Investing in Platinum and Palladium

On the morning of Monday, March 7, 2022, eleven days after Russian troops crossed into Ukraine, the palladium price touched $3,440.76 an ounce, the highest it has ever been. The logic was simple and, for a few weeks, unanswerable: Nornickel, the Russian mining group, produces about 40% of the world’s palladium, sanctions were arriving by the day, and the metal has one dominant use, cleaning the exhaust of petrol engines, for which the car industry had no fast substitute. Traders who had watched palladium rise from about $500 in January 2016 to $2,800 before the pandemic had six years of evidence that the price only went one way.

It went to $876.60 instead, on our tape’s close of April 7, 2025, a fall of nearly three-quarters over three years. The sanctions came and Russian metal kept flowing, routed through third countries; the carmakers, who had spent a decade swapping platinum out of petrol catalysts because palladium was cheaper, spent the next three swapping it back in; and the electric-vehicle forecasts of 2022 implied that the metal’s only large market was on a path to zero. Nothing in the geology changed. What changed was the demand curve, and palladium has no second buyer to catch it when the first one steps away.

That is the shape of every platinum-group metal, and this guide is about living with it: the car fleet that sets demand, the single complex and single company that set supply, the price histories, the sister metals whose 2021 spike is the same lesson in miniature, and the premiums, fees and taxes that make these the most expensive precious metals to own. The hub’s flagship, Investing in Precious Metals, surveys the four metals together and carries the full dealer, storage and US tax map; Investing in Gold and Investing in Silver cover the monetary metals. This guide goes deeper on the two industrial ones and assumes you have not read the others.

A long-term holder of platinum has lost money. A long-term holder of palladium has made or lost a fortune depending on whether they bought before or after the car industry changed its mind. Here is what each has actually made, before the case for the metals is argued.

Platinum’s previous record was set in March 2008: the London PM fix reached $2,276 on March 4, 2008 and the NYMEX (New York Mercantile Exchange) future traded at $2,308.80 intraday, on a power crisis at Eskom, South Africa’s state utility, that shut the mines for five days in late January (Auronum and First National Bullion, on the 2008 record; USGS on the January closure; feeds quote the peak variously between $2,273 and $2,300). Eight months later it was near $750, a 67% fall, as the financial crisis emptied the car showrooms.

On September 8, 2026 the front-month future (the nearest-dated contract) closed at $1,838 on our tape, 19% below the 2008 peak in nominal dollars after eighteen and a half years. Over the same period the US consumer price index rose from 213.5 in March 2008 to 333.9 in July 2026 on the Bureau of Labor Statistics series, so the 2008 buyer has lost roughly 48% of purchasing power, before premiums, storage and tax. Gold, which touched $1,011 in the same month of 2008, closed at $4,443.90 on the same day of 2026, up 340%. That pair of numbers is the whole platinum discount in one sentence: an ounce of platinum bought 2.25 ounces of gold in March 2008 and bought 0.41 ounces on September 8, 2026.

Palladium’s record is shorter and stranger. From about $500 in January 2016 the metal rose to roughly $2,800 in February 2020 and to its record on March 7, 2022, a five-and-a-half-fold gain in six years that beat every other metal and most equities. A buyer at the record has lost 59% at the September 8, 2026 close of $1,399; a buyer at the 2016 low has made 180% in ten and a half years, about 10% a year, in the neighbourhood of the S&P 500’s long-run nominal return (11.2% a year since 1971 on officialdata.org’s series, via our silver desk) and with several times its volatility. The metal’s whole return sits inside a single demand shift, tightening emissions rules from 2016 to 2020, and the whole drawdown inside its reversal.

Two things belong beside those figures. The spot returns in fund literature omit the premium on the way in, the spread on the way out, the ETF’s 0.60% a year and the 28% federal rate; Section 13 puts them back. And the volatility is not a footnote: on our tape platinum’s worst session was a 19.0% fall on January 30, 2026 and palladium’s a 16.9% fall on December 29, 2025, against 11.4% for gold’s worst day in the same window. These metals move like small-cap stocks, and the rest of this guide explains why.

Gold’s demand is jewellery, central banks and investors, three buyers who want the metal for what it is. Platinum and palladium have one buyer who wants them for what they do, and that buyer is the car industry. That is what sets the price, and every other section answers the same question in more detail.

