Copper

The question

Is the copper complex priced for a cyclical mid-cycle with incremental grid and EV demand, or for a structural deficit in which the electrification build-out, data-center power density, and sub-par mine-supply growth compound on the same five-year curve?

Base rates

The reference class is prior copper supply-deficit episodes: 2003-2008, 2010-2011, and 2020-2022. In each prior cycle, deficit to stockpiles ran for three to five consecutive years, LME prices moved 60-110% above the marginal incentive price, and diversified miner equity returns averaged +20-28% annualized over the deficit window with significant dispersion between low-cost and high-cost operators. The base rate for a diversified copper-weighted basket entering year one of a sustained deficit has historically been roughly the 85th percentile of commodity sub-sectors for the first three years.

The current deficit is structurally deeper. Commissioned copper project pipeline through 2030 totals roughly 2.5-3.0 million tonnes of incremental annual capacity against forecast demand growth of 5-7 million tonnes. The gap is large, and new greenfield copper projects take 12-18 years from discovery to first production; the permitting and capital intensity issues that produced the current gap do not resolve quickly.

The lithium sub-base-rate is different. The spot price has fallen roughly 80% from the 2022 peak and is now below the estimated all-in sustaining cost for roughly 40% of global supply. Prior episodes of lithium below marginal cost -- 2019-2020 and parts of 2015-2016 -- resolved within four to six quarters as supply was curtailed. The reference class does not require demand to inflect; it only requires supply to respond, which is the observable variable.

For data-center copper, the reference class is thinner because the demand segment is newer. The arithmetic is nonetheless specific: a hyperscale facility with 100 MW of IT load requires roughly 4,000-6,000 tonnes of copper in busbars, switchgear, transformers, and cabling. Announced North American hyperscale pipeline through 2028 is roughly 25 GW of incremental IT load. That is a discrete new demand leg the 2010-era copper models did not include.

Why consensus is wrong

Consensus is modeling copper with a demand curve drawn from the 2010-2020 decade, when EV penetration was the marginal story and grid spending was a slow-grind variable. Three pieces of the story are mis-specified.

First, grid capex. North American, European, and Chinese transmission and distribution spending has inflected to a growth rate of 8-11% from the prior 3-5%. The copper intensity per dollar of grid capex is roughly double the copper intensity per dollar of general industrial capex. The category mix shift is itself a demand tailwind independent of headline capex growth.

Second, data-center electrical density. The move from 10-20 kW per rack to 100+ kW per rack for AI training compresses the copper-per-megawatt ratio upward because higher current density at a given voltage requires thicker conductors and more redundant paths. The sell-side copper model treats data centers as a general-industrial line item at historical intensity.

Third, the lithium curve. Consensus extrapolates spot prices and estimates equity fair value against those prints. The cost-curve mechanic -- that sub-marginal tonnes exit and that the incentive price to build new spodumene or brine capacity is multiples of current spot -- operates on a two-to-four-quarter lag. The equity is being valued as if the cost curve does not exist.

Position construction

The book has three 20% anchors, a copper pure-play cluster, a lithium trough cluster, and an adjacency tail.

Anchors. RIO at 20% and BHP at 20% are the diversified majors with the lowest-cost iron ore books subsidizing a growing copper franchise -- Oyu Tolgoi at RIO and Escondida plus the South Australia copper province at BHP. Both generate sufficient free cash flow at current commodity prices to fund their copper growth programs without equity issuance, which is the scarce attribute in the sector. TECK at 20% is the third anchor -- post-coal-divestiture pure-play copper operator with QB2 ramping, a clean balance sheet, and the cleanest near-term production growth curve in the sector. Note that FCX and SCCO are explicitly excluded from this book -- they sit in the AR Critical Minerals preset where the thesis is different.

Copper pure-play cluster (~21.1%). HBM at ~15.6% is the Canadian and Peruvian copper and zinc producer whose Copper Mountain acquisition and Snow Lake operations give above-peer volume growth. ERO at ~5.5% is the Brazilian copper producer with the Tucuma project just entering commercial production.

Lithium trough cluster (~7.8%). LAC at ~4.0% owns Thacker Pass, the largest permitted North American lithium project with a DoE loan and an offtake with a major OEM. SGML at ~1.7% is the Brazilian spodumene producer with industry-low cash costs. SLI at ~2.1% is the direct-lithium-extraction development story in the Smackover formation.

Adjacency and optionality (~11.2%). TMC at ~8.0% is the seabed-nodule copper, nickel, cobalt, and manganese resource with a material optionality payoff on a successful ISA regulatory path. KRO at ~1.3% is the titanium dioxide producer at cycle trough -- a pigment adjacency that tracks the same industrial construction cycle. CMP at ~1.8% is the salt and plant-nutrition producer with specialty sulfate-of-potash exposure; the position is small because the lithium adjacency did not deliver.

Asymmetric payoff

If copper deficits widen as the pipeline suggests, grid and data-center demand compound through 2028, and lithium spot recovers toward the marginal cost of new supply, the weighted book returns roughly 20-30% annualized over three years. If copper deficits narrow on a Chinese industrial slowdown and lithium stays below cost curve for another year, the book returns roughly -5% to +5% with diversified-major free cash flow providing a floor. If a major new copper project slips or a producing asset is curtailed on geopolitical grounds, the right tail is 38-50% with the copper pure-plays carrying most of the convexity.

At 55% base, 25% bear, and 20% bull, expected value is roughly +15 to +22% annualized against an SPY base rate near +8%. The payoff is asymmetric because the diversified-major anchors truncate the bear case through cost position and free cash flow, while the pure-play cluster, the seabed optionality, and the lithium trough cluster each carry uncapped right tails on independent drivers.

Three things that would change our mind

1. A Chinese industrial policy pivot that translates into a sustained slowdown in grid capex and property-adjacent copper consumption, with State Grid capex guidance rolling over for two consecutive annual plans -- this would remove the largest single tailwind on the demand side.
2. Commissioned copper project pipeline expanding materially through 2028, with two or three greenfield projects clearing permitting and financing milestones ahead of schedule and management commentary pointing to accelerating rather than stalling supply response.
3. Lithium spot prices staying below the estimated all-in sustaining cost of the 40th percentile tonne for more than six additional quarters, with supply curtailments failing to materialize -- which would indicate demand rather than supply is the binding variable and collapse the trough-cycle thesis on the lithium sleeve.

What we're explicitly NOT betting on

We are not betting on a specific copper price target. We are not betting on the timing of any single project -- QB2 ramp, Thacker Pass first production, Tucuma commissioning, or ISA seabed permitting. We are not betting on a particular EV penetration trajectory; the grid and data-center copper legs carry the thesis without incremental EV assumption. We are not betting on a specific lithium price recovery date. We are not holding FCX or SCCO here -- those sit in the AR Critical Minerals preset. The thesis requires only that the copper supply deficit persists on its announced trajectory, that grid and data-center demand stays on its announced path, and that the lithium cost curve behaves like every prior commodity cost curve. All three are weaker claims than picking project timing, and the book is sized for them.