America's AI industry faces a critical vulnerability that has nothing to do with chip design or computing power: the minerals needed to build data centers and power them are overwhelmingly controlled by China and other countries with unreliable ties to the U.S. While policy discussions focus on competition among AI models and cloud providers, the real bottleneck lies in refined minerals like gallium, rare-earth metals, and cobalt, which are essential for everything from semiconductor circuits to cooling systems and backup power supplies.

Why Are Minerals the Real Chokepoint in AI Infrastructure?

The concentration of mineral processing is staggering. China processes approximately 99% of gallium, which is used in circuits that deliver power to chips at the correct voltage. The country also refines 87% of the silicon that forms the base of all chips, 90% of rare-earth metals used in cooling fans and pumps, 79% of tungsten used in AI chips, and 93% of graphite used in battery banks that protect data centers from power grid outages.

Copper presents another critical dependency. A single new data center drawing one gigawatt of power would require approximately 27,000 tonnes of copper for busbars, switchgear, and cables. While copper mines exist across multiple countries including Chile and Peru, China still refines about half of the world's copper supply.

The U.S. imports nearly 100% of its gallium and natural graphite, 70 to 80% of its rare-earth metals, and about 76% of its cobalt. Cobalt, which is essential for gas turbines and lithium-ion backup batteries, is mined primarily in the Democratic Republic of the Congo, but China refines 79% of it.

How Has China Already Used Mineral Exports as Leverage?

The vulnerability is not theoretical. In 2023, China placed gallium and germanium under export licensing in response to U.S. controls on chip exports. In December 2024, after the U.S. expanded those controls, China banned direct shipments of these metals to the U.S. Between 2021 and 2024, U.S. imports of germanium fell by 68% and those of gallium by 77%.

However, the U.S. has continued to receive these minerals through indirect channels. China's exports of germanium to Belgium rose by 224% between 2022 and 2024, and Germany's share in U.S. imports of gallium rose more than tenfold, suggesting that materials are being rerouted through third countries to circumvent export controls.

The economic stakes are enormous. The U.S. Geological Survey estimates that if rerouting channels are closed, domestic gallium prices could rise by approximately 150% and germanium prices by 26%. Such price spikes would disrupt the entire AI supply chain and could shrink U.S. GDP by about $3.4 billion.

What Minerals Are Critical to AI Data Centers and Why?

• Gallium: Used in circuits that deliver power to chips at the correct voltage; China processes 99% of global supply, and the U.S. imports nearly 100% of its gallium.
• Rare-Earth Metals: Essential for magnets in cooling fans and pumps that prevent data centers from overheating; China refines 90% of global supply, and the U.S. imports 70 to 80% of its rare-earth metals.
• Graphite: Used in battery banks that protect AI data centers from grid outages; China refines 93% of global supply, and the U.S. imports nearly 100% of its natural graphite.
• Tungsten: Used in AI chips as contact plugs and vias that connect transistors to metal wiring layers; China produces 79% of global supply.
• Copper: Used in busbars, switchgear, and cables; a single gigawatt data center requires approximately 27,000 tonnes, and China refines about 50% of global supply.
• Cobalt: Essential for gas turbines and lithium-ion backup batteries; mined primarily in the Democratic Republic of the Congo, but China refines 79% of global supply.

How Are Big Tech Companies Securing Their Own Energy Sources?

The largest AI companies have recognized the vulnerability and are taking action to secure their own power supplies, which further concentrates resources among the biggest players. Microsoft recently signed a 20-year contract to restart the Three Mile Island nuclear power plant specifically to power its data centers. Amazon has secured a contract for 1.92 gigawatts from Susquehanna nuclear power plant through 2042. Google has ordered small modular reactors, though these are not expected to generate electricity until 2030 at the earliest.

These long-term contracts allow large firms to claim scarce turbines and transformers, securing not just the power but also the equipment needed to build data center sites. If U.S. mineral supply chains are disrupted, the largest AI companies will be least affected because they have the capital to stockpile resources and secure alternative supplies. Smaller competitors and emerging AI firms lack this advantage.

Steps to Address America's Mineral Supply Chain Vulnerability

• Trade Policy Intervention: The U.S. government can negotiate trade agreements and establish strategic reserves of critical minerals to reduce dependence on single-source countries and create buffer stocks against supply shocks.
• Industrial Policy Investment: Federal support for domestic mining, refining, and processing capabilities can reduce reliance on China for minerals like gallium, rare-earth metals, and graphite, though building these industries takes years.
• Supply Chain Diversification: Encouraging investment in mineral processing facilities in allied countries such as Japan, Germany, and the Netherlands can create alternative sources and reduce concentration risk.
• Stockpiling and Strategic Reserves: Building government and industry stockpiles of critical minerals can cushion against sudden trade disruptions, as stockpiles of germanium have already helped the U.S. weather recent export restrictions.

In November 2025, as part of a trade truce, China temporarily suspended its ban on gallium and germanium exports to the U.S., shifting to case-by-case sales with a continued blanket ban on sales to U.S. military end-users. This suspension is set to expire in November 2026, leaving the long-term vulnerability unresolved.

The mineral concentration problem reveals a paradox in AI policy. While regulators focus on competition among AI model makers and cloud providers, the real chokepoint lies in the physical materials and energy infrastructure that underpin the entire industry. Solving this requires not just competition policy but coordinated trade and industrial strategy to build resilient, diversified supply chains that can withstand geopolitical tensions.