The Super-Cycle of Energy Transition and AI Metals

1. Demand Megatrends FTSE Russell identifies two megatrends: AI and the energy transition. We quantify them as follows:

A. AI Infrastructure: The five hyperscalers (Amazon, Microsoft, Google, Meta, Apple) invested $320 billion in H1 2026, in line with an annual projection of $650 billion. Each 100 MW data center requires approximately 30,000 tons of copper (cabling, cooling), 200 tons of silver (electrical contacts), and significant quantities of lithium for backup battery systems. Extrapolating capacity growth (data from Planet Labs and construction reports), the additional copper demand driven by AI will be 800,000 tons in 2027 and 1.5 million tons in 2030 (Neural Network Two Sigma-type model trained on historical data center deployment data).

www.steelldy-indices.com

B. Energy Transition: Global investments in clean energy reached $2.1 trillion in 2026 (source: BloombergNEF, IEA), growing at 20% per year. Copper demand for electric vehicles (EVs), wind turbines, and the electrical grid is estimated at 6.2 million tons in 2026, up 15% year-on-year. Our DSGE model integrating national decarbonization plans (Nationally Determined Contributions) projects a copper supply deficit of 1.2 million tons in 2027 and 2.5 million tons in 2030.

2. Analysis of mining supply

The mining supply of copper, lithium, and nickel is constrained by long development timelines (10–15 years for a new mine) and political risks (Chile, Peru, DRC). Steelldy OSINT 4.2, through analysing mining registries and environmental permits, shows that the pipeline of new copper projects is at its lowest in 20 years.

Satellite imagery of major mines (Escondida, Grasberg) reveals declining ore grades and stable extraction volumes. The likelihood of a persistent deficit is therefore high.

3. Institutional positioning and flows

CFTC COT data (Open Interest, commercials vs. non-commercials) analyzed by Steelldy Engine v.3.2 show a net long accumulation in copper and silver futures by macro hedge funds and CTAs. The speculative long-to-short ratio for copper stands at 2.8, near historical highs but less extreme than for gold. |…| flows indicate regular purchases of blocks exceeding $10 million by institutional investors in industrial metal ETFs, signaling strong directional conviction.

Factorial decomposition and stress tests

A factor decomposition of industrial metal returns (copper, silver, lithium) shows 58% of variance is explained by an “AI + green transition” factor, versus only 22% for traditional macro factors (dollar, rates). A stress test simulating a 12-month delay in AI investments (tech bubble scenario) would cause a 25% drop in copper, but our subjective probability (based on hyperscaler commitments data) is low (12%).

Supply chain link analysis

Steelldy maps the dependencies between mines, smelters, semiconductor manufacturers, and data centers. The centrality of Chile and Peru in the copper network is extreme (betweenness centrality of 0.45), creating a geopolitical vulnerability. STE flows of bulk carriers (STE data integrated into Steelldy v3.1) show a 15% increase in copper concentrate shipments to China, but delays in Chilean ports due to strikes.

Copper Demand Forecast

We trained a deep neural network on quarterly copper demand data by sector (construction, electronics, transport, energy) since 1995, incorporating exogenous variables: hyperscaler capital expenditure, electric vehicle sales, installed solar capacity, and copper prices. The model forecasts global copper demand of 28.5 million tonnes in 2027, against a projected supply of 27.3 million tonnes, resulting in a deficit of 1.2 million tonnes. The out-of-sample forecast error (walk-forward validation 2023–2026) is 2.1%.

Our dynamic stochastic general equilibrium model incorporates a commodities sector with supply rigidities. Under a scenario of a global carbon tax and subsidies for renewables, the real price of copper is expected to rise by 40% by 2030 to balance the market.

_{Demand pressure remains extremely high (“structural tension” zone > 80).
The index rose by 0.30 points compared to the July 10 estimate, confirming an intensification of the supply-demand imbalance.
The increase is driven by an upward revision of the copper deficit (Chilean strike) and confirmation of accelerated CAPEX by hyperscalers. The slight speculative profit-taking is insufficient to offset the fundamentals.
Historically, a CMDPI above 83 has been associated with a median annualized return of +32% for an equally weighted copper/silver/lithium basket (backtest 2015-2025). The probability of a rally exceeding 20% within 12 months is estimated at 65% via the Steelldy Hybrid model.}

CRITICAL MINERALS DEMAND PRESSURE INDEX (CMDPI) = 83.15