Ruthenium Supply Shock: AI Chip Shortage Exposes Critical Bottleneck in Semiconductor Metals

Ruthenium prices hit $1,750 per ounce on March 13, tripling year-over-year as AI-driven demand for advanced chip packaging collides with supply concentrated in South Africa and Russia. The metal, essential for next-generation interconnects and 3D chip stacking, now represents a binding physical constraint on AI scaling—one that cannot be coded around or financed away.

Supply Cliff Meets AI Infrastructure Boom

Ruthenium jumped from $560 per ounce in March 2025 to $1,750 by March 2026, per Reuters data citing Johnson Matthey benchmark prices. Wilma Swarts, director of platinum group metals at Metals Focus, projects a 203,000-ounce deficit in 2026—roughly 6.3 metric tons against global production of approximately 35 metric tons annually.

South Africa supplies 91% of global ruthenium, with Russia and North America accounting for the remainder, according to Statista 2021 data. Over 90% of primary production comes as a byproduct from platinum and nickel operations in South Africa, Russia, and Canada, meaning supply is inelastic to ruthenium’s own demand and instead dictated by host-metal economics.

South African platinum-group metal output fell 3.8% year-over-year in January 2025 due to electricity supply disruptions and rail transport issues, compounding structural underinvestment. Russian operations, led by Norilsk Nickel, contribute over 25% of global ruthenium output, but sanctions complexity and trade flow redirection to Asia create procurement uncertainty for Western buyers.

Semiconductor Imperative: Copper’s Successor at 2nm

Ruthenium’s price explosion reflects its emerging role as copper’s replacement in leading-edge chip interconnects. IBM Research and Samsung demonstrated ruthenium interconnects that require no diffusion barrier and can scale to smaller widths without astronomical resistance increases, addressing copper’s fundamental scaling limits below 20nm pitch.

Chipmakers are likely to replace copper at some metal levels with ruthenium or molybdenum as soon as the 2nm node, with ruthenium deployment expected at 1nm (20nm metal pitch), according to Semiconductor Engineering analysis. Applied Materials introduced an industry-first ruthenium-cobalt liner combination that reduces electrical line resistance by up to 25% and enables copper scaling to 2nm and beyond.

Advanced packaging programs under the US CHIPS Act encourage investments in chiplet-based architectures, heterogeneous integration, and advanced interconnects—all ruthenium-intensive technologies. Bloomberg Intelligence projects the advanced semiconductor packaging market will grow eightfold to $80.5 billion by 2033, with a 26% compound annual growth rate significantly outpacing the 10% projected for overall Semiconductors.

3D Stacking Multiplies Demand

Chiplet disaggregation and vertical integration amplify ruthenium intensity beyond traditional 2D scaling. Wafer-to-wafer hybrid bonding, a key technology for 3D system-on-chip integration at micrometer interconnect density, yields substantially smaller pitches compared to 2.5D microbumps, driving higher metal loading per die stack.

TSMC’s CoWoS advanced packaging capacity is expanding from 330,000 wafers in 2024 to 660,000 in 2025—a 100% annual increase driven by Nvidia, AMD, AWS, Broadcom, and cloud service provider demand, per IDC semiconductor Supply Chain intelligence. Each wafer processed consumes ruthenium for barrier layers, interconnect metallization, and via fill across multiple die-to-die interfaces.

The AI server market is forecast to climb from $140 billion in 2024 to $850 billion by 2030, according to Creative Strategies’ “giga cycle” analysis. Each gigawatt of AI data center capacity requires approximately $42 billion in capex including $23.6 billion in chips alone, with hyperscalers deploying 20-30 GW through 2030. That translates to trillions in semiconductor demand, much of it at nodes where ruthenium becomes essential.

Geopolitical Leverage and Margin Pressure

South Africa’s mining industry faces electricity supply instability, deep-level operating risks, and policy uncertainty, while Russian platinum-group metal production decreased in 2024 due to natural disasters, lower metal grades, and issues related to the Russia-Ukraine conflict. These disruptions hand producing nations pricing power at a moment when semiconductor supply chain resilience is a strategic priority.

The EU removed Russian refiners from the London Platinum and Palladium Market delivery lists in 2022, establishing precedent for platinum-group metal restrictions. The EU is now considering a ban on Russian imports of several platinum-group metals and copper as part of new sanctions targeting Moscow, per Bloomberg reporting from February 2026. Ruthenium was not explicitly listed, but supply chain bifurcation would further tighten Western access.

Foundries face direct cost pressure. Samsung optimized a ruthenium-cobalt liner to improve copper gap fill at 3nm, achieving 87% fewer voids and 14% better line resistance, but ruthenium procurement at current levels versus historical $400–600 ranges erodes process economics. TSMC, Samsung, and Intel compete for allocations from the same narrow supplier base, creating potential bidding wars that inflate wafer costs and compress chipmaker margins.

No Quick Substitutes

Molybdenum precursors are an order of magnitude less expensive than ruthenium, but neither material will likely be needed before the 2nm node, and full data sets to choose between them remain incomplete, according to imec researchers. Cobalt serves as an interim solution for some applications but cannot match ruthenium’s resistivity profile at sub-10nm dimensions.

Annual ruthenium supply totals around 30 metric tons globally, with scarcity making prices highly sensitive to minor supply disruptions or demand fluctuations. Recycling rates remain low—below 30% for end-of-life consumer electronics—due to inadequate collection systems and the complexity of separating platinum-group metals from diverse waste streams. Scaling secondary supply to offset primary deficits requires multi-year infrastructure investment.

What to Watch

N2 node ramp timing: TSMC began 2nm risk production in July 2024, targeting second-half 2025 mass production. Volume shipments in 2026 will establish ruthenium consumption baseline for leading-edge logic.

South African output stabilization: Eskom’s power delivery and rail logistics performance through mid-2026 will determine whether the 203,000-ounce deficit widens or moderates. Any labor disputes at Anglo American Platinum or Impala Platinum operations would exacerbate shortages.

Sanctions evolution: EU deliberations on Russian platinum-group metal restrictions, expected in February 2026, could formalize supply bifurcation or leave Western buyers in regulatory limbo.

Foundry pricing actions: Watch for TSMC, Samsung, and Intel wafer price adjustments in Q2 2026 earnings calls. Ruthenium cost pass-through to fabless customers would ripple into AI accelerator economics and hyperscaler capex models.

Recycling infrastructure investment: US and EU programs under the CHIPS Act and Critical Raw Materials Act could accelerate platinum-group metal recovery from electronics waste, but capacity additions won’t materialize before 2027–2028.

The ruthenium constraint differs from cyclical chip shortages: it’s a materials-science problem with decade-long supply cycles meeting quarterly demand volatility. Unlike fab capacity, which can be financed, or software optimization, which can be engineered, ruthenium availability is governed by geology, geopolitics, and the economics of host metals mined half a world away. AI scaling now depends on whether a 30-ton-per-year specialty metal market can triple output before 2nm nodes hit volume production.