The Pull of Data Centers on Copper Demand Is Overestimated

Overview

This in-depth report from Bernstein takes a comprehensive look at the long-term supply and demand dynamics of copper and raises three critical questions about the market’s hottest narrative—the role of data centers in driving copper demand. The report argues that although global electrification trends (electric vehicles, renewable energy, etc.) will drive copper demand growth in the long term, the market’s expectation for data centers alone may be overly optimistic. The core conclusion is that, due to technological advances and grid bottlenecks, the actual copper demand driven by data centers will fall far short of market expectations; the real bright spot lies in energy storage systems (ESS). On the supply side, multiple structural challenges will create an expanding supply-demand gap after 2030, providing long-term support for copper prices—but current stock valuations have already priced in most of the positive factors.

Core views

Demand Side: Correcting Three Misconceptions About Data Centers The report highlights three major misunderstandings about data centers driving copper demand. First, although global data center capacity is expected to reach nearly 200 GW by 2030, actual copper demand won’t grow proportionally. This is because the introduction of 800V DC distribution architecture—especially its early adoption in AI training facilities—will significantly reduce copper intensity. NVIDIA estimates that this technological shift could cut copper demand by around 45%. As a result, the report forecasts that copper demand from data centers will peak around 2028 (about 500,000 tons) and then gradually decline to about 200,000 tons by 2040, contributing only about 1% of global demand overall. Second, the expansion of data centers doesn’t have a one-to-one direct relationship with investment in grid transmission and distribution (T&D) and copper demand. Due to long queue times for grid interconnection in the U.S. and Europe—often exceeding four years—more and more data center operators are turning to “behind-the-meter” (BTM) self-generation solutions. Surveys show that by 2030 and 2035, 33% and 44% of data center power needs, respectively, could be met by 100% on-site generation. Because these solutions are closer, they use less copper per unit of power, thus weakening immediate demand for grid expansion in the short term. Third, compared to data centers, energy storage systems (ESS) represent a more important and stable new source of copper demand. As the share of solar and wind power grows, the electricity system’s need for flexible regulation surges, driving ESS capacity from about 550 GWh in 2025 to about 1,500 GWh by 2030. Battery storage mainly uses lithium iron phosphate (LFP) technology, which has a copper intensity of about 1.02 kg/kWh. By 2030, this is expected to add about 1.5 million tons of incremental copper demand. More importantly, there’s no sign yet of a decline in copper intensity for ESS batteries in the short term. The Real Long-Term Demand Drivers: Electric Vehicles and Renewable Energy Beyond data centers and energy storage, the long-term growth of copper demand remains firmly driven by electric vehicles (EVs) and renewable energy. The report forecasts that copper demand from EVs will rise from 2 million tons in 2025 to 4.4 million tons by 2030; renewable energy (wind, solar) demand will increase from 3.3 million tons to 4.7 million tons. Overall, global copper demand is expected to grow at a compound annual rate of 2.7%, from about 34 million tons in 2025 to about 40 million tons by 2030, before slowing down to 2.4% and reaching about 51 million tons by 2040.

Analysis framework

Bernstein’s analytical framework is a typical combination of top-down and bottom-up approaches. On the demand side, the report first deconstructs and corrects the ‘hot narratives.’ Rather than simply linearly extrapolating data center capacity growth, it delves into technical details, quantifying the disruptive impact of the 800V VDC architecture on copper intensity (tons of copper per megawatt). It models separately for AI training, AI inference, and non-AI data centers. Then, it expands its analysis from within data centers to the grid level, citing U.S. grid interconnection queue data and operator strategies (such as Digital Realty and Equinix) to demonstrate the diminishing effect of the data center → grid investment → copper demand chain. On the supply side, the report employs a probability-weighted project tracking method. It builds a detailed global database of copper mine projects, categorizing them according to their certainty into “definitely progressing,” “highly likely,” and “possible,” assigning different realization probabilities of 100%, 50%, and 15%, respectively, to aggregate future supply. The advantage of this approach is that it looks beyond just planned project capacity—it also incorporates realistic assessments of licensing, financing, and geopolitical risks. In terms of pricing and valuation, the report combines classic industry marginal analysis methods. It estimates the C1 cash cost curve and long-term industry profit margins (around 55% post-China era) to determine the long-term support level for copper prices (about $10,700 per ton). Finally, by combining supply-demand balance, cost support, and current industry EBITDA margins, it derives its long-term copper price forecast and compares it with the implied copper price expectations embedded in current stock prices, thereby giving a ‘Market-Perform’ rating.