The six platinum-group metals, platinum, palladium, rhodium, ruthenium, iridium and osmium, occur together in the same ores and share a chemistry that makes them the best catalysts known: they speed reactions without being consumed. A catalytic converter, the honeycomb in a car’s exhaust, carries a few grams of platinum, palladium and rhodium that turn carbon monoxide, unburnt fuel and nitrogen oxides into carbon dioxide, water and nitrogen. Petrol engines have used mainly palladium since the late 1990s; diesel engines, whose cooler exhaust and excess oxygen suit it better, use mainly platinum. On the Precious Metals desk’s reading of the balances published by Johnson Matthey, the refiner, and the World Platinum Investment Council (WPIC), the producer-funded industry body, autocatalysts take roughly 40% of platinum demand and more than 80% of palladium demand; the rest of platinum goes to jewellery, mainly in China and Japan, to glass-making, petroleum refining and other industrial catalysis, and, in a good year, to investors.

Two consequences follow. Price discovery is a negotiation between miners with high fixed costs and carmakers with long contracts, priced off the London benchmark, and the investor sits at the edge of it: the exchange price you see is set by the marginal ounce, and the marginal ounce is small. And the demand curve is inelastic in the short run and ruthless in the long run. A carmaker pays whatever a converter costs in a given year, because a car without one cannot be sold, and spends the next five years engineering the expensive metal out. Palladium’s rise to $3,440 and its fall to $876 are the two halves of that sentence.

What the metals lack matters as much. There is no official-sector bid; central banks hold gold. Platinum jewellery was about a quarter of demand in 2025 (2,226 koz of 8,297 koz on the WPIC’s count), and the WPIC’s May 2026 forecast has it shrinking 12% in 2026 because the price rose. The investment markets are thin: abrdn’s Physical Platinum Shares (PPLT), the largest platinum fund, held about $2.3 billion on September 4, 2026 (Seeking Alpha, citing the fund) and its palladium sibling PALL about $769 million (Yahoo Finance, September 2026), against roughly $140 billion in SPDR Gold Shares alone. When investors arrive in a market that size, as in 2025, they move the price by tens of percent; when they leave, as in the first half of 2026, the WPIC’s annual balance flips sign. Section 8 shows how.

About 70% of the world’s mined platinum comes from South Africa, from a handful of deep, ageing mines in the Bushveld Igneous Complex north-west of Johannesburg, where the Merensky and UG2 reefs are worked more than a kilometre underground by Valterra Platinum (the former Anglo American Platinum), Impala Platinum, Sibanye-Stillwater and Northam. Supply is the most concentrated of any commodity you can buy, and that concentration explains both the deficits and the spikes. South Africa, Russia and Zimbabwe together supply roughly 73%, 10% and 8% of platinum on the WPIC’s breakdown (as summarised by Discovery Alert, 2026), which leaves about 9% for the rest of the world, mostly North America. Palladium is the mirror image: Russia’s Nornickel, mining the nickel-copper ores of the Norilsk region in the Arctic, produces about 40% of world palladium, and South Africa another 35% or so (Nornickel annual reporting and Reuters, via the Precious Metals desk).

South African supply has been falling for twenty years. Primary platinum output was about 5.3 million ounces in 2006 and about 3.9 million in 2025, a 26% decline (Discovery Alert, 2026, on the WPIC series). The reefs are narrow, the mines deep and hot, labour organised and expensive, and the metal is sold as a basket of platinum, palladium, rhodium, gold and nickel, so a mine’s economics depend on the basket price rather than platinum alone.

On top sat the electricity. Eskom’s rolling blackouts, load-shedding, interrupted smelters and ventilation intermittently from 2008 and almost continuously from 2018, and were at their worst in 2022–2023; the platinum spike of 2008 was itself triggered by an Eskom crisis that shut the mines for five days that January. The relief is dated: on May 16, 2026 Eskom marked a full year without load-shedding, last achieved in September 2018. That is good for costs and bad for the supply-shock thesis at once, which is the usual pattern with these metals.

Russia is different. Palladium there is a by-product of nickel and copper, so Nornickel keeps producing whatever the palladium price does, and the sanctions that followed February 2022 did not stop the metal reaching the market; they redirected it through Asian and Middle Eastern hubs at a discount that accrued to intermediaries.

The metal itself has never been sanctioned by the United States, as of September 2026, and US imports of Russian palladium rose to 27.6 tonnes in 2024 from 20.4 tonnes in 2022. The live threat is a tariff rather than a ban: in April 2026 the US Commerce Department set an anti-dumping duty of 132.83% and a countervailing duty of 109.1% on Russian palladium, to take effect only if the International Trade Commission finds injury to US producers (Reuters via Mining.com and Kitco, May 21, 2026). Meanwhile Nornickel guided its 2026 palladium output to 2.415–2.465 million ounces (Moz) from 2.725 Moz in 2025, which would be its lowest in twenty years (Reuters, April 2026). A Western investor cannot own Nornickel and should not assume the Western majors’ shares are insulated from its output.

