebuilding over 1-5 years.

4.2 China: Neither Aggregate Data nor Property Alone Is Sufficient

Fact | China's National Bureau of Statistics reports that fixed-asset investment fell 5.7% year on year in 1H26. Excluding real estate, it fell 2.7%; infrastructure investment declined 2.4%; manufacturing investment declined 1.2%; and real estate development investment declined 18.0%. High-tech industry investment rose 4.6% over the same period. [8]

Source: National Bureau of Statistics of China, fixed-asset investment for 1H26. Commercial floor area sold declined 11.6% year on year.

Inference | Simultaneous weakness in property, traditional infrastructure and manufacturing investment means the copper rally lacks support from a broad macro demand expansion. High-tech investment provides structural growth, but its near-term scale and deployment pace may not fully offset weakness in traditional sectors. The higher copper prices rise, the stronger the incentive for end users to manage demand by reducing inventories, delaying deliveries and substituting materials.

4.3 Global Electrification and AI Infrastructure: Capital, Compute and the Grid Form One Chain

Fact | Under the IEA STEPS pathway, total global copper demand rises from approximately 26.72 Mt in 2024 to 31.35 Mt in 2030. Clean-technology demand rises from approximately 7.74 Mt to 10.91 Mt, while secondary supply and reuse increase from approximately 4.44 Mt to 5.43 Mt. [6]

Source: IEA Copper Data Explorer, STEPS. Units are million tonnes. The figures are an IEA scenario, not Mortise Capital forecasts.

Fact | On 2026-08-10, NVIDIA announced an independent compute-financing platform with Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR. The platform aims to mobilize more than USD 500 billion of third-party capital over time to finance AI factories. The announcement describes a memorandum-of-understanding framework; final agreements, individual funding commitments and deployment schedules have not been determined. The USD 500 billion target therefore cannot be treated as an existing copper order. [16]

Fact | The IEA's 2026 update projects global data-center electricity use to rise from approximately 485 TWh in 2025 to approximately 950 TWh in 2030, while electricity use by AI-focused data centers roughly triples. The IEA previously estimated that 2030 data-center expansion could require approximately 512 kt of copper, or as much as 2% of total 2024 demand. Grid connections, transformer availability and permitting constraints could nevertheless delay roughly 20% of planned capacity. [9][17]

Sources: IEA Energy and AI and the IEA's 2026 update. Regional shares refer to 2024 data-center electricity use; 2030 is the IEA baseline scenario. The 512 kt estimate corresponds to capacity expansion by 2030, not a committed annual procurement volume.

Fact | Regional projects are beginning to provide verifiable MW and GW milestones. NVIDIA, NAVER and Brookfield plan to expand the GAK Sejong AI factory in Korea from 55 MW to 200 MW, with a 1 GW long-term pathway. NVIDIA's multi-year partnership with SK hynix covers next-generation memory, advanced manufacturing and the AI-factory supply chain. [18][19]

Inference | AI is not a single-point question about how much copper a GPU contains. Once financing eases the capital constraint, demand propagates through busways, power distribution, backup power, cooling and expansion of the external grid. The direction is bullish, but near-term trading requires final agreements, final investment decisions, grid connections and equipment orders. Until then, news raises the probability of future demand but does not automatically tighten the current-month physical market.

4.4 Order of Operations for Demand Analysis

5. Inventory: Aggregate Stock, Deliverable Stock and Geographic Location Must Be Separated

Claims that global inventories are falling often combine four different concepts: exchange-registered warrants, unregistered inventory in exchange warehouses, bonded or social inventories, and hidden industrial inventories. Only inventory that matches a specific delivery location, approved brand, tax treatment and transport time has equivalent significance for nearby pricing.

Verifiable Inventory Signals

• Bullish confirmation: LME cash backwardation persists; registered warrants and off-warrant stocks decline together; Asian physical premiums strengthen; and COMEX inventory stops absorbing metal from the rest of the world in one direction.

• Bearish confirmation: aggregate visible inventories across the three major exchanges rise consistently; the LME curve moves into contango; Chinese spot moves to a discount; and refined production continues to increase.

• Neutral or noise: a single week of warrant cancellations, concentrated inflows or outflows at one warehouse, or cross-market spreads that do not adjust for taxes and transport costs.

6. Global Policy and Geopolitics: The Largest Effect Is Inventory Relocation

6.1 US Copper Tariffs: The Direction Is Clear, but Refined Copper Requires Separate Treatment

Fact | The United States announced in July 2025 that semi-finished copper products and certain copper-containing derivatives would face a 50% tariff from 2025-08-01. The announcement also proposed a phased tariff path for refined copper beginning in 2027, but this should not be described as a comprehensive tariff on refined cathode copper already in force in 2026. [10]

Fact | In June 2026, the United States further adjusted its metal-product tariff regime: selected metal products remained at 50%; some derivatives were subject to 25%; certain industrial machinery and electrical equipment temporarily faced 15%; and the US-manufactured content threshold was revised. [11]

Inference | The first-order effect of tariffs is not to create global copper demand, but to change the optimal delivery destination, raise the value of holding metal in the United States and widen cross-market spreads. Only the second-order effect operates through downstream cost pass-through, import substitution and project delays. Treating the entire COMEX premium as evidence of a global shortage would overstate aggregate tightness.

