By Penny Langford
The global copper market is undergoing a structural realignment as artificial intelligence (AI) transitions from a software-driven hype cycle to a physical infrastructure imperative. While the "green energy" narrative: focused on electric vehicles (EVs) and renewables: has long been the primary driver of copper’s bull case, the rapid expansion of hyperscale data centers is emerging as the most inelastic component of demand.
As we move toward 2026, industry analysts and commodity strategists are increasingly pricing in a market where $15,000 per metric ton is no longer a speculative peak, but a probable floor. This shift is driven by a fundamental reality of the AI era: you cannot process data without a massive, copper-intensive physical footprint.
The copper-intensity of AI infrastructure
To understand why copper demand is decoupling from traditional industrial cycles, one must look at the anatomy of a modern AI data center. Traditional data centers are power-dense, but AI-ready facilities, powered by high-performance GPUs like those from NVIDIA, require significantly more robust power distribution systems.
A single hyperscale AI data center is now estimated to require between 1,500 and 2,500 tonnes of copper. This metal is not hidden in the chips; it is the backbone of the power delivery system. Copper busbars, high-voltage cabling, and advanced liquid cooling systems are essential to manage the immense power loads and heat generated by AI training clusters.
The intensity of this demand is driven by three primary factors:
- Power Density: AI racks can consume five to ten times more power than traditional server racks, requiring thicker copper conductors to prevent overheating and energy loss.
- Grid Connection: Bringing gigawatts of power to a single site requires massive investments in transformers and substation infrastructure, all of which are copper-heavy.
- Cooling Systems: As AI chips run hotter, the shift toward liquid cooling and high-efficiency heat exchangers adds another layer of copper requirements.
The Math: Projecting 2026-2027 demand
According to recent sector analysis, copper demand from AI-powered data center facilities is expected to average roughly 400,000 tonnes per year over the next decade. However, the growth is front-loaded. By 2026, aggregate data center copper demand is projected to reach 740,000 tons annually.
For context, the total global "call on copper" is expected to rise from 28 million metric tons in 2025 to over 42 million by 2040. While data centers currently represent a single-digit percentage of global consumption, their share is growing faster than any other sector. Some forecasts suggest data centers will account for 8.5% of total copper demand by 2027, up from just 4.1% in 2025.
| Year | Data Center Copper Demand (Est. Tonnes) | Share of Global Demand (%) | Average Copper Price Forecast (USD/t) |
|---|---|---|---|
| 2024 | 280,000 | 1.1% | $9,200 |
| 2025 | 450,000 | 4.1% | $10,500 |
| 2026 | 740,000 | 6.8% | $12,500 – $14,500 |
| 2027 | 920,000 | 8.5% | $15,000+ (Bull Case) |
Data compiled from Bank of America, Citi, and S&P Global Research.
Supply-side bottlenecks: Why mines can’t keep up
The demand spike from AI infrastructure is hitting the market at a time when primary supply is increasingly fragile. The mining industry is struggling with declining ore grades, rising geopolitical risks, and a chronic lack of new "tier-one" discoveries.
Major projects intended to bridge the supply gap are facing significant headwinds. For example, the Oyu Tolgoi project in Mongolia has faced logistical and political blockades that threaten global flow. Similarly, while the Zambian copper resurgence offers hope, the timelines for bringing massive new capacity online: such as the Konkola turnaround: are measured in years, not months.
Exploration successes, such as the Mocoa extension in Colombia, are critical for the long term, but they cannot solve the immediate squeeze projected for 2026. The lead time from discovery to first production for a copper mine now averages 15 to 20 years. This means the supply available in 2026 and 2027 is essentially locked in today; it cannot be meaningfully increased regardless of how high the price climbs.
Price forecast: The path to $15,000
The convergence of inelastic AI demand and constrained supply has led major financial institutions to revise their 2026-2027 outlooks.
- Citi’s Bull Case: Analysts at Citi have issued a scenario where copper tests $15,000 per metric ton by late 2026. They argue that the market is underestimating the "intensity of the squeeze" as data center developers prioritize speed-to-market over metal costs.
- Bank of America: BofA strategists project an average price of $13,501 in 2027 but note that in a "tight-market spike," prices could easily clear the $15,000 threshold.
- Goldman Sachs: While framing $15,000 as a 2030 target in real terms, Goldman acknowledges that the structural deficit could pull that timeline forward if the AI build-out continues at its current 165% power-demand growth rate.
For operators and investors, the path to $15,000 copper represents a fundamental shift in project economics. Higher prices make previously marginal deposits viable but also increase the CAPEX for the very data centers driving the demand.
Substitution and the "Aluminum Threat"
The primary risk to the $15,000 floor is substitution. As copper prices rise, engineers look toward aluminum, which is lighter and cheaper. However, aluminum is only 60% as conductive as copper. In the high-density environment of an AI data center, where space is at a premium and heat management is the primary constraint, switching to bulkier aluminum conductors is often technically unfeasible.
For the "last mile" of power delivery within a server rack, copper remains the only viable material due to its superior thermal and electrical properties. This makes AI demand far more price-inelastic than traditional sectors like residential wiring or general consumer electronics.
Operational impact on the mining sector
To meet this demand, the industry must look beyond traditional extraction. We are seeing a move toward more automated, data-driven mining to squeeze efficiency out of existing operations.
The focus for 2026 will likely shift to:
- Brownfield Expansions: Increasing throughput at existing mines where infrastructure is already in place.
- Secondary Recovery: Improving tailings processing to recover copper that was previously considered waste.
- M&A Activity: Larger miners acquiring junior explorers with advanced-stage assets to replenish their pipelines.
Conclusion: A structural reset
The "AI copper squeeze" is more than a temporary price spike; it is a structural reset of the global copper market. As data centers evolve into the most critical infrastructure of the 21st century, their reliance on copper creates a permanent floor for the metal’s valuation.
For decision-makers in the mining and tech sectors, the message is clear: the era of cheap, abundant copper is over. The race is now on to secure the supply chains that will power the intelligence of the future.
Social Media Snippet (LinkedIn/X):
Is $15,000/t the new reality for copper? ? As AI data centers scale, they are consuming up to 2,500 tonnes of copper per facility. With supply constrained by aging mines and geopolitical hurdles, the "AI Copper Squeeze" is no longer just a theory. Penny Langford breaks down the 2026 outlook for Skillings Mining Intelligence. #MiningNews #Copper #AI #CriticalMinerals #EnergyTransition
