Precious & industrial metals · Research note

Twenty years of gold, silver & copper — and the ten-year theses that follow.

A rand-investor's backtest of the three metals ETFs in USD and ZAR, read against the June 2026 macro, technology and financial landscape — distilled into four high-conviction positions for the decade ahead.

Coverage GLD · SLV · CPERPublished 24 June 2026

16.2%

Gold's 21-yr CAGR in rand (10.6% in USD)

R2,504

What R100 in gold became (from Dec 2004)

+185%

USD/ZAR move · R5.75 → R16.40

−26%

Gold's worst ZAR drawdown (vs −43% USD)

Over two decades, a rand-based holder of dollar-denominated metals earned the global price move plus the rand's secular slide — and, counter-intuitively, suffered shallower drawdowns than a dollar investor. That single structural fact reframes offshore hard assets from a diversifier into a core holding for the South African investor.

Gold is the standout compounder — 10.6%/yr USD, 16.2%/yr ZAR over 21.5 years; best risk-adjusted return in either currency.
The rand added ~5 points of return a year, for free — USD/ZAR ran R5.75 → R16.40, a 5.0%/yr depreciation stacked on the USD return.
The rand is a built-in hedge — every metal’s maximum drawdown was smaller in rand than in dollars.
Silver is a sizing problem, not a buy-and-hold — 8.0% USD CAGR but 32% volatility and a −72% peak-to-trough fall.
The vehicle matters as much as the view — copper-the-metal is in deficit, yet CPER-the-ETF returned just +58%; roll cost ate the thesis.

Methodology & data integrity

Prices: month-end adjusted closes (dividends & splits reinvested) for GLD, SLV, CPER via Alpha Vantage Premium. Each ETF runs from its own inception; a 14.5-year common period is shown separately for a like-for-like read.

Currency: the month-end USD/ZAR path is reconstructed from documented anchors with endpoints verified against live data. Headline ZAR returns are anchored to real endpoints; intra-anchor points are interpolated, so ZAR volatility/drawdown are indicative.

Twenty years, two currencies

The rand conversion lifts the annual return by roughly the currency’s own depreciation rate, and simultaneously compresses the worst drawdown, because rand weakness cushions dollar-price falls for a local holder.

Annualised return: USD vs ZAR — full available history

The ZAR–USD gap (~5%/yr) is the rand’s annualised depreciation — a structural tailwind for the rand-based holder.

Growth of 100 invested in gold (GLD), Dec 2004 = 100 — log scale

R100 in gold became ~R2,500; the same $100 became ~$880. The gap between the lines is the rand.

Five findings that drive the theses

1 · Gold compounds; the rest oscillate. Only gold delivered a smooth, institutional-grade risk-adjusted return.

2 · The rand is a return source and a shock absorber. ~5%/yr of tailwind and a permanently shallower drawdown.

3 · Silver is the highest-torque metal — in both directions. A sized, tactical expression, never a core hold.

4 · Copper-the-metal ≠ copper-the-ETF. Roll cost in CPER turned a strong commodity into a weak return. Own the miners.

5 · Diversification is real. Copper’s low correlation to gold (0.32) means a metals sleeve needs a monetary and an industrial leg.

The decade ahead

The theses

Each thesis follows the MCH macro-lens discipline: a single testable claim, the indicators that confirm or refute it, the mechanism and its counter, US-listed expression vehicles, review triggers, and a kill switch. Probabilities are explicit and sum to 100%. US-listed comparatives only.

Monetary · The reserve metal

Fiscal dominance & de-dollarisation drive a structural gold bull

Claim. Gold sustains a real (above-US-CPI) uptrend through 2035 as central-bank reserve diversification and deficit monetisation override the headwind of higher-for-longer policy rates; a rand holder compounds in the low-to-mid teens in ZAR.

Conviction split

Base 60%

Up 20%

Down 20%

Regime indicators

Indicator	Confirms thesis	Refutes thesis
Central-bank net gold buying	>800 t/yr sustained	Turns net selling
US deficit / GDP	Holds >6%	Falls below 3%
10-yr TIPS real yield	Drifts below 1.5%	Holds >2.5%
Gold share of global reserves	Rising vs USTs	Reverses to Treasuries

Mechanism

Post-2022 reserve weaponisation pushed non-aligned central banks (China, India, Türkiye, Gulf, CEE) to swap Treasuries for gold — a price-insensitive, structural bid. With debt >$37tn and 6%+ deficits, the Fed cannot durably hold real rates high without crushing debt service, so the market discounts eventual financial repression. Gold prices that discount.

