Artificial intelligence’s rapid expansion could run into a critical mineral supply crunch as demand for copper, lithium and rare earths outpaces mining production, threatening the infrastructure needed to support the technology, according to new research from Global X.
The investment manager’s Charting Disruption 2027 report forecasts that mined demand could exceed supply by 17% to 22% across the three mineral categories by 2035. A single one-gigawatt AI data centre could require nearly 180,000 tonnes of critical materials, including about 27,200 tonnes of copper.
The findings point to a widening gap between the speed of technological development and the mining industry’s ability to deliver raw materials. New mines take an average of 10 to 18 years to develop, leaving producers struggling to respond to demand from data centres, power grids, electric vehicles and advanced manufacturing.
Global X research analyst Madeline Ruid said access to essential resources is becoming increasingly difficult as technological adoption accelerates. Concentrated mineral supply chains, lengthy project development timelines and water constraints could raise costs, delay construction and expose infrastructure projects to geopolitical disruptions.
The challenge could shift investor attention beyond semiconductor manufacturers and technology companies towards miners, mineral processors and energy infrastructure developers. Securing reliable supplies of industrial metals is becoming increasingly important to the next phase of AI investment.
$550 billion needed
Copper stands out as a key material because of its widespread use in electrical wiring, transformers, transmission networks and cooling equipment. Expanding AI computing capacity requires more than constructing data centres; it also depends on upgrading the electrical systems that supply them.
Global X estimates that the copper, lithium and rare earths industries could require nearly $550 billion in investment through 2040 to meet growing demand and address supply chain vulnerabilities.
The broader requirement for critical minerals and water infrastructure could exceed $2 trillion over the same period.
Supply risks extend beyond the availability of mineral deposits. Production and processing are concentrated in relatively few countries, exposing manufacturers to trade restrictions, geopolitical tensions and disruptions that can affect access to refined materials.
Rare earths are particularly important for permanent magnets used in industrial equipment, electric motors and other advanced technologies. Lithium remains central to battery storage systems that can help balance electricity supply as grids accommodate additional demand.
Higher projected consumption does not guarantee shortages or sustained price increases, the report warns. New discoveries, recycling, substitution and improvements in processing efficiency could ease pressure on some commodities. Mining companies also face permitting delays, financing constraints and construction costs that can undermine project economics.
The investment implications will depend on which materials face genuine shortages, how quickly new production becomes available and whether demand forecasts translate into actual consumption.
Infrastructure pressures
Mineral availability is only one potential constraint on AI growth. Electricity generation and transmission capacity could become equally important in determining how quickly developers bring new computing facilities online.
Global X projects that data centres could account for 20% of US electricity demand by 2035, compared with about 6% currently.
Planned additions of dispatchable power generation, meanwhile, are expected to meet only 42% of new US electricity demand through 2030.
The imbalance could accelerate investment in nuclear energy, natural gas, renewable generation and grid modernization. Each option carries different construction timelines, capital requirements and mineral needs, creating additional demand across the resources sector.
The report also identifies a shortage of immediately available computing facilities. US data centre vacancy rates are projected to fall to about 1% in 2026, while average asking rents have risen 73% from 2021 levels.



