Ruthenium, a rare platinum-group metal that spent decades as an obscure industrial byproduct, has become one of the hottest commodities of the artificial intelligence era. By late July 2026, the metal traded at approximately $62.59 per gram, a cumulative gain of 787 percent since January 2018, when it changed hands for just $7.05 per gram. The global ruthenium market is now valued at over $740 million and is projected to surpass $1.42 billion by 2034, growing at a compound annual rate of 7.5 to 8.1 percent. The rally's decisive inflection point came between 2024 and 2025. After drifting from $20.28 per gram in early 2022 down to $16.20 by January 2024, the price exploded 162.57 percent in a single year to reach $48.68 by January 2026, before climbing another 28.57 percent in the first seven months of the year. The timing was no coincidence: the surge tracked precisely with the semiconductor industry's mass validation of ruthenium chip interconnects and the commercial rollout of new AI-driven data storage technologies.
Regional pricing tells the same story of scarcity. In China, the world's largest consumer of semiconductor precursors and industrial catalysts, 99.95 percent purity ruthenium stabilized at a premium of 370 RMB per gram in June 2026, while Johnson Matthey's global base price index recorded $1,675 per troy ounce. Intraday trading data from Heraeus Precious Metals in July 2026 revealed hourly price swings in the European market spanning more than fourteen euros per gram, the signature of a thin, illiquid market where buyers vastly outnumber available ounces.
Three demand vectors, all tied to the AI and clean energy boom, have converged on the metal simultaneously. The most powerful driver comes from advanced chipmaking. As fabrication nodes shrink below 10 nanometers toward 2 nm architectures, copper, the industry's workhorse interconnect metal for a quarter century, is failing physically. At those dimensions, wire widths fall below copper's 40-nanometer electron mean free path, causing electrons to scatter off grain boundaries and sidewalls, spiking resistivity and bottlenecking AI processors. Ruthenium outperforms copper below line widths of 10 to 17 nanometers, resists diffusion without bulky barrier layers, and can be subtractively etched to enable air-gap architectures. Intel has built it into its 20A and 18A nodes, with TSMC, Samsung, and IBM pursuing parallel schemes. For the tightest sub-2nm logic interconnects, no physical substitute currently exists.
The second shock comes from data storage. Training large language models requires petabytes of capacity, accelerating the industry's shift to heat-assisted magnetic recording hard drives. These drives depend on ruthenium in multiple layers as epitaxial templates for the iron-platinum recording media, as thermal management interfaces that channel laser heat away in under a nanosecond, and as spacer layers that enable multi-level 3D recording. AI-targeted data center deployments have already driven a reported 9% increase in orders for ruthenium-based components for high-capacity drives.
The third front is the hydrogen economy. Ruthenium sits at the apex of the catalytic "volcano plot" for ammonia cracking, the reaction that releases hydrogen from its most practical carrier molecule. While cheap iron and nickel catalysts require temperatures above 900 degrees Celsius, ruthenium achieves near-complete conversion at just 400 to 500 degrees Celsius, and advanced membrane reactor systems have reduced residual ammonia to below 1 part per million. As green hydrogen mandates scale up in India and the European Union, the energy sector is now bidding against chipmakers for the same scarce metal.
What makes this rally structural rather than speculative is the peculiar economics of ruthenium supply. Global primary production is capped at roughly 30 to 40 metric tons annually, and the metal is never mined for its own sake; it is recovered exclusively as a minor byproduct of platinum and palladium mining, meaning production is fundamentally inelastic to ruthenium's own price. No matter how high the price climbs, no one digs a ruthenium mine.
Over ninety percent of primary supply comes from South Africa's Bushveld Igneous Complex, with most of the remainder from Russia, where sanctions affecting Nornickel's Norilsk-Talnakh operations have further tightened flows. The richest ruthenium source is South Africa's UG2 reef, which carries up to 12 percent ruthenium in its PGM basket, but its fine-grained, chromite-heavy ore limits recoveries to 80-85 percent, with significant ruthenium lost to tailings. Worse, the supply base is actively shrinking. Platinum and palladium, which constitute nearly ninety percent of PGM demand, are consumed primarily by automotive catalytic converters. As battery electric vehicles displace internal combustion engines, prices for those primary metals have stagnated, compressing margins at Sibanye-Stillwater, Anglo American Platinum, and Impala Platinum. Squeezed further by soaring electricity costs from South Africa's fragile Eskom grid, the majors have responded with shaft closures, labor restructuring, and deferred expansions. The result is a crippling paradox: the mining sector is scaling back the only operations capable of producing ruthenium at exactly the moment the AI industry needs it most.
Recycling offers only partial relief. Secondary recovery currently accounts for less than fifteen percent of supply, and the metal's 2,250-degree melting point and near-total resistance to acid dissolution make reclamation extraordinarily difficult. Refiners must resort to complex smelting flowsheets or chlorination roasting at up to 850 degrees Celsius while managing ruthenium tetroxide, a volatile, toxic, and explosive intermediate that reacts violently with common organics. The capital and regulatory burden of handling it safely limits the number of facilities worldwide that can process ruthenium scrap, capping how fast urban mining can scale even at $60-plus per gram.
With Johnson Matthey forecasting continued deficits in ruthenium, platinum, and iridium through 2026, and no strategic government stockpiles existing outside the major PGM producers, downstream manufacturers have been forced into defensive positions. Buyers in the Asia-Pacific region, which consumes roughly 47 percent of global electronics supply, have moved aggressively into inventory restocking and long-term offtake agreements, locking up future production and cementing a high pricing floor beneath the market.
The pressure is already rippling through the AI hardware ecosystem. While the mass of ruthenium in any single chip is microscopic, the high cost of ultra-pure organometallic deposition precursors and sputtering targets, multiplied across the enormous chip volumes required for global AI training infrastructure, aggregates into meaningful bill-of-materials inflation. Sustained high prices are also accelerating substitution research: molybdenum, cheaper and barrier-free, is poised to capture the memory-chip interconnect market, leaving ruthenium ring-fenced for the critical logic routing and high-density storage layers where nothing else works.
The ruthenium rally is not a bubble; it is the market repricing a metal whose strategic importance has fundamentally and permanently changed. The physics of sub-2nm transistors, the thermodynamics of AI data storage, and the catalytic demands of green hydrogen have all converged on ruthenium at once, while the supply side remains hostage to the declining fortunes of the internal combustion engine. High prices cannot summon new mines because ruthenium's fate is chained to the platinum-and-palladium economics that the EV transition has broken. Barring a major technological substitution breakthrough, the structural deficit will persist, prices will remain elevated and volatile, and securing ruthenium will shift from a procurement afterthought to a boardroom-level strategic imperative for chipmakers, data center operators, and energy companies alike. The era of cheap, accessible ruthenium is over.
