The US-China AI competition is increasingly being decided not by innovation or talent, but by control over physical materials and energy supplies that neither country can easily replace. While policymakers focus on export restrictions and visa policies, a more immediate crisis is unfolding: critical bottlenecks in tungsten, oil, and semiconductor manufacturing are creating vulnerabilities that could disrupt AI development on both sides of the Pacific .

What Is Tungsten and Why Does It Matter for AI Chips?

Tungsten is a metal with an exceptionally high melting point and density, making it essential for manufacturing advanced semiconductors. It appears both inside chips themselves and in the equipment used to produce them, particularly at the cutting-edge nodes required for artificial intelligence processors. Unlike many other materials, tungsten has no practical substitute at scale .

The problem is stark: China controls roughly 79% of global tungsten mine production. In February 2024, Beijing added tungsten to its export control list amid escalating US trade tensions. The results have been dramatic. Tungsten prices have skyrocketed to $2,250 per metric ton unit, representing a 557% increase in just over a year, outpacing gains in gold, copper, and even oil by a wide margin. Chinese exports of restricted tungsten products also fell by roughly 40% in 2025 .

BMO analysts have forecast another supply deficit in 2026. With production highly limited by geography and Western efforts to build alternative supply chains still years away from meaningful capacity, there is no near-term mechanism to offset the shortfall.

How Are Energy Disruptions Threatening Chip Production?

The geopolitical crisis extends beyond rare materials. Military conflict in the Middle East has disrupted oil shipments through the Strait of Hormuz, creating energy security risks that directly threaten semiconductor manufacturing. South Korea is particularly vulnerable, importing roughly 70% of its crude oil from the Middle East, with virtually all of it traveling through the Hormuz strait .

This matters because South Korea is home to Samsung and SK Hynix, two companies that together control approximately 80% of global HBM (high-bandwidth memory) production and roughly 70% of the DRAM (dynamic random-access memory) market. When energy supplies tighten, manufacturing and transportation costs rise. Whether chipmakers will absorb these costs or pass them on to buyers remains unclear, but either scenario threatens the affordability and availability of AI chips globally .

The Carnegie Endowment has observed that the conflict in Iran has exposed a long-standing vulnerability in the South Korean chip sector created by the country's dependence on imported fossil fuels. As the conflict persists beyond one month, the ripple effects are being felt across the entire AI supply chain.

Why Are Export Controls Failing to Stop China's AI Chip Demand?

Despite years of US export restrictions, demand for AI chips in China has not diminished. Instead, it has found shadowy new pathways. At Nvidia's GTC conference in San Jose, CEO Jensen Huang confirmed that the company has received a new wave of purchase orders from Chinese customers for its H200 processors and has restarted manufacturing to fulfill them. Chinese tech firms had collectively placed orders for more than two million H200 units for 2026 delivery, according to Reuters reporting .

This represents a significant shift. China once accounted for at least one-fifth of Nvidia's data center revenue. The announcement ended a roughly ten-month freeze on advanced chip exports to China, during which the Trump administration required export licenses and Nvidia took a $5.5 billion charge tied to stranded inventory. A December arrangement now allows H200 sales to approved Chinese customers, with the US government taking a 25% cut of revenues .

However, the legal reopening of channels has not eliminated the black market. A recently unsealed Department of Justice indictment reveals the scale of illegal smuggling when legitimate supply is constrained. US prosecutors charged three men with conspiring to smuggle approximately $2.5 billion in Supermicro servers containing restricted Nvidia GPUs to Chinese buyers. At least $510 million in hardware allegedly reached its destination .

The individuals charged include Yih-Shyan "Wally" Liaw, a co-founder of Super Micro Computer; Ruei-Tsan "Steven" Chang, its Taiwan sales manager; and contractor Ting-Wei "Willy" Sun. They face charges carrying maximum sentences of up to 30 years. The trio reportedly collaborated to sidestep the Export Control Reform Act by placing purchase orders as though the servers were destined for Supermicro's operations, then had a third-party firm repackage the hardware in unmarked boxes before concealing the trail with fabricated audit documentation .

How to Navigate Supply Chain Risk in the AI Era

• Diversify Material Sources: Companies should work to identify alternative suppliers for critical materials like tungsten and reduce dependence on single-source countries, even if it increases short-term costs.
• Monitor Energy Security: Organizations with significant chip manufacturing or procurement operations should track geopolitical developments affecting oil supplies and energy prices, particularly in the Middle East and other strategic regions.
• Secure Inventory Early: Procurement teams should identify exposure to supply constraints and secure critical components before allocation windows close, avoiding last-minute sourcing that drives up both cost and counterfeiting risk.
• Implement Compliance Controls: Companies must establish robust audit and documentation systems to ensure supply chain partners are not engaged in illegal export activities, protecting themselves from liability and reputational damage.

What Does This Mean for the Future of AI Competition?

The convergence of these pressures reveals a fundamental truth about the US-China AI race: it is not primarily a competition of algorithms or talent, but of physical resources and supply chain control. Export controls alone cannot address the problem when critical materials are geographically concentrated and energy supplies are vulnerable to geopolitical disruption .

For hyperscalers, original equipment manufacturers (OEMs), and memory chip buyers, the implications are immediate. Manufacturing and transportation costs are likely to rise if Middle East oil disruptions continue. Tungsten shortages will persist through 2026 and beyond. And the black market for restricted chips will remain active as long as legitimate supply is constrained or expensive.

The real vulnerability is not that China cannot get AI chips, but that the entire global semiconductor supply chain is becoming more fragile, more expensive, and more dependent on geopolitical stability. Both the US and China have strong incentives to stabilize these supply chains, yet both are also pursuing policies that destabilize them further. The result is a race to secure critical materials and manufacturing capacity that may ultimately benefit neither side.