6.2 Middle East, Sulfur and Smelting Costs

Fact | The IEA's 2026 outlook states that the Middle East accounts for approximately one-quarter of global sulfur supply and roughly half of seaborne sulfur trade passes through the Strait of Hormuz. Sulfuric acid is an important input in some copper smelting and hydrometallurgical processes. Regional conflict, shipping disruption or export-policy changes can therefore affect copper supply costs through sulfur and sulfuric-acid prices. [4]

Inference | Geopolitical risk can reach copper without passing through mines. Energy, sulfuric acid, marine insurance, ports and sanctions can all raise marginal costs or delay delivery. Trading analysis should monitor physical premiums and regional spreads rather than relying only on headlines.

6.3 Resource Nationalism and Social License

Tax regimes, export arrangements, community relations and infrastructure conditions in major producing jurisdictions such as Chile, Peru, the Democratic Republic of the Congo, Zambia and Indonesia can change project net present value and commissioning schedules. For listed equities, the key variables are attributable production, cash repatriation, fiscal stability and execution of expansions, not headline resource tonnage.

7. What Is Priced In and What Still Requires Proof

Source: Mortise Capital judgment. Scores range from -2 to +2 and are used to organize evidence; they are not a statistical model or price forecast.

Earnings, Valuation and Positioning Linkages

Earnings | For an unhedged miner with 1 Mt of annual copper sales and no price-participation terms, a USD 1,000/t move in copper has a mechanical revenue effect of approximately USD 1 billion. The actual EBITDA effect must also account for royalties, cost inflation, grade and by-product changes. This is a first-order sensitivity, not a company earnings forecast.

Valuation | High spot prices mechanically compress spot EV/EBITDA multiples for upstream companies and can create the illusion that they are cheap. A more robust approach is to calculate P/NAV and free cash flow under spot, base-case and stress prices while including expansion capex, taxes and minority interests.

Positioning | Managed-money net length indicates a strong directional consensus. If marginal new buying weakens, price can correct through profit-taking and higher volatility even without a deterioration in fundamentals. Position triggers should therefore depend on physical evidence, not only on a macro narrative.

8. Trading Framework for the Next 3-12 Months

Risk Control: Convert a Structural View into Stop-Lossable Hypotheses

• Define the risk budget before selecting the instrument. Options suit high volatility and policy tail risk; futures are better suited to clearly defined maturity or regional spreads.

• Every directional position must be linked to at least one physical indicator (premium or discount, inventories or operating rates) and one demand indicator (orders, investment, production or sales).

• Do not chase the first volatility expansion after treatment charges, cash backwardation or a news shock reaches an extreme; wait for the structure to stabilize.

• Relative-value trades require full-cost accounting, especially for tariffs, storage, financing, grade, foreign exchange and the delivery route.

9. Public-Equity Mapping: Who Truly Benefits and Who Merely Tracks Copper?

Without the user's portfolio or access to a live valuation terminal, this section provides research priorities and earnings transmission rather than individual stock ratings. Equity candidates are ranked by copper-price sensitivity, volume certainty, cost position, balance-sheet strength, and policy and execution risk.

Equity Ranking Conclusion

• First priority: upstream companies with low costs, credible production guidance and controlled capital expenditure. Price sensitivity converts into free cash flow only when volumes hold and costs remain controlled.

• Second priority: diversified leaders with copper volume growth, which reduce single-asset accident and country risk at the cost of lower copper-price sensitivity.

• Caution: independent smelters that rely on purchased concentrate and lack sufficient by-product income. Negative TC/RC first compresses their earnings; it does not directly benefit the entire copper value chain.

• Selective watch: grid and data-center equipment companies. They are genuine beneficiaries only when order growth, processing margins and working capital improve together; otherwise revenue can grow while cash flow deteriorates.

• Do not treat Apollo, BlackRock, Blackstone, Brookfield, Goldman Sachs and KKR as direct copper proxies. Their principal exposure is through assets under management, fees, credit and infrastructure investment, not the copper price itself.

10. Scenario Matrix: Price Is the Outcome; Triggers Are Tradable

The probabilities and price ranges below are research assumptions for risk management, not deterministic forecasts or target prices. All scenarios use LME copper as the benchmark.

The Most Sensitive Variables Across Scenarios

• Whether negative TC/RC causes an actual reduction in refined-copper output rather than only margin pressure.

• Whether LME cash backwardation persists together with inventory declines rather than reflecting a short-lived warrant event.

• Whether Chinese copper-semifabrication orders and inventories show high-price demand destruction.

• Whether AI financing converts into final agreements, GW or MW projects reaching FID, equipment orders and grid connections.