Counter-mechanism

If Warsh genuinely restores positive real rates and fiscal consolidation narrows the deficit, the opportunity cost of a zero-yield asset rises and gold de-rates — the 2013–15 analogue, when GLD fell −43% in USD.

Expression vehicles · US-listed

Ticker	Why this vehicle	Sizing logic
GLD / IAU	Physical core; purest, most liquid exposure	Core
GDX	Senior miners — operating leverage to the gold price	Satellite
GDXJ	Juniors — higher torque, higher idiosyncratic risk	Small / tactical
NEM · AEM	Best-in-class operators; dividend + execution quality	Core-satellite
WPM · FNV · RGLD	Royalty / streaming — gold-price margin without operating risk	Core (quality)

▲ Confirmation triggers

CB buying accelerates >1,000 t/yr
10-yr real yield breaks below 1.0%
A sovereign-credit or debt-ceiling event

▼ Refutation triggers

Real yields hold >2.5% for two quarters
Deficit/GDP falls below 3%
Central banks turn net sellers

Kill switch

If 10-yr TIPS real yields hold above 2.5% for two consecutive quarters while the federal deficit narrows below 3% of GDP, the debasement thesis is broken and the gold overweight comes off.

Aligns with the standing MCH "USD devaluation" macro book (gold miners + royalty/streaming as primary expression).

Industrial · The bottleneck metal

The electrification supercycle makes copper the binding constraint — own the miners, not the ETF

Claim. A structural refined-copper deficit drives copper to new real highs over the decade — but the return is captured through producers and physical, not the roll-cost-laden CPER, which lagged spot badly across the backtest.

Conviction split

Base 55%

Up 25%

Down 20%

Regime indicators

Indicator	Confirms thesis	Refutes thesis
Refined-copper balance	Deficit widening	Swings to surplus
LME+COMEX+SHFE inventories	Drawing down	Building for 2+ quarters
Copper vs ~$9,000/t incentive price	Sustained above	Holds below
Data-centre power under construction	Rising	Stalls / cancellations

Mechanism

AI data centres use 27–33 t of copper per MW, and the grid to feed them needs 3–4× more again. Hyperscalers are now outbidding grid suppliers on transformer units. On supply, the IEA sees existing+planned mines covering only ~70% of 2035 demand; new ore is lower-grade, deeper and slower to permit. Price must rise to ration demand.

Counter-mechanism

Much of the 2025 rally is expectation-driven; US tariff front-running swelled visible inventory. If the AI build stalls on power/permitting (12 GW announced, 5 GW building) or China demand stays soft, no true deficit bites until ~2029 and copper caps ~$10–11k.

Expression vehicles · US-listed

Ticker	Why this vehicle	Sizing logic
FCX	Freeport — purest large-cap copper leverage, US-listed	Core
SCCO	Southern Copper — low-cost, high-margin (Peru risk)	Core-satellite
TECK	Teck — diversified copper-transition story	Satellite
COPX	Global X Copper Miners — basket exposure	Satellite
CPER	Tactical only — roll decay makes it unsuitable for buy-and-hold	Avoid for core

▲ Confirmation triggers

Exchange inventories draw below ~3 weeks of use
Major Chile/Peru/DRC mine disruption
Copper holds above $12k

▼ Refutation triggers

Hyperscaler capex guidance cut
Data-centre power approvals stall further
China grid/property demand contracts

Kill switch

If visible exchange inventories build for two consecutive quarters while copper trades below the ~$9,000/t incentive price, the deficit is not yet live — cut the copper-producer overweight to a watch position.

Infrastructure · AI's real constraint

The bottleneck is electrons, not chips — own the power and the grid

Claim. Through 2030 the binding constraint on AI is electrical capacity and equipment, not GPUs. The durable compounders are the owners and builders of generation, transmission and data-centre power, where order books are visible years out.

Conviction split

Base 60%

Up 25%

Down 15%

Regime indicators

Indicator	Confirms thesis	Refutes thesis
Transformer / switchgear lead times	Lengthening past 5 yrs	Normalising
Hyperscaler forward capex guidance	Rising	Cut >20% YoY
Power-equipment book-to-bill	>1.0 and rising	Falls below 1.0
Data-centre interconnection queue	Growing	Clears / reform floods supply

Mechanism

Electrical equipment is <10% of data-centre cost but 100% of the bottleneck. Transformer lead times have blown out from ~2 years to ~5. Of 12 GW announced for 2026, only ~5 GW is building — the gate is power. The winners locked in PPAs and equipment orders 3–4 years ago; this is a multi-year, supply-constrained order book.