Recycling is the third source and the one that responds to price: spent autocatalysts carry a few grams each, and the WPIC forecast platinum recycling up 9% for 2026 in its May 2026 Quarterly. Because the recycled ounce was once an autocatalyst, volumes track the age and scrappage of the fleet as much as the price; a decade of fewer diesels in Europe becomes a decade of less recycled platinum from 2030.

Diesel, then petrol, then the hybrid, then the battery: the market learned the demand side in that order, and each change in the fleet re-priced one metal against the other. The next one will too.

Until 2015 the story was diesel. Diesel cars took roughly half of new-car sales in Western Europe, each carried a platinum-heavy converter, and platinum traded above gold in most years before 2011; the two metals reached parity in 2011–2012, and gold has been the dearer of the two every day since mid-January 2015 (BullionVault, 2017). Then on September 18, 2015 the US Environmental Protection Agency issued its notice of violation to Volkswagen for the defeat devices in its diesel engines, and the European diesel share fell for a decade, to 8.9% of EU new-car registrations in 2025 and 7.5% in the first half of 2026 (ACEA). Less diesel meant less platinum per car sold, and the discount to gold opened; by 2024 an ounce of platinum bought about 0.4 ounces of gold on the Precious Metals desk’s record, a ratio without precedent in the twentieth century.

The petrol engine’s metal is palladium, and from 2016 tighter emissions rules in China, Europe and the United States raised the loading per car just as diesel’s share fell. That is the demand shift behind palladium’s rise from $500 to $2,800 between January 2016 and February 2020, and Johnson Matthey’s balances show it as a persistent deficit from 2012 to 2025 (JM PGM Market Report, May 2026, via Mining Weekly, May 14, 2026). Substitution is the counter-mechanism: a petrol catalyst can be reformulated to use platinum in place of some palladium, and once palladium cost twice as much the catalyst makers did exactly that. On our tape palladium still cost 1.18 times platinum on October 28, 2024 and 0.76 times on September 8, 2026. Substitution takes years, because a formulation is certified with the engine, and in practice it is one-directional.

The hybrid is the complication the electric-vehicle bears missed. A battery-electric car uses no platinum-group metal, and the 2020–2022 forecasts of a fast transition were the biggest reason palladium lost three-quarters of its value. A hybrid carries a full combustion engine and a catalyst that must work from cold more often, so it uses more PGM per car than a conventional one, 10–15% more platinum on the WPIC’s estimate, and hybrids took share as the EV forecasts were cut in 2024 and 2025. The net effect is a slow decline rather than a cliff: the WPIC’s May 2026 Quarterly has automotive platinum demand down 2% for 2026, and Johnson Matthey has palladium’s automotive demand contracting with petrol-car production and total palladium demand down 9% (JM, May 2026). Hold those against the marketing on both sides: the car market for these metals is shrinking by a few percent a year with a very long tail.

About a third of platinum demand does not depend on a car, and the bull case leans on a further slice that the balances do not yet show. The demand that exists and the demand that is promised need separating, because the promise is doing a lot of the work.

The genuine growth is industrial. The WPIC’s Q1 2026 Platinum Quarterly (May 19, 2026) forecast industrial demand up 9% to 2,238 koz in 2026, mainly on glass, where platinum-rhodium alloys line the vessels used to make LCD glass and glass fibre and Asian capacity expansions pull metal in for years at a time. Refining, nitric acid, silicones and medical devices are the steady users, and they are why a third of platinum demand does not depend on a car; Johnson Matthey expects platinum to stay in deficit in 2026 on exactly that (JM, May 2026).

Hydrogen is the promise. A proton-exchange-membrane fuel cell uses platinum as its catalyst; a PEM electrolyser uses platinum on one electrode and iridium on the other. If the hydrogen economy of the roadmaps arrives, it is a new platinum demand of the same order as the car industry and an iridium demand the market cannot supply, since iridium is a by-product produced in the hundreds of thousands of ounces.

The WPIC’s own hydrogen page (2024) projects almost 900 koz a year of platinum demand from electrolysers and fuel cells by 2030, about 229 koz of it from PEM electrolysers on a 38% PEM share. That would be roughly an eighth of 2025’s total demand; it is an industry body’s projection, and the WPIC has cut earlier editions of it. The published balances are more useful: hydrogen appears in 2025 and 2026 as a small line inside industrial demand, and JM’s 2026 ruthenium and iridium deficits are attributed to data storage and the energy transition together (JM, May 2026). Electrolyser demand for iridium is already real at the margin of a tiny market; fuel-cell demand for platinum is not yet visible in a large one. Hydrogen is an option embedded in platinum, priced near zero in the 2020s: you do not pay for it and should not count on it.

