As the artificial intelligence (AI) boom extends its reach into the resource sector, copper, rare earths, and strategic minor metals have emerged as the ultimate focal points. These three categories share a common rationale: demand is tangible, supply is inflexible, and pricing is dictated by structural factors. Copper is about bridging supply shortfalls, rare earths hinge on market penetration rates, and minor metals command scarcity premiums. The sharp decline in copper prices in mid-September, from all-time highs, effectively eliminated speculative premiums.

From AI chips to optical modules, from printed circuit boards (PCBs) to liquid cooling systems, every advancement in the AI narrative brings us closer to the physical world. While chips can be mass-produced, minerals are subject to geological constraints. The resource sector represents the final link in this chain—the last to be priced and the most resistant to easy refutation.

On the evening of September 10, London copper prices abruptly reversed course after reaching a record peak, plummeting nearly 4% intra-day and oscillating in subsequent days. Markets interpreted this downturn as a "cooling of the AI narrative," but what was truly being re-evaluated were loosening policy expectations. The plunge revealed a fundamental truth: copper's rally had been propelled by two pillars—tangible demand from AI and power grids, and speculative premiums driven by expectations. For months, these two forces had been intertwined in trading. When one weakened, the market misconstrued the entire structure as collapsing.

01. The Resource Sector: Priced Last Due to Inelastic Supply

AI's demand for metals is entirely physically driven: data centers require power, cooling, and cabling; humanoid robots need joint motors. Copper bears the brunt of this demand—a single AI training server consumes 3-6 times more copper than a standard server; a large AI facility demands 50,000 tons of copper; a single compute cabinet's liquid cooling plate uses 800-1,000 kg of pure copper. CICC projects that the AI data center industry will drive approximately 1 million tons of new copper consumption by 2026. Structural upgrades are even more profound: high-frequency PCBs necessitate low-profile, high-ductility copper foil; AI server motherboards and accelerator cards use 260% more copper foil year-on-year. Every link—power supply copper bars, connector copper alloys, liquid cooling plates—is transforming copper from a mere commodity into a functional component. The driver of copper price growth is shifting from volume to value-added.

Rare earths revolve around market penetration. Humanoid robot joint servo motors demand rapid response, high starting torque, and precise control—high-performance neodymium iron boron is indispensable. Coupled with demand from air conditioners, new energy vehicles, and the low-altitude economy, the average price of praseodymium neodymium oxide surged 72.67% year-on-year in the first half, with a clear upward shift in price levels. This logic mirrors that of growth stocks: demand hinges on penetration rates, while supply is constrained by quotas.

Strategic minor metals are all about scarcity. Tungsten is used in AI server PCB micro-drills and high-density alloys; germanium in fiber optics and infrared; gallium in compound semiconductors; antimony in flame retardants and military applications. Supply is highly concentrated with limited reserves, and output responds extremely sluggishly to price changes. By 2026, antimony prices are projected to rise approximately 600% from early 2025 levels; tungsten prices could surge over sixfold—price elasticity far exceeds that of traditional cyclical commodities, as markets reprice these as strategic assets.

The common thread across all three categories is their structural foundation: demand is initially priced on expectations, while supply can only be priced over time. It takes an average of 17.5 years from exploration to production for new mines. Global copper supply-demand gaps are projected to widen to 40,000, 350,000, and 430,000 tons annually from 2026-2028. Metals do not adhere to Moore's Law—ore grades and mine depths are fixed in geological reports. In past compute booms, resource stocks always rallied last and delivered last.

02. Why Zijin and China Molybdenum's Valuations Differ by 100% at the Same Copper Price

Setting narratives aside and focusing on the numbers, during the same copper price rally, two industry leaders exhibited vastly different valuations—Zijin Mining trades at approximately 16-21 times price-to-earnings (P/E) based on market capitalization, while China Molybdenum trades at approximately 12-13.5 times. Where does this disparity originate? Breaking it down, there are four key variables: production base, price elasticity, earnings delivery, and resource reserves.

The production base determines viability. Zijin plans to produce 1.2 million tons of copper by 2026 (1.09 million tons in 2025); China Molybdenum guides 760,000-820,000 tons, having produced 388,000 tons in the first half (+9.73% year-on-year). The leverage pivot is established here.

