AI's 66-Fold Data Center Boom Could Trigger a Critical Minerals Supercycle-But the Chips Are Only the Beginning

Highlights

• AI data centers depend on rare earth magnets, germanium, gallium, copper, and graphite—not just GPUs—making mineral supply chains a hidden bottleneck.
• Structure Research projects AI-focused data center capacity will surge from 2.3 GW today to 150 GW by 2030, with power, financing, and infrastructure as key constraints.
• Most critical mineral refining and downstream manufacturing remains concentrated in China, posing significant supply chain risk for Western AI infrastructure buildout.
• The AI revolution is a mine-to-data-center story: investors who understand the full industrial ecosystem stand to identify strategic value beyond chip designers.
• North America, Europe, and allied nations must scale separation, refining, and recycling of critical minerals at industrial levels to support AI infrastructure growth.

Every investor knows artificial intelligence runs on GPUs. Far fewer realize it also runs on rare earth magnets, germanium-doped fiber optics, gallium semiconductors, copper, graphite, and a growing list of strategic critical minerals. As AI infrastructure accelerates, the race for computing power is quietly becoming a race for materials. Rare Earth Exchanges® believes the companies controlling those supply chains may prove just as important as the companies designing the next generation of AI chips.

Every AI Data Center Begins in a Mine

The semiconductor captures the headlines. The supply chain that keeps it powered, connected, and cooled does not. Halfway around the world from Silicon Valley, miners, refiners, chemical processors, and magnet manufacturers determine whether AI infrastructure can actually be built. That reality becomes clearer in a new report from Structure Research (opens in a new tab), which projects AI-focused data center capacity will expand from approximately 2.3 gigawatts today to 150 gigawatts by 2030. Using a bottom-up methodology that reconciles infrastructure ownership with compute demand, the firm argues that power availability, financing, and infrastructure will become the industry's defining constraints.

Rare Earth Exchanges agrees with the direction of travel, although the precise 150 GW forecast ultimately depends on electricity availability, permitting, semiconductor supply, and capital deployment.

The Critical Minerals Hidden Inside AI

What the report largely leaves unexplored is the extraordinary mineral intensity required to build that future. As REEx has previously documented, AI's fiber-optic expansion is driving strong demand for germanium, while advanced cooling systems increasingly depend on neodymium-iron-boron permanent magnets enhanced with dysprosium and terbium. AI also drives growing demand for gallium in advanced electronics, copper for power transmission and networking, graphite, lithium, nickel, cobalt, and manganese for grid-scale battery systems, and high-purity aluminum, titanium, and specialty steels for data center construction.

Most of these supply chains remain heavily dependent on Chinese refining and downstream manufacturing.

Rare Earth Exchanges Take

Structure Research correctly identifies electricity as AI's principal bottleneck. Water could also be a material bottleneck depending on location. We would add another: industrial materials and their supply chains. The challenge is no longer whether enough critical minerals exist underground. It is whether North America, Europe, Japan, South Korea, Australia, and trusted partners can separate, refine, manufacture, recycle, and commercialize them at industrial scale.

The AI revolution is no longer simply a semiconductor story—it is a mine-to-data-center story. Investors who understand the entire industrial ecosystem—not just the chips—will be better positioned to identify where the next generation of strategic value is likely to emerge.

Register today: REEx Marketplace™ (opens in a new tab)