Palladium’s cycle is the cleanest example in any commodity of a demand shock met by substitution, and the same mechanism began running in platinum’s favour in 2025. Here is the whole of it, with dates.

The metal was about $500 an ounce in January 2016, cheaper than platinum and largely ignored. Emissions rules did the rest. By February 2020 it was near $2,800, and eleven days after the invasion of Ukraine it printed its record of $3,440.76 on March 7, 2022 as the market priced the loss of Nornickel’s 40% (APMEX price history; CNBC, March 7, 2022, via the Precious Metals desk; other feeds print the day’s high at $3,429.50). The record lasted a morning. The three forces the cold open described, re-routed Russian metal, substitution and the EV forecasts, then ran for three years: palladium fell below $1,000 in late 2023 and below $900 in 2024, and our tape puts the low at $876.60 on April 7, 2025, 74.5% below the record.

Then the rebound, which most readers missed. From that April 2025 low the future rose 147.5% to a close of $2,169.90 on January 26, 2026, the same day platinum, silver and gold’s peaks clustered, as the EV forecasts were themselves cut, hybrids took share and Western investors who had bought the platinum story added palladium as the cheap cousin. It then fell 46.5% in 149 days to $1,161.30 on June 24, 2026 and closed at $1,399 on September 8, 2026, 35.5% below the January high and 22.8% above a year earlier. Calendar 2025 was an 80.0% gain on our tape; 2026 to September 8 a 16.1% loss.

The balances explain the shape. Johnson Matthey’s May 2026 report has palladium in a 416 koz deficit in 2025, nearly double 2024’s 218 koz, which is the fundamental story behind the rally, and then a 214 koz surplus in 2026 as ETF holders sold in the first quarter and petrol-car production fell (JM, May 2026). Nornickel’s own 2026 outlook, via TASS, is a 100 koz deficit excluding investment and 200 koz including it, and its forecasts have swung between surplus and deficit within a year (via the Precious Metals desk). A market whose balance flips sign on one quarter’s ETF flow is one where the fundamentals and the price are different subjects, and palladium is that market in its purest form.

Platinum rose from $894 to a $2,852.40 close in thirteen months and lost 45.7% in the next five, on our tape. That run is the market event the reader most wants explained, and the mechanism, not the number, is what will still be true when the price is somewhere else.

Our tape records the front-month NYMEX platinum future at the close, stored daily since September 2021 with a 500-session chart window from September 12, 2024. It opened that window at $981.90, spent the autumn drifting to a low of $894 on December 31, 2024, and then did something it had not done since 2008. It closed 2025 at $2,034.50, a 123.8% calendar-year gain from the January 2 close, and rose a further 40% in January 2026 to a peak close of $2,852.40 on January 26, above the 2008 record by any measure (the WPIC and spot feeds put the intraday record at about $2,923 on the same day, January 26, 2026; the future’s close is ours).

It then fell 19.0% in the single session of January 30, 2026, the day gold fell 11.4% and silver 31.3%, and kept falling to a trough of $1,550.20 on June 30, 45.7% below the peak in 155 days. On September 8, 2026 it closed at $1,838: 35.6% below the peak, 18.6% above the trough, 87.2% above the start of the window and 33.1% above a year earlier.

Four things made the run, each dated. Tariffs first: between December 2024 and March 2025, fear that the new US administration would tax imported platinum sent metal from London and Zurich into NYMEX warehouses in New York, tightening the London market that prices everything else; when the exemption became apparent, NYMEX stocks fell 317 koz through the second quarter, then rose 201 koz to 531 koz in July 2025 as the trade reversed (WPIC Platinum Quarterly Q2 2025, September 10, 2025). Second, the lease market: three-month platinum lease rates, the interest a borrower of physical metal pays to the lender of it, averaged 6–16% through Q2 2025 against 1–3% in 2024, peaked at 22.7% in June 2025, and the one-month rate passed 35% in July (WPIC; International Precious Metals Institute). A lease rate that high means London has run out of lendable ounces, and it is the best confirmation that a deficit is physical rather than statistical.

Third, China: priced out of gold, Chinese buyers moved down a metal, imports rose 26% year on year in Q2 2025, global platinum jewellery demand rose 32% in the quarter, and the WPIC forecast Chinese jewellery fabrication up 42% for the year (WPIC Q2 2025). Fourth, the investor, who arrived last, bought the ETFs and futures through the autumn and drove the final 40% in January.