Counter-mechanism

An AI-capex air-pocket — a DeepSeek-style efficiency shock that collapses compute intensity, or model-economics disappointment — would deflate the entire chain fast. Policy mismatch (accelerating AI while throttling renewables) could also raise power costs.

Expression vehicles · US-listed

Ticker	Why this vehicle	Sizing logic
GEV	GE Vernova — grid equipment + gas + nuclear; cleanest pure-play	Core
ETN	Eaton — electrical distribution & data-centre power	Core
VRT	Vertiv — data-centre power & thermal management	Satellite (higher beta)
PWR	Quanta — transmission & grid construction	Satellite
CEG · VST	Independent power / nuclear baseload for AI load	Core-satellite
CCJ	Cameco — uranium fuel for the nuclear leg	Small / thematic

▲ Confirmation triggers

Transformer lead times extend past 5 yrs
New multi-GW PPAs & nuclear restarts/SMRs
Backlogs and book-to-bill keep climbing

▼ Refutation triggers

Hyperscaler capex guidance cut
A step-change in compute efficiency
Interconnection reform that floods supply

Kill switch

If the top-four hyperscalers collectively cut forward capex guidance by more than 20% year-on-year, the demand pillar is breaking and the infrastructure overweight comes off.

Dovetails with the standing MCH "AI supply-side infrastructure" book (TSM · CEG · GEV).

Hybrid · The torque trade

Silver is the highest-torque expression of both themes — a sizing problem, not a hold

Claim. Silver outperforms gold in the cyclical up-leg as the gold:silver ratio mean-reverts — it is monetary and industrial (solar, electronics) — but its −72% historical drawdown and 32% volatility make it a tactical, position-sized expression, never a core hold.

Conviction split

Base 50%

Up 30%

Down 20%

Regime indicators

Indicator	Confirms thesis	Refutes thesis
Gold:silver ratio vs long-run mean	Elevated (silver cheap)	Compressed / expanding
Solar / PV installation growth	Rising	Thrifting / demand shock
Gold's primary trend	Leading higher	Rolls over
Silver market balance	Deficit widening	Surplus

Mechanism

Silver carries ~50% industrial demand (PV now the largest, fastest-growing slice) on top of a monetary bid that activates in precious-metal bull markets. The float is small, so investment flows amplify it — the backtest shows a best 12-month return of +200%.

Counter-mechanism

Silver is not a reserve asset — central banks don’t buy it — so it lacks gold’s structural, price-insensitive bid. A demand shock or a gold top removes both legs at once; the reflexivity that powers the up-leg savages the down-leg.

Expression vehicles · US-listed

Ticker	Why this vehicle	Sizing logic
SLV / PSLV	Physical silver exposure	Small core
SIL · SILJ	Silver miners — extreme operating torque	Tactical, stops in place
Long SLV / short GLD	Express the ratio view directly, hedging precious-metals beta	Pair (relative value)
SLV call spreads	Defined-risk upside that caps the −72% tail	Via us-options-analyser

▲ Confirmation triggers

Gold:silver ratio compresses from elevated levels
PV demand surprises to the upside
Silver deficit widens

▼ Refutation triggers

Gold rolls over
Ratio expands (silver lagging)
Industrial demand disappoints

Kill switch

If gold itself breaks its primary uptrend — the monetary leg fails — silver’s dual-identity case collapses and the position comes off immediately. Silver does not hold value without gold leading.

Disclaimer. This document is produced by MCH Advisory Services for informational and research purposes only. It does not constitute investment, financial, legal or tax advice, nor an offer or solicitation to buy or sell any security. All comparative instruments referenced are US-listed. Past performance is not a reliable indicator of future results; backtested figures are hypothetical and do not represent the return of any actual account. The author may hold positions in instruments mentioned and is not a registered financial adviser. Conduct your own due diligence and consult a qualified, registered adviser before making any investment decision.

Data & sources. Price history: Alpha Vantage Premium. Macro: US Federal Reserve / FOMC, Trading Economics, CNBC. Commodities: Wood Mackenzie & IEA critical-minerals outlooks; JPMorgan, Goldman Sachs, Bank of America copper research; industry reporting on AI power. All third-party figures paraphrased.