Urban Mining
and Resource Recovery

The copper tally alone showed Japan's accumulated in-use stock exceeding the measured reserves of many producing mines in Chile and Australia. The concept that emerged from this and from parallel work in Europe is urban mining: treating the technosphere, the roughly 30 trillion tons of human-made stuff on the planet, as a resource deposit assessable by the same criteria you would apply to a greenfield mineral prospect, meaning tonnage, grade, spatial distribution, and extraction cost.

A server circuit board runs 18 to 22 percent copper and 250 to 350 ppm gold. Grasberg, one of the richest mines operating, works ore at about 1 percent copper and under 1 gram gold per ton. Everyone who writes about urban mining cites this comparison, and everyone who cites it seems to think it clinches the argument, when what it actually does is set up the puzzle that the rest of the field is trying to solve. The board is richer per ton, and the board is also scattered across five million households in a metropolitan area, in devices made by hundreds of manufacturers using different solder alloys and laminate chemistries and component packages, and collecting a thousand tons of it requires an infrastructure that looks nothing like a mine haul road.

The economics of electronic waste processing are driven by precious metals, and this is not a minor detail or a footnote to the main story. It is the organizing principle of the entire sector, and most of the policy literature either does not engage with it or mentions it in passing before moving on to more comfortable topics like collection targets and extended producer responsibility.

Umicore runs an integrated copper smelter-refinery at Hoboken in Belgium that is one of maybe a dozen facilities worldwide capable of processing complex electronic scrap. When Hoboken accepts a shipment of circuit board scrap, the purchase price is calculated from an assay report, and the assay that matters is precious metals: gold on the edge connectors, palladium in the MLCCs, silver in solder and conductive adhesive. The copper is the process vehicle, not the revenue source. During electrolytic refining, copper deposits on the cathode and the precious metals concentrate in anode slime, which Hoboken routes to its own precious metals refinery on the same site. In a batch of high-grade server boards, the precious metal fraction can exceed 80 percent of total recoverable value per ton. A ton of boards from a data center decommissioning might assay at $8,000 to $12,000 in precious metals, with the exact figure varying by board generation because Intel boards from 2014 have different gold loading than Dell boards from 2018 than Cisco telecom cards from 2016, and Hoboken prices each lot off its own assay rather than paying a flat rate for a product category.

The flip side is that a washing machine or a laser printer, mostly steel and plastic with a small control board carrying negligible precious metal content, costs more to smelt than the recovered metals are worth. The smelter will process it if someone pays a gate fee. That fee comes from EPR levies, municipal budgets, or consumer disposal charges depending on the country, and what it means in practice is that the majority of electronic waste by mass, the low-grade fraction, moves through the recycling system only because someone is subsidizing a money-losing operation, while the high-grade fraction needs no regulation and never did.

The pricing differential plays out in scrap yards and container shipping manifests in ways that policy analysts rarely track. A scrap aggregator in Guangdong doing business with a European e-waste collector will cherry-pick the boards with high gold loading and decline the rest. The collector then needs to find another outlet for the remaining steel-heavy, precious-metal-poor fraction, and if the EPR scheme in the collector's country does not fully cover the processing cost, the material sits in a warehouse or gets shipped to a mechanical shredder that recovers the steel and copper and landfills the plastic and ceramic fraction. At each stage, the high-grade material is pulled toward precious metals recovery and the low-grade material drifts toward the cheapest disposal option available. This sorting happens continuously, at every transaction point in the chain, driven by assay values rather than by regulatory intent.

This subsidy structure is also weakening over time. Connector gold plating has gotten thinner with each product generation. Palladium is being replaced in some MLCC formulations. Miniaturization means less absolute precious metal mass per board. Smelters have responded by tightening their feed acceptance specs and turning away material they would have bought fifteen years ago. The fraction of e-waste that generates positive revenue at the smelter gate is shrinking, which means a larger and larger share of the electronic waste stream depends on external subsidy for any processing at all, and the trend is structural because it tracks product design evolution that nobody is going to reverse for the sake of making recycling economics work better.