Price elasticity dictates the slope. For every $1,000/ton change in copper prices, amplified by production: Zijin's 1.2 million tons correspond to $1.2 billion in pre-tax incremental revenue (approximately 8.5 billion RMB); China Molybdenum's 800,000 tons correspond to $800 million (approximately 5.7 billion RMB). After income tax and minority interests, net elasticity is approximately 60-70%. Every $1,000/ton price increase triggers billions in profit revaluation—a mechanism precise to every 10,000 tons and every $1,000.

Earnings delivery determines valuation. Zijin's first-half net profit attributable to shareholders was 39.2 billion RMB (+68% year-on-year), with full-year consensus at 70-80 billion RMB; China Molybdenum's first-half net profit was 16.15 billion RMB (+86.27% year-on-year), with full-year consensus at 28-32 billion RMB. Market capitalizations: Zijin at 1.3-1.5 trillion RMB, China Molybdenum at approximately 378 billion RMB. In the same rally, different combinations of market capitalization and earnings naturally create valuation tiers.

Resource reserves determine the floor. The reserves-to-market capitalization ratio is the simplest benchmark: Zijin's copper resources number in the hundreds of millions of tons, with approximately 14,000 RMB in market capitalization per ton; China Molybdenum's copper resources number in the tens of millions of tons, with less than 10,000 RMB per ton. This ratio reflects mine quality and location—grade, cost, and mine life are all embedded in the denominator.

Integrating these four variables into a unified framework yields a clear conclusion: the framework is shared, but parameters vary. This model applies to any resource company, calculating its unique numbers. The numbers do not dictate trades—they transform vague questions of "overvalued or undervalued" into verifiable calculations.

On the industrial chain, a company's position determines which segment of elasticity it captures. Upstream miners capture price elasticity; companies with high self-sufficiency and large output are direct beneficiaries of rising copper and rare earth prices. China Molybdenum achieved record first-half profits, while Zijin is regarded as the sector leader with its copper-gold dual mainstay. Jiangxi Copper and Western Mining provide elasticity through reserves and output; rare earth resources are controlled by Northern Rare Earth, China Rare Earth, and GHS Rare Metal. Midstream materials capture penetration rates: copper foil firms benefit from AI servers' demand for low-profile, high-ductility products—Tongguan Copper Foil, Jiayuan Technology, and Nord Holding have carved out positions. Magnetic materials firms benefit from robot orders—JL Mag, Zhenghai Magnetic Material, and China Three Rings are mainstream choices. Minor metals capture scarcity: Hunan Gold and Huaxi Rare Metal have dual mainstays in antimony and gold; Xiamen Tungsten, Zhangyuan Tungsten, and China Tungsten High-Tech cover the full tungsten chain; Yunnan Germanium and Chihong Zn&Ge control germanium resources; Eastern Tantalum enjoys a monopoly on tantalum-niobium scarcity.

03. Three Lines, Three Ledgers: Copper Awaits Supply Gaps, Rare Earths Volume Growth, Minor Metals Disruptions

Copper follows a long-term ledger, with rhythm dictated by AI capital expenditures (capex) and inventory cycles—track capacity utilization and price premiums/discounts. Rare earths follow a growth ledger, with rhythm tied to humanoid robot shipments and quota policies—track orders and inventories. Minor metals follow a speculative ledger, with prices hypersensitive to supply disruptions—the highest volatility and strictest discipline on entry/exit timing.

All three lines share an anchor for judgment: when the AI narrative truly shifts, watch for three signals—whether AI capex is slashed, whether inventories accumulate continuously, and whether spot premiums turn to discounts. As of mid-September, none had occurred. The mid-September copper price plunge squeezed out speculative premiums, but supply gap structures, hardware production schedules, and copper foil orders showed no directional changes.

AI's endpoint is electricity; electricity's endpoint is metals. Chip computing power can be diluted by Moore's Law, but mineral supply is constrained only by geology and time. Copper awaits supply gaps to materialize; rare earths await robot volumes to surge; minor metals await strategic revaluation—each line moves to its own rhythm. This AI rally, having reached the resource sector, is no longer about narratives—it's about ledgers. Markets paid a premium for "AI"; they will reprice for real supply gaps. When trading noise fades, prices will return to centers defined by supply gaps, scarcity, and penetration rates.

- End -