TheVortiq
Inteligencia Artificial

The Rare Earth Crisis: The Invisible Bottleneck of AI

China threatens to restrict access to critical materials for data centers, jeopardizing the global deployment of artificial intelligence infrastructure.

August 31, 2026 · 3 min read

aerial photo of desert

TL;DR: China controls the supply of critical materials like yttrium and erbium, essential for fiber optics and AI server infrastructure. With export laws under review in November, the tech industry faces a potential blockade that current strategic reserves cannot yet mitigate.

The Hidden Dependence of Digital Infrastructure

The race for hegemony in artificial intelligence is no longer a purely algorithmic contest. While media focus has centered on GPU computing power and the efficiency of Large Language Models (LLMs), the architecture supporting this digital edifice is vulnerable. According to reports from IEEE Spectrum, the rare earth supply chain has become the invisible bottleneck for the expansion of global data centers. The crisis is not hypothetical: data confirms a 75% drop in yttrium supply compared to 2025, an indicator revealing how geopolitical tensions are directly impacting the operationality of physical cloud infrastructure.

The Strategic Role of Yttrium and Erbium: The Foundations of the Network

The dependence is not merely commercial, but technical and structural. Yttrium and erbium are critical elements that lack scalable substitutes in the current industry. Yttrium is an essential component in the manufacturing of multi-layer ceramic capacitors (MLCCs), present in virtually all modern server printed circuit boards, in addition to being vital for solid-state lasers. On the other hand, erbium is the pillar of erbium-doped fiber amplifiers (EDFA), the technology that allows data signals to travel across oceans and data centers without losing power. Without these elements, the high-speed interconnection infrastructure that powers generative AI simply could not scale. China not only extracts 90% of global yttrium, but according to the International Energy Agency (IEA), it controls 91% of separation and refining stages worldwide. This absolute concentration makes Beijing the only viable supplier, creating a situation of asymmetric dependence unprecedented in the history of computing.

Consequences of the Trade War and the November Horizon

The current crisis is the direct result of the tariff escalation initiated in 2025 in response to U.S. restrictions on advanced lithography equipment from ASML. Beijing has implemented a mandatory licensing scheme that acts as a mirror to Washington's export control policies. The temporary suspension of restrictive measures—agreed upon in November 2025 after a brief diplomatic truce—comes to an end this month. The technical uncertainty is total: if China decides to reactivate or tighten controls on erbium and other materials, the tech market will face an immediate supply shock. Historically, this scenario recalls the 2010 crisis, when China restricted rare earth exports to Japan, although the current scale is exponentially larger due to the insatiable demand of AI clusters.

Technological sovereignty no longer resides solely in software, but in the ability to secure access to the elements of the periodic table that enable data processing.

U.S. Response and Long-Term Impact: An Inevitable Lag

The Trump administration has attempted to mitigate this risk through initiatives like Project Vault, endowed with $10 billion to ensure supply chain resilience. However, at TheVortiq we warn that there is an insurmountable time lag: mining and, above all, refining rare earths are processes that require years of infrastructure development and complex chemical processes. Although there are efforts to diversify supply to regions like Brazil or reactivate domestic refining capacity, these measures will not compensate for the immediate dependence that exists today in server racks. The reality is that the transition toward a diversified supply chain is a long-term process, while the demand for AI infrastructure is an urgent present need. Companies must prepare for an environment of high volatility where access to critical materials will be the new determinant of competitive advantage in the SaaS and cloud infrastructure market.

Keep reading