Access to Critical Minerals is the Achilles’ Heel of Trump’s AI Ambitions

On June 11, the Trump administration announced a 55% tariff on Chinese imports. This came as a dramatic reversal from the initially proposed 145% rate proposed on “Liberation Day.” The reason: to restore America’s access to critical and rare earth minerals.

Beijing had responded to the “Liberation Day” tariffs by halting exports of rare earth metals to the US. The move sent shockwaves through American and European industries, especially the automobile and defense manufacturing sectors. This months-long trade war revealed just how dependent the US is on China for these critical resources.

Rare earths are a group of 17 elements indispensable to many advanced technologies, including artificial intelligence (AI). For example, terbium and dysprosium are vital to producing 3D scanning devices used in autonomous vehicles and robotics. Other critical minerals, including gallium and germanium, are equally important, especially for AI hardware.

Critical minerals are used in every layer of the AI stack. At the most fundamental level, advanced semiconductors require gallium compounds. Then, in data centers where AI models are trained, dysprosium prevents data storage devices from overheating. Lithium-based batteries store energy long enough to power autonomous vehicles and robots.

China dominates the production of critical minerals and rare earths. It controls 98% of the world’s supply of gallium and processes nearly 90% of all rare earths. The US produces no gallium and less than 2% of the world’s germanium. These minerals are key ingredients in US defense systems, making supply chain vulnerabilities a serious threat to US national security.

Less known is that the next wave of AI depends even more heavily on critical minerals. Leading AI companies are ushering in an era of ‘Physical AI.’ Tesla is investing heavily in its humanoid robot, Optimus. OpenAI’s first AI device is also in the works. Driverless Waymos are entering more and more cities. Citi predicts that there will be 1.3 billion AI-powered robots by 2035. Physical AI may soon become a permanent feature of our roads, our workplaces, and even our homes. In such a world, critical minerals supplies will be crucial to modern life.

Most people today experience AI through large language model (LLM) applications such as ChatGPT. While significant compute resources are required to train LLMs, their diffusion is relatively cheap and easy. Anyone can access ChatGPT on a computer or smartphone without buying a new physical device. On the other hand, democratizing Physical AI means every household might have its own AI home assistant and autonomous car. This puts exponential pressure on the demand for critical minerals and rare earths.

Consider humanoid robots. Their motors, vision sensors, and batteries all depend on critical minerals. Autonomous vehicles, most of which are electric, also use significantly more critical minerals than a conventional car. A critical minerals shortage will hamper America’s progress in Physical AI. Meanwhile, this bottleneck does not exist for China.

The US cannot hope to “win” the AI race, which is the stated policy of the Trump administration, without solving its minerals problem. As AI becomes increasingly prominent in Americans’ lives, the US’s dependence on China for critical minerals will always be its Achilles’ heel.

Earlier this year, Trump signed two executive orders targeting this challenge. The first aimed to boost US production of critical minerals, and the second ordered a more aggressive exploration of deep-sea resources. On August 13, the Department of Energy announced plans to commit $1 billion towards securing American critical minerals supply chains. The Department of Defense has spent over half a billion dollars on critical minerals projects, and recently became the largest shareholder of the only active rare earths mine operator based in the US. The Trump administration is also reportedly considering redirecting $2 billion in CHIPS Act funding towards minerals projects.

In addition to developing America’s domestic capabilities, the US must also work with international partners. Turning raw minerals into usable metals is both resource and capital-intensive. In particular, China has a chokehold on rare earth separation and processing, and operationalizing these production facilities is notably time-consuming. No country alone can overtake China’s minerals dominance. Brazil has the second largest rare earths reserves after China and is geographically closest to the US compared to other major reserves. However, its complex minerals licensing process, requiring around 16 years for approval, coupled with recent political tensions with the Trump administration, makes Brazil an unreliable option in the near-term.

Other countries with favorable political environments and robust mining infrastructure — such as Australia, Saudi Arabia, and Canada — are better-suited partners. The Australian government has invested heavily in its mining sector and seeks to become a leader in critical minerals production. In May 2025, Australian firm Lynas Rare Earths became the first company outside China to successfully produce dysprosium. Saudi mining company Maaden and US firm MP Materials also signed a memorandum of understanding to work together on developing resilient rare earths supply chains. Canada, too, is well-positioned to become a critical minerals power. In the first Trump administration, the US and Canada had already established a Joint Action Plan on Critical Minerals Collaboration.

China is already touting its strength in robotics development. It recently opened the world’s first humanoid robot store and hosted more than 500 robots at the inaugural World Humanoid Robot Games. With the shift towards Physical AI already underway, staying competitive in the AI race requires critical minerals security. Without strategic and bilateral critical minerals partnerships, the US risks ceding its AI leadership.