As the global AI race intensifies, “copper and electricity” are rapidly becoming structural constraints. In its latest report, Goldman Sachs warns that surging power demand from AI-driven data centers is pushing the US electricity system toward its limits, potentially becoming a key bottleneck to AI expansion. At the same time, copper—an indispensable metal for electrification and AI infrastructure—retains a compelling long-term bullish outlook. The emerging “copper–power equation” may ultimately reshape global AI competitiveness.
US Power Supply Under Pressure: AI Data Centers Strain Reserve Margins
Goldman Sachs data show that US data center capacity hit a record high in November, with annualised power demand growth nearing 3%, well above GDP growth. However, electricity supply is struggling to keep pace:
· Shrinking reserve margins: In most US power markets, reserve generation capacity is already at or below safe operating thresholds. Last summer, this led to sharp real-time power price spikes, notably in the PJM market serving Virginia.
· Highly concentrated demand: Around 72% of US data centers are located in just 1% of counties, significantly increasing local grid overload risks, especially in northern Virginia, the country’s largest data center hub.
· Worsening outlook: Coal plant retirements (around 5% annually) and slower renewable buildout, combined with continued data center expansion, are expected to further erode reserve margins by 2026, raising the risk of regional power outages and directly constraining AI compute deployment.
Copper’s Long-Term Bull Case: AI Infrastructure Meets Structural Supply Limits
Running parallel to power constraints is tightening copper fundamentals. As the backbone metal of electrification, copper faces growing demand and rigid supply limits:
· Demand: Nearly 50% of global copper demand already comes from power-related sectors, including data centers, grids, and renewable energy. AI infrastructure—such as server wiring and transformers—is emerging as a new growth driver.
· Supply: Copper mine development requires 10–15 years, with production highly concentrated—Chile and Peru account for about 40% of global output. China is unlikely to achieve rapid supply expansion through overseas investment alone.
· Price outlook: Goldman Sachs forecasts copper prices to remain elevated in 2026, averaging USD 11,400 per tonne, and rising to USD 15,000 per tonne by 2035. The bank reiterates copper as its top long-term commodity pick, outperforming aluminium, lithium, and iron ore, which could fall 15–25% by 2026.
China’s Advantage: Power Reserves Strengthen AI Competitiveness
In contrast to the US, China benefits from robust power reserves. Goldman Sachs estimates that by 2028, China’s effective power reserve capacity will reach 360 GW—more than three times projected global data center power demand (105 GW). This provides a strong energy security buffer for continued AI expansion.
Conclusion: The AI Race Is Also a Resource Race
At its core, the global AI competition is a competition for critical resources. Power constraints and copper supply pressures may slow AI development in the US, while China’s ample power reserves and resilient copper demand position it more favourably. Goldman Sachs’ message is clear: copper’s long-term value is embedded in every incremental gain in AI computing power. As data centers illuminate the digital future, copper remains the invisible conductor enabling it all.
Source：Changjiang Nonferrous Metals Network