Indium is a $250-to-$400-per-kilogram metal that cannot be economically recovered from end-of-life products. Sitting with that contradiction for a while is useful because the same mechanism, at varying severity, explains why gallium, tantalum, neodymium, and a dozen other specialty elements have functional recycling rates below 5 percent despite being designated "critical" by every government that maintains such a list.

The indium in an LCD display is present as indium tin oxide, sputtered onto the glass substrate in a film about 150 nanometers thick. A 42-inch television holds about 0.3 grams. The recovery chemistry is well established: HCl leaching dissolves the ITO, solvent extraction or cementation separates indium from tin, and further purification yields commercial-grade metal. Groups at NIMS in Tsukuba, KU Leuven, and KIST in Seoul have published complete process flowsheets, and Fraunhofer has run pilot lines. The chemical engineering has been solved multiple times over, and the remaining barrier is elsewhere.

Collecting a television, transporting it, separating the LCD panel from the housing, running the panel through acid leaching, and processing the leachate through separation and purification: the cost of doing this for one unit vastly exceeds 8.5 cents. Scaling up does not fix the problem because collection and disassembly costs scale linearly with unit count while chemistry costs come down only modestly. The Belgian and Japanese pilot operations confirmed both the technical viability and the economic impossibility at the same time.

The reason this matters beyond indium is that the gap is not a policy failure or an investment shortfall. Someone designed the ITO coating at 150 nanometers because that is the minimum thickness for the electrode to function. Someone chose glass because it is the right substrate for the display optical stack. Someone distributed the product to hundreds of millions of households because that is the business model. Each of these decisions was good engineering and good commerce, and their combined effect was to disperse a trace quantity of a moderately valuable metal across a continental-scale population of consumer products in a form bonded to its substrate at the atomic level. The thermodynamics of reconcentrating material from that state of dispersion impose energy costs that exceed the commodity value of the recovered metal, and this is a physics constraint rather than an economics one, meaning it does not respond to the usual policy levers of subsidy, mandate, or R&D funding. Process optimization can shift the cost curve at the margin without closing a gap whose origin is in the second law.

At the pilot plant level, Fraunhofer and the Japanese national labs can and do recover indium from display waste, and the technical literature contains dozens of optimized flowsheets, and none of these demonstrations has produced an operation that sustains itself financially. The per-unit revenue from 0.3 grams of indium does not cover the per-unit cost of extraction, and scaling up fails to fix this because collection and disassembly costs grow linearly with unit count while the chemistry gets only marginally cheaper at volume. Gold recovery from circuit boards does not have this problem because gold's per-gram value is high enough to absorb handling costs. Indium's is not.

UNEP assessed 60 metals in 2011 and found indium below 1 percent recovery, gallium in the same range, rare earths as a group the same, and tantalum in the low twenties, with the tantalum figure driven mostly by manufacturing scrap rather than consumer product recovery. The results track what thermodynamics would predict and have not changed much since.

Grade destruction is fast. A mixed electronics batch goes through an industrial shredder in about thirty seconds and comes out as a particle size distribution in which a tantalum capacitor is no longer distinguishable from a steel bracket or a cable fragment. Downstream separation equipment, magnets, eddy current, air tables, optical sorters, recovers some fractions. The fine material below 2 to 4 millimeters, where the specialty metals disproportionately end up because the components that contained them were small to begin with, goes one of two ways: to a copper smelter, where gold and palladium are partially captured and everything else enters slag, or to landfill. Umicore has published slag composition data, and the tantalum and neodymium are in there, quantifiable, permanently locked in silicate.

Japanese recyclers feeding Dowa and Mitsubishi Materials pull out batteries and some capacitors and identifiable ICs before shredding, and for uniform high-grade feeds from data center tear-downs this clearly pencils out. For the municipal collection bin full of mixed consumer junk, sorting by hand does not cover labor cost in any country where labor is expensive, so the default in practice is the shredder, and what the shredder produces is a mixture that feeds the copper smelter for partial precious metal recovery while everything else goes to slag or landfill.