Methodology notes

• Industry/Industrial Analysis FrameworkQuantity-Price Decomposition

  Data Center Copper Demand = Installed Capacity × Copper Intensity (tons of copper per megawatt)

  This is an analytical method that breaks down total demand into ‘quantity’ (installed capacity) and ‘price’ (unit intensity). Bernstein precisely demonstrates that even though ‘quantity’ is growing, ‘price’ is falling faster, proving that total demand may not grow synchronously and correcting the market’s linear extrapolation mistake.

• Cycle and Cyclical FrameworkCyclical Turning Point Analysis

  Determining the long-term copper price equilibrium by analyzing the mean-reversion characteristics of industry profit margins and the support role of cost curves.

  Bernstein studies historical data and finds that copper industry EBITDA margins exhibit mean-reversion behavior. Using the estimated cost curve for 2030 and a post-China-era profit margin target of about 55%, it retroactively calculates the long-term equilibrium copper price at $10,700 per ton. This provides a valuation anchor for determining whether current prices are high or low.

• Company Fundamentals and Financial FrameworkROIC–WACC spread

  The relationship between mining companies’ return on invested capital and their cost of capital, as well as the impact of capital expenditure intensity on new supply.

  When analyzing supply, the report implicitly uses ROIC analysis. It calculates the capital costs of new or restarted mines (e.g., the Hudbay project’s capital intensity of about $4,400 per ton-year capacity) and operating costs (about $7,000 per ton) to assess the economic feasibility of projects at given copper prices, thereby determining the pace of future supply release.

• Industry/Industrial Analysis FrameworkCost curve analysis

  Finding price floor support by analyzing the distribution of global mine cash costs (C1 costs).

  On the left side of the cost curve are low-cost mines, and on the right side are high-cost mines. When commodity prices fall into the high-cost range, some mines will shut down due to losses, reducing supply and providing price support. Bernstein uses this principle, estimating costs at the 25th, 50th, 75th, and 90th percentiles to build a dynamic model of copper price floors.

• Company Fundamentals and Financial FrameworkPE/PEG valuation

  Investors pay a premium based on companies’ future earnings expectations; high-growth companies typically enjoy higher valuation multiples.

  The report notes that most copper mining companies’ EV/EBITDA multiples have exceeded their five-year historical average and are higher than the levels implied by the market consensus copper price (about $12,500 per ton), meaning the market has already priced in higher copper price expectations (> $13,500 per ton). This is a typical application of using valuation multiples to retroactively infer market-implied expectations.

Asset mapping & comparison

Structured mapping from thesis to named assets (strengths, weaknesses, peers, risks).

• Antofagasta PLC (ANTO.LN)

  As a major pure copper miner, its performance is highly correlated with copper prices. The report views it as a direct proxy for copper price expectations.

  Strengths

  It owns large mines such as Los Pelambres and Centinela, with clear prospects for production growth.

  Weaknesses

  Its current stock price trades about 1.7 standard deviations above its five-year EV/EBITDA average, indicating overvaluation.

  Comparison

  Compared to FCX, its valuation deviation from historical averages is greater; the target price implies about 27% downside potential.

  Risks

  Downside risk from copper prices, project execution risks, and rising Chilean mining royalties.

• Freeport-McMoRan Inc (FCX.US)

  One of the world’s largest publicly listed copper producers, owning world-class mines such as Grasberg and Morenci. It benefits from rising copper prices but also faces the risk of overvaluation already reflecting positive factors.

  Strengths

  It has low-cost leaching projects in the U.S. that are expanding, such as MEH at Morenci and the Jetti Resources partnership, which can improve resource recovery rates.

  Weaknesses

  Although its valuation is relatively milder than ANTO (0.6 standard deviations above the average), its target price still implies downside potential.

  Comparison

  Compared to ANTO, FCX’s valuation bubble is smaller, but its short-term upside potential remains limited.

  Risks

  Fluctuations in copper prices, operational risks in Indonesia (Grasberg mine), and changes in mining costs.

