The Philadelphia semiconductor index is up 170 per cent in the last 12 months with half of that increase this year and 25 per cent in the last month alone. The AI boom is increasing demand for chips, and creating demand all the way up the supply chain from mining to power generation to data centre and hyper-scale campus construction. Let’s dig into the commodities that will be in play to see if the AI boom shifts the supply/demand profile beyond their current moves.
Data centres:
As of early 2026, there are approximately 12,500 operational data centres globally, with roughly 5,500 located in the United States and 350 in Canada. The U.S. continues to account for nearly half of global data centre capacity and industry forecasts suggest the number of U.S. data centres could approach 8,000 facilities by 2030.
Total installed computing capacity is expected to grow much faster than the number of physical sites as hyper-scale campuses become larger and more power intensive. A hyper-scale campus is a very large data centre complex designed and operated by major cloud and AI companies. Unlike a traditional data centre that might occupy a single building, a hyper-scale campus typically consists of multiple data centre buildings, dedicated electrical substations, massive backup power, advanced cooling infrastructure and fibre optic networks.
Commodities:
Copper is the single most important commodity beneficiary from what is happening with AI. AI servers consume significantly more power than traditional servers, requiring power distribution equipment, transformers, switchgear, cooling systems, backup generators and considerable electrical infrastructure. A large AI data centre can contain thousands of tonnes of copper.
Most forecasts suggest that data centres could add somewhere between 0.5 million and 1.0 million tonnes of copper demand by 2030, with additional copper required for the surrounding electrical grid infrastructure. Current global copper demand is roughly 26 million tonnes/year, in line with current supply. Annual AI demand for copper is projected to be over three million tonnes by 2050, with data centres consuming 6 to 7 per cent of global supply by then.
A technician works at an Amazon Web Services AI data centre in New Carlisle, Ind.Noah Berger/Reuters
Supply chain constraints, geopolitical risks and the slow pace of new copper mine development will further compound price increase risks. Increased demand from AI-related industries looks like it will add to a long-term bullish outlook for copper, leading to possible strategic stockpiling by tech firms and manufacturers.
The uranium market is already in a structural deficit with a shortfall of 19 million pounds of U₃O₈ or 11 per cent of current demand. As AI growth continues, many forecasts suggest nuclear generation will be needed because it is carbon-free, runs 24/7 and provides stable baseload power.
Several large technology companies, including Microsoft, Amazon and Google have announced investments (in small modular reactor technology) or partnerships (through power purchase agreements) aimed at securing long-term nuclear power supplies for future data centre operations.
A combination of existing reactor life extensions, new large-scale nuclear projects, and the deployment of small modular reactors will be needed to help meet the rapidly growing electricity requirements of AI infrastructure. As a result, uranium demand forecasts have strengthened, with many market observers anticipating sustained growth in nuclear fuel consumption through the 2030s and beyond. While most of the growth in uranium demand is expected to come from reactor construction in Asia and the extension of existing reactor fleets, the rapid growth of AI data centres is emerging as an additional demand driver.
Aluminum demand will rise alongside copper demand. High-voltage transmission projects increasingly utilize aluminum conductors because they are lighter, less expensive and more readily available than copper for long-distance power transmission. In addition, aluminum is widely used in data centre construction, including structural components, cable trays, heat exchangers, cooling systems and power distribution equipment. As hyper-scale AI facilities grow in size and electrical capacity, aluminum is expected to benefit not only from increased data centre construction but also from the substantial expansion of the electrical grid needed to support these energy-intensive operations.
Silver is used in high-performance electronics, semiconductors, electrical contacts and solar infrastructure (that may support data centre power demand). A modern AI server could contain 10 grams of silver. AI alone won’t drive silver demand dramatically, but will add to existing demand from electrification and solar.
Rare earth elements such as neodymium, praseodymium and dysprosium are needed for magnets, electric motors, cooling systems, fans and power electronics. Like silver, these are more of a secondary beneficiary to the AI data centre build out. The biggest supply risk for these rare earth elements is that processing capacity is concentrated in China, creating supply-chain vulnerability and the potential for sudden price spikes
What about the data centres themselves?
The biggest commodity demand from data centres is often not inside the chips – it’s the electricity, copper, steel, aluminum, concrete and cooling infrastructure needed. Beyond the metals commodities, we are seeing demand for electricity and/or natural gas, given a single modern AI campus can require as much electricity as a medium-sized city.
The AI build out and subsequent supply chain shows us demand will be increasing for a number of commodities, some of which are supply constrained currently. The impact looks to be drawn out over the next five to 25 years, as additional data centre capacity is brought online.
Brian Donovan, CBV, is the president of StockCalc, a Canadian fintech based in Miramichi, N.B.