The Royal Society of Chemistry estimated tens of millions of unused devices in UK households. People keep old phones because they worry about data residue, because the device might be useful someday, because the collection point is out of the way, because there is no financial incentive to bother. Deposit-refund systems for beverage containers get return rates above 90 percent in Germany and Scandinavia from exactly the same populations that let electronic devices accumulate in kitchen drawers, which isolates the incentive variable pretty cleanly.

Nobody has tried it for phones at scale.

The informal waste sector figured out collection economics on its own. In Accra, in Delhi, in Guangdong before the crackdown starting around 2015, waste pickers and itinerant buyers built networks that reached into neighborhoods at a depth no municipal schedule matches, because their income tracked directly with volume gathered. The processing was dangerous, cable fires, acid in the open, contaminated dust everywhere. When China shut Guiyu down, local pollution fell and so did collection, and the material that informal operators would have picked up sat in households or went to landfills where nothing got recovered. Colombia's Constitutional Court tried a different approach, mandating integration of Bogotá's waste picker cooperatives into the formal system, and Brazil gave catadores legal standing. Both attempts preserved some collection reach while pushing toward safer processing, and neither has fully sorted out the tension between the informal sector being better at gathering material and the formal sector being better at not poisoning people with it.

EV battery recycling was financed on cobalt, which trades at roughly thirty times the per-kilogram price of lithium carbonate and dominated the recovered value in NMC 111 cathodes. Li-Cycle, Redwood Materials, SungEel HiTech, Brunp, and the battery lines at Umicore and Glencore's Horne smelter all built capacity around NMC economics.

That bet is aging poorly, and the chemistry shift is already visible in the incoming waste stream.

NMC 811 cuts cobalt to a third of NMC 111 levels, and LFP, with no cobalt and no nickel, is gaining share through Tesla's standard-range models, BYD's full lineup, and the majority of Chinese EV production. Lithium carbonate went from over $80,000 per ton in late 2022 to roughly $10,000 by late 2023. When LFP packs reach recyclers in volume in the early 2030s, the feed will be lithium and iron phosphate, and the lithium revenue may not cover the hydrometallurgical processing cost. The three processing routes, pyrometallurgy losing lithium to slag, hydrometallurgy recovering it at chemical cost, and direct recycling recovering cathode crystals through relithiation with the constraint of needing sorted single-chemistry input, will coexist across feed tiers, and the overall revenue per ton processed is heading in a direction that should concern anyone who financed a recycling plant on 2021-era cobalt projections.

Connected Energy in the UK and B2U Storage Solutions in California are building on this second-life model, which means a recycling plant sized for 2035 retirement volumes may find half its expected feed diverted to grid storage and not arriving until the 2040s.

Palladium from autocatalysts, refined by Johnson Matthey, BASF, Heraeus, and Tanaka Kikinzoku, covers roughly 25 to 30 percent of annual global palladium supply. Mature collection through the dismantling chain, no subsidy needed, the whole system running on commodity value.

The theft wave from 2019 onward was a price signal about the formal system's speed. Rhodium passed $25,000 an ounce, a rhodium-bearing converter hit $1,000 scrap value, and a reciprocating saw took ninety seconds under a parked SUV and paid cash at a scrap yard the same day, while the formal pathway from insurance write-off to auction to licensed dismantler takes weeks or months and the vehicle's last owner sees none of the PGM value. That the fastest collection mechanism for a $1,000 recoverable component was theft is an indictment of how the formal chain handles end-of-life vehicles, and deposit-refund schemes for converters have been proposed in US state legislatures without going anywhere.

The EU Critical Raw Materials Act and the US Inflation Reduction Act both fold urban mining into supply security language, and at 10 to 15 percent of annual consumption for most critical metals it provides disruption insurance, not self-sufficiency, and lithium demand growth to 2035 makes the math especially unfavorable for any claim that battery recycling closes the supply gap. Landfill mining has been piloted at Belgium's REMO site and elsewhere and does not cover costs for typical municipal deposits. Gold tailings reprocessing in South Africa's Witwatersrand is the exception, commercially practiced for decades on waste from 19th-century gravity operations. The EU Battery Regulation mandates digital product passports, and whether data survives 20-year product lifetimes and platform migrations remains to be seen.