• Whether US tariff and inventory incentives continue to create structural divergence between COMEX and LME.

11. Monitoring Dashboard and Falsification Conditions

Five Conditions That Falsify the Bull Case

• LME cash-to-3-month moves from backwardation into sustained contango while deliverable inventories accumulate consistently.

• Chinese copper-semifabrication orders and operating rates decline, while the refined-to-scrap spread widens and scrap supply rises materially.

• Smelters deliver refined-output growth above expectations despite negative TC/RC.

• Grasberg, Kamoa-Kakula and other new projects recover or ramp materially faster than expected.

• A change in US policy removes the incentive to hold inventories on COMEX, while AI financing remains at the MOU stage for an extended period and project grid connections continue to be delayed.

Research Queue: Data Required for the Next Update

12. Bottom Line

The long-term structural case for copper remains intact. Declining ore grades, long project cycles, resource-country policy and infrastructure constraints prevent supply from growing as smoothly as demand models imply. Demand should be assessed through a two-anchor framework: China determines whether high prices destroy short-cycle demand, while US AI infrastructure and global grid capex determine the medium- to long-term demand slope.

The core question over the next 3-12 months is not whether USD 500 billion of AI financing will require copper, but when capital converts into final agreements, FIDs, equipment orders and grid connections. Record prices, LME cash backwardation, negative TC/RC and managed-money net length show that the market is already pricing tightness. Weak traditional Chinese investment, AI-project execution risk, scrap elasticity and regional inventory dislocation provide the counterweights.

One-Sentence Version

The long-term copper shortfall sets the direction; Chinese demand and physical market structure determine timing; the conversion of AI financing into grid and equipment orders determines the demand slope; and volume delivery plus cost position determines which equities can convert the theme into cash flow.

Appendix A | Definitions, Calculations and Source Freshness

Source and Freshness Statement

Market data use public snapshots through 2026-08-14; CFTC positioning is through 2026-08-11; Chinese macro data cover 1H26; company analysis uses the latest official operating disclosures available through 2026-08-14; and the AI-infrastructure event update is through 2026-08-15, using official NVIDIA announcements and public IEA research. Long-term supply and demand rely on public IEA 2026 materials. The latest monthly ICSG database is subscription-restricted, so the report does not use unverifiable monthly inventory or balance figures as load-bearing conclusions.

Boundary Between Fact and Inference: sentences labeled Fact are drawn from the official sources listed below; sentences labeled Inference are Mortise Capital's interpretations of those facts; probabilities, price ranges and thresholds are Scenario Assumptions, not realized outcomes.

Principal References

[1] London Metal Exchange, LME Official Prices / LME Copper, 2026-08-14. Link

[2] CME Group, Copper Futures Quotes, 2026-08-14 snapshot. Link

[3] U.S. CFTC, Disaggregated Commitments of Traders—Futures Only, 2026-08-11. Link

[4] IEA, Global Critical Minerals Outlook 2026—Executive Summary, 2026-07-16. Link

[5] IEA, Copper prices have hit record highs, but smelters face mounting strategic pressures, 2026-03-02. Link

[6] IEA, Copper data explorer / STEPS demand and supply table. Link

[7] International Copper Study Group, Copper Market Forecast 2026—2027 Press Release, 2026-04. Link

[8] National Bureau of Statistics of China, Investment in Fixed Assets from January to June 2026. Link

[9] IEA, Data centre electricity use surged in 2025, 2026-04-16. Link

[10] The White House, Adjusting Imports of Copper into the United States, 2025-07. Link

[11] The White House, Further Adjusting the Tariff Regimes for Imports of Aluminum, Steel, and Copper, 2026-06. Link

[12] BHP, Operational Review for the year ended 30 June 2026, 2026-07-16. Link

[13] Freeport-McMoRan, Update on Restart Plans for Grasberg Minerals District, 2025-11-18. Link

[14] Ivanhoe Mines, Q2 2026 Results Presentation, 2026-07-29. Link

[15] U.S. Geological Survey, Mineral Commodity Summaries 2026—Copper. Link

[16] NVIDIA, AI Compute Infrastructure Financing Platforms to Mobilize Over $500 Billion, 2026-08-10. Link

[17] IEA, Energy and AI, 2025; data-centre electricity, critical minerals and grid-connection analysis. Link

[18] NVIDIA, NAVER and Brookfield to Expand Korea's National AI Factory Infrastructure Buildout, 2026-07-24. Link

[19] NVIDIA and SK hynix, Multiyear Technology Partnership for AI Factories, 2026-06-07. Link

Appendix B | Disclaimer

This report is for research and discussion purposes only. All prices, probabilities, thresholds and trade structures are analytical examples and do not constitute an offer, solicitation or investment advice for any investor. Commodities and equities may be highly volatile. Futures and options may result in the loss of all invested capital or require additional margin. Readers should form independent judgments based on their own objectives, risk tolerance, tax position and legal environment.