Key data

• Global Copper Demand in 2025About 34 million tonsBernstein’s estimate of the actual demand base for 2025.
• Global Copper Demand Forecast for 2030About 40 million tonsCompound annual growth rate from 2025 to 2030 is about 2.7%, lower than GDP growth.
• Global Copper Demand Forecast for 2040About 51 million tonsCompound annual growth rate from 2030 to 2040 slows to about 2.4%.
• Peak Copper Demand from Data Centers and TimingAbout 500,000 tons, 2028After that, demand is expected to decline due to the widespread adoption of 800V VDC technology, dropping to about 200,000 tons by 2040.
• Reduction in Copper Demand from 800V VDC ArchitectureReduced by about 45%Higher voltage reduces current, allowing thinner conductors and significantly cutting copper usage.
• Incremental Copper Demand from Energy Storage Systems (ESS) by 2030About 1.5 million tonsDriven by ESS capacity growth from 550 GWh to 1,500 GWh, making it a more important copper demand source than data centers.
• Increase in Transmission and Distribution (T&D) Capital Expenditures from 2030 to 2050About $50–80 billion per yearDue to data center construction, the midpoint baseline of the previous STEPS/APS scenario has been adjusted toward the more optimistic APS scenario.
• Long-Term Copper Price Forecast (Bernstein)$10,700 per tonBased on the 2030 global cost curve and a long-term industry profit margin of about 55% post-China era.
• Copper Price Forecast for Second Half of 2026 (Bernstein)$11,500 per tonAssuming the U.S. does not impose additional tariffs on imported refined copper, the release of Comex inventories will put downward pressure on prices.

Impact & implications

Bernstein’s report directly challenges the market’s mainstream narrative. Through detailed data, it proves that data centers aren’t the ‘super growth engine’ for copper demand investors expect—they contribute far less than electric vehicles and renewable energy. In the short term, if the market re-prices this view, the valuations of copper concept stocks that have been highly anticipated could come under pressure. The report specifically points out that companies like Lundin and Southern Copper are trading at more than two standard deviations above their five-year average, making them relatively more vulnerable to corrections. In the medium to long term, although data center demand has been lowered, the severe supply shortage starting from 2030 hasn’t changed. Long-term supply-side challenges—such as declining ore grades, scarcity of high-quality projects, geopolitical risks, and rising capital costs—are still building a structurally tight market. This means copper prices have a solid long-term floor, and the current correction could offer better entry opportunities for long-term investors. For relevant companies, ANTO and FCX, as major pure copper miners, have been given a ‘Market-Perform’ rating. Although analysts have raised their target prices based on the latest commodity prices to £28 and $58.50, they believe the stock prices have already reflected the positive factors, leaving limited upside potential and requiring further improvement in earnings (EBITDA) or more attractive valuation levels to provide room for growth.

Risks

• Price Risk: If the U.S. imposes additional tariffs on refined copper imports, it could lead to further accumulation of Comex inventories, putting significant downward pressure on short-term copper prices.
• Supply Risk: Key mine projects like Escondida OGP2 depend on licensing approvals, posing risks of execution delays. The Reko Diq project has had its advancement probability lowered due to worsening geopolitical security conditions.
• Demand Risk: The long-term outlook for copper demand from electric vehicles faces technological substitution risks, such as centralized electrical architectures, aluminum wiring harnesses, and higher-voltage systems, which could reduce copper usage per vehicle.
• Technological Change Risk: If sodium-ion batteries accelerate penetration into downstream ESS markets (with copper intensity about 30% lower than LFP), it could weaken the growth expectations for copper demand from energy storage systems.

What to watch

• U.S. Refined Copper Tariff Decision: Whether the U.S. government imposes additional tariffs on imported refined copper this month will be a key variable affecting short-term copper prices and inventory destocking.
• Cobre Panama Mine Restart Progress: The Panamanian government’s review results for the mine owned by First Quantum are expected to be announced mid-year; its restart will bring significant incremental supply to a tight market.
• FID for Escondida OGP2 Project: BHP plans to make a final investment decision on this project in 2027 or 2028, making it an important leading indicator for future supply.
• AI Chip and 800V Platform Progress: Watch the deployment pace of next-generation platforms like NVIDIA’s Rubin Ultra—this will be the catalyst triggering a sharp drop in copper intensity in data centers.
• China’s Demand Dynamics: As the world’s largest copper consumer, China’s economic and real estate market recovery remains the biggest variable on the demand side.