What ended it was what ended silver’s run the same week: leverage. The January 30, 2026 liquidation, triggered by a Fed chair nomination and a stronger dollar and amplified by exchange margin increases, hit the metals with the newest speculative length hardest. The ratio to gold tells it cleanly: an ounce of platinum was worth 0.385 ounces of gold on September 12, 2024, 0.280 at the low on April 22, 2025, 0.562 at the peak on January 26, 2026, and 0.414 on September 8, 2026. It has given back two-thirds of its recovery against gold while keeping most of its gain in dollars, because gold rose too.

“Deficit” was the word that carried the platinum story. Here are the numbers behind it, and the reason the WPIC’s own forecast flipped from deficit to surplus within six months of 2026 without anything happening in a mine.

The WPIC’s quarterly balance, researched by Metals Focus and funded by the South African producers, is the series behind every deficit headline. On its count 2025 was the third consecutive deficit and the largest in its series: 1,082 koz, after 921 koz in 2024, the second-largest (WPIC Platinum Quarterly, as reported by Discovery Alert and Kitco, March 2026). Against 2025 demand of 8,297 koz on the WPIC’s count that is a shortfall of 13% in a single year, met from above-ground stocks: the metal held in vaults, exchange warehouses and ETFs that is not in a catalyst or a ring.

The Q1 2026 Quarterly (May 19, 2026) estimated those stocks at about four months of demand; the Q2 2026 edition, released September 9, 2026, expects them to end 2026 at 1,747 koz, just under three months of cover (WPIC Q2 2026 release, September 9, 2026). For a metal whose supply is 70% one country, three months of stock is thin by any commodity’s standard.

Now the forecast. On March 4, 2026 the WPIC expected a fourth consecutive deficit in 2026 of 240 koz (Kitco; Newswire). On May 19, 2026 it raised that to 297 koz, with total demand down 9% to 7,674 koz because 2025’s exceptional investment and exchange-stock inflows were not expected to repeat (WPIC Q1 2026). On September 9, 2026 it forecast a 265 koz surplus, having cut its full-year investment demand forecast by 601 koz, and said in its own release that the shift was “overwhelmingly due to investment outflows” in the first half of the year (WPIC Q2 2026). Mine supply barely moved between those three documents. What moved was the investor: the same ETF and exchange-warehouse buying that the WPIC counts as demand when it comes in is counted as supply when it goes out, so a market that is in deficit on fabrication alone can print a surplus in a year when the funds sell.

That is not a criticism of the WPIC’s arithmetic; it is the arithmetic. But it means a platinum “deficit” headline describes the investor’s mood as much as the miner’s output, and a deficit forecast is not a floor. Johnson Matthey, whose May balance treats investment differently, put 2025’s platinum deficit at 951 koz on its own count, up from 559 koz in 2024, and still had platinum in deficit for 2026 on industrial demand and constrained supply (JM, May 2026), and we trust the JM fabrication balance more than the headline for that reason. The readings that matter are stock cover in months, which only falls when industry uses more than mines and scrap produce, and the lease rate, which only spikes when London’s ounces have run out. Both told the truth in 2025. Neither stopped the price halving in 2026, because the marginal ounce is an investor’s.

Rhodium rose tenfold in twenty-six months to March 2021 and then lost more than 80%. The four sister metals’ spike is the platinum story with the volume turned up, and a dealer will offer them to you at some point; this is what the offer is.

Rhodium, iridium, ruthenium and osmium are by-products of platinum and palladium mining, produced in quantities of hundreds of thousands of ounces a year rather than millions, with rhodium’s supply about a million ounces and most of it South African. They trade dealer-to-dealer, between refiners, catalyst makers and a few specialist houses; there is no futures contract, no meaningful ETF and no sovereign coin, and the reference prices are the daily quotes published by Johnson Matthey, Umicore and Heraeus rather than an exchange.

Rhodium is the one that matters. It is the metal that removes nitrogen oxides from petrol exhaust, there is no substitute for it at the loadings the rules require, and its market is a small fraction of platinum’s. When China’s stricter emissions rules met a South African supply interruption in 2020 and 2021, the Johnson Matthey base price went from under $3,000 an ounce in January 2019 to $17,000 at the end of 2020, above $21,000 in January 2021 and to a record $29,800 on March 23, 2021, a tenfold rise in twenty-six months (Johnson Matthey base prices, as compiled by Capital.com and FindBullionPrices).

It then fell more than 80% over two years as carmakers thrifted and the supply returned. Umicore’s published price was $9,600 on September 8, 2026, about a third of the peak and more than three times the 2019 level (Umicore Precious Metals Management price pages). Johnson Matthey’s balance had rhodium in a 50 koz deficit in 2025, against 9 koz in 2024, and forecast a 15 koz surplus for 2026 (JM, May 2026).

Iridium spiked the same spring for a different reason. It is the anode catalyst in PEM electrolysers, the tip of a spark plug and the crucible in which crystals for phones and 5G components are grown; its base price went from about $1,670 in December 2020 to a record $6,000 on March 19, 2021 on the same South African interruption and the hydrogen story (S&P Global Commodity Insights, March 19, 2021). Unlike rhodium it never gave most of it back: Umicore quoted $7,900 on September 4, 2026, and JM has it in deficit for 2026.

Ruthenium, for decades the cheapest of the group, spiked above $800 in 2007 and 2021 (Daily Metal Price) and has become the interesting one: it is the metal in high-capacity hard-drive platters, the data-centre build-out has tightened a market of a few hundred thousand ounces a year, and Umicore’s price was $1,700 on August 25, 2026, after $1,750 on July 22. Osmium has no liquid market and we would not call it an investment.

The way they trade is the reason to leave them alone. A one-ounce rhodium or iridium bar carries a premium of roughly 10–20% over the published price and is bought back at a similar discount (a general dealer range the Precious Metals desk could not pin to one quote): a round trip of a fifth to a third of the position, on a sawtooth price series, at 28% tax. For nearly everyone the correct allocation is zero. If you want the exposure, it is inside the South African miners’ basket, where a rhodium spike arrives as a dividend rather than a bar you cannot sell.

Platinum’s price is made in London, New York and, since November 2025, Guangzhou as well as Shanghai, and knowing where it is made tells you which price you are actually getting when you buy a coin, a fund or a future.

The physical benchmark is London. Platinum and palladium have their own market association, the London Platinum and Palladium Market, alongside the London Bullion Market Association (LBMA), and their own twice-daily benchmark prices, the LBMA Platinum Price and LBMA Palladium Price, set in auctions at 9:45 and 14:00 London time and used by refiners, carmakers and the ETFs to value their metal. Good-delivery bars are the standard: platinum and palladium plates and ingots of 1–6 kilograms at .9995 fineness from accredited refiners. Unlike gold, much of the industrial trade moves as sponge, the powder form that catalyst makers use, which is priced off the same benchmark at a small premium or discount and which a retail investor never sees. The lease market sits on top of London: refiners and fabricators borrow metal, and the lease rate described in Section 7 is the interest.

The paper price is New York. The NYMEX platinum future (50 troy ounces) and palladium future (100 troy ounces) on CME Group are where speculative positioning lives, where the January 30, 2026 liquidation happened and where the exchange warehouses that absorbed the 2025 tariff flow sit. At a $1,838 platinum price a contract is about $92,000 of metal, and at $1,399 a palladium contract about $140,000; there is no micro contract of the kind gold and silver have, so the futures are an institutional instrument. Open interest and volume are a fraction of gold’s, which is why a fund that would barely register in gold can move platinum by 5% in a session.

The third venue is Asia. China is the largest platinum jewellery market and, in 2025, the largest marginal buyer of bars; Shanghai’s exchanges price the metal in yuan, and the premium or discount to London is the cleanest read on Chinese demand. Japan has a long retail tradition of platinum bars. The Guangzhou Futures Exchange listed platinum and palladium futures on November 27, 2025 and options the next day, China’s first domestic PGM derivatives, with a first-day turnover of 42.28 billion yuan (Reuters; WPIC Platinum Perspectives); what matters for a Western holder is that Chinese demand has been the swing buyer of physical platinum since 2025 and shows in the Shanghai and Guangzhou prices before it shows in London.

Between the venues sit the refiners and fabricators, Johnson Matthey, Heraeus, Umicore, BASF and the miners’ own refineries, who turn concentrate into sponge and ingot and publish the daily prices for the minor metals. The market is small enough that its participants know each other and, in dollar terms roughly a third of silver’s at September 2026 prices (our arithmetic on the WPIC, Johnson Matthey and Silver Institute demand totals), small enough to be squeezed. For an outsider there is no retail-friendly price: every wrapper below pays a premium to London, a fee to a sponsor who holds London bars, or the size and leverage of a New York contract.

The wrapper decides more of your return in these metals than the direction does, so here are the ways in, ranked, with their published costs.

Coins and bars

The American Platinum Eagle, minted since 1997 at .9995 fineness, is the reference coin in the United States and the only PGM coin the Code names for IRA purposes. It is expensive. In a September 2026 capture of APMEX’s pricing a random-year one-ounce Platinum Eagle listed at $2,244.69 against a spot price of $1,855.60, a 21% premium; the Canadian Platinum Maple Leaf listed at $2,094.69, a 12.9% premium (APMEX product pages via search; the capture day is not shown and premiums move, so treat these as a September 2026 reading). APMEX’s own guide puts the Maple Leaf between generic bars and the Eagle, and that is the ordering: Eagle dearest, Maple in the middle, refiner bars cheapest. The Eagle’s premium reflects the US Mint’s small, irregular bullion mintages, which give it a semi-numismatic bid, and its IRA eligibility. The Precious Metals desk’s general range for PGM coin premiums is 5–10%; the September 2026 quotes are the dated evidence that the range is a floor.

Palladium coins are thinner still. The Canadian Palladium Maple Leaf was struck in only a few years and trades on the secondary market; dealer guides put government-minted palladium coins at 8–15% over spot (JM Bullion); and the US Mint priced its 2026-W uncirculated American Palladium Eagle, a collector issue in a series that began in 2017, at $1,895 on September 3, 2026 against a spot price near $1,400, about 35% over on our arithmetic. That last figure is what a collector strike costs and is not a bullion premium, which is the reason to avoid it.

One-ounce and larger platinum bars from PAMP, Valcambi, Argor-Heraeus or Johnson Matthey, in assay packaging, run a few percent over spot, and dealer buybacks on both coins and bars sit at or a little under spot. A round trip in platinum coins is therefore in the range of 15–25% at the September 2026 premiums, against 2–4% for a Gold Eagle. That single figure is why physical is the wrong wrapper for a platinum position you intend to trade.

The physical ETFs

abrdn’s Physical Platinum Shares (PPLT) and Physical Palladium Shares (PALL) are grantor trusts holding allocated London good-delivery metal at an expense ratio of 0.60% each; PPLT held about $2.3 billion on September 4, 2026 and PALL about $769 million (Seeking Alpha; Yahoo Finance, September 2026). GraniteShares’ Platinum Trust (PLTM), at about $141 million and a 0.50% expense ratio, is the small, slightly cheaper alternative (ETF Database; AAII, 2026).

Sprott’s Physical Platinum and Palladium Trust (SPPP) is a different structure: a closed-end fund that charges a 0.50% management fee plus expenses, a total expense ratio above 1% on third-party tallies, and trades at a premium or discount to its net asset value that has ranged from about +7% to −20% over its life (Sprott; Cbonds). We did not verify its discount on a given day, and a buyer should, because that range moves more in a year than the fee costs in a decade. The open-ended trusts cost 0.60% plus a spread and carry the gold ETFs’ tax surprise: the IRS looks through the share to the metal, and a long-term gain is taxed at 28%.

Futures

The NYMEX contracts give leverage, the tightest spread and the best tax treatment, Section 1256’s 60/40 blend, at the cost of size ($92,000 and $140,000 a contract at September 8, 2026 prices), a roll every quarter and contango, the premium of the next contract over the expiring one that a holder pays at each roll, which runs near the short-term interest rate. For a reader with a six-figure PGM view and a futures account they are the cheapest wrapper; for everyone else they are the wrong size.

The miners

This is where the 2025–2026 cycle paid. The South African majors sell a basket, so a rise in platinum, palladium and rhodium together lands on a cost base that barely moves. Sibanye-Stillwater’s first-half 2026 results (September 1, 2026) showed adjusted EBITDA up 111% to R31.8 billion and a swing to a R18.8 billion profit from a R3.9 billion loss a year earlier, with its South African PGM operations’ EBITDA up 302% to R19.2 billion on a 4E basket price (platinum, palladium, rhodium and gold, sold together) 67% higher; it declared a R5.7 billion dividend (Mining Weekly).

Valterra Platinum, demerged from Anglo American in 2025, reported headline earnings of R21.5 billion for the six months to June 2026 against R1.24 billion a year earlier, up 1,634%, and a R15.1 billion interim dividend at 70% of headline earnings, helped by the recovery of its Amandelbult mines (Daily Maverick, July 29, 2026; Business Day). Impala Platinum declared R17.1 billion of dividends for its year to June 2026 (results of September 3, 2026). Those three declarations, made between July 29 and September 3, 2026, sum to R37.9 billion; Business Day put the three majors’ 2026 dividends at about R44.5 billion on a wider count (September 4, 2026).

Those are numbers from the top of a cycle, and the reason to own miners only with the cycle in mind. A basket price 67% higher took Sibanye from a loss to R18.8 billion; the same arithmetic ran in reverse in 2023 and 2024, when the palladium collapse pushed the sector to losses, closures and the restructuring of Sibanye’s US Stillwater mines (about 700 layoffs announced in September 2024 after more than $350 million of Montana losses since the start of 2023, and a R7.6 billion impairment in the first half of 2024).

A miner is a leveraged, rand-denominated, single-country bet on the basket with equity tax at 20%: the right wrapper for a view on the cycle and the wrong one for a store of value. All three have American depositary receipts, so the position is a brokerage click, which is more than can be said for any other route into rhodium. Investing in Gold Miners and Royalty Companies covers the equity route in full; the PGM majors are the same trade with a smaller basket and one country.

The US treatment is gold’s in outline and different in three details that matter for these metals. The flagship’s tax chapter and Investing in Gold carry the full map, including state sales tax and the foreign-account report (FBAR); what follows is what a PGM holder needs.

Platinum and palladium bullion and coins are collectibles under IRC §1(h)(4)–(5), so the maximum federal rate on a gain held more than a year is 28% rather than the 20% that applies to shares, with the 3.8% net investment income tax on top for higher earners. Short-term gains are ordinary income. Because PPLT, PALL and PLTM are grantor trusts, the look-through applies to them too, and each year’s small sales of metal to pay the sponsor’s fee generate a reportable gain or loss that the trust reports on its annual tax schedule. Futures are Section 1256 contracts, 60% long-term and 40% short-term whatever the holding period, marked to market at year-end, a blended top rate of about 26.8%. Mining shares and ADRs are ordinary equities at 20% with qualified dividends, subject to South African withholding on the dividend.

The IRA fineness test

The test under IRC §408(m) is .9995 for platinum and palladium, stricter than gold’s .995, and the American Platinum Eagle is admitted by name. A self-directed IRA can hold Eagles, Maple Leafs and .9995 bars through a trustee at an approved depository and, under IRS private letter ruling 200732026, the physical trusts.

Dealer reporting thresholds

The thresholds are different from gold’s. A dealer files Form 1099-B when a customer sells it platinum bars totalling 25 troy ounces or more or palladium bars totalling 100 troy ounces or more, at .9995 fineness, because those are the quantities that satisfy a CFTC-approved futures contract, which is the test the IRS instructions use (Instructions for Form 1099-B, 2025–2026, with the IRS’s own correction notice; dealer lists at Scottsdale Mint and CoinWeek, 2026). Platinum and palladium Eagles and Maple Leafs are not on the reportable list. As with gold, this is dealer reporting, not an exemption; the gain is taxable regardless.

The premium is basis

The premium is part of your basis: a coin bought at 21% over spot and sold at spot has a smaller taxable gain than the metal’s move implies, which is cold comfort but should be on the invoice you keep.

One platinum position, ten ounces bought at our September 8, 2026 close and held for ten years, run through four wrappers with every cost and the tax shown, because the ranking in Section 11 only becomes real in dollars. The assumptions are stated so you can change them: platinum at $1,838 compounds at 5% a year nominal to $2,994 in September 2036; the top federal rates apply (28% collectibles, 20% equity, 26.8% blended for futures) with no state tax and no NIIT; the coin prices follow the September 2026 APMEX capture, the Maple Leaf at its 12.9% premium over $1,838, or $2,075 a coin, and the Eagle at its $2,244.69 list rounded to $2,245, which is 22% over $1,838; dealer buyback is 98% of spot; a safe-deposit box costs $100 a year; PPLT charges 0.60%. The metal itself gains 62.9%.

Ten Platinum Maple Leafs

At $2,075 a coin the outlay is $20,750. In 2036 ten ounces at $2,994 are worth $29,940 at spot and a dealer pays 98%, $29,341. The taxable gain is $29,341 less the $20,750 basis (the premium is in the basis), $8,591, and the tax at 28% is $2,405. Ten years of the box cost $1,000. Net cash: $25,936, a 25.0% gain on the outlay, 2.26% a year, against the metal’s 62.9%. The 12.9% premium and the 2% buyback discount cost about 15% of the position on the way in and out, and the tax took another $2,405.

Ten Platinum Eagles

The same ten ounces in Eagles at $2,245 cost $22,450. The sale is the same $29,341, the gain $6,891, the tax $1,929, the box $1,000. Net $26,412 on $22,450: a 17.6% gain, 1.64% a year. The Eagle’s extra premium is pure cost unless a future buyer pays it back, which the semi-numismatic bid sometimes does and which you cannot plan on.

PPLT in a taxable account

$20,750 buys the equivalent of 11.289 ounces at spot, with no premium and a one-cent spread. The 0.60% fee comes out in metal, so by 2036 the shares represent 11.289 × 0.994 to the tenth power, 10.630 ounces, worth $31,826. The gain is $11,076, the tax at 28% is $3,101, and the net is $28,725: a 38.4% gain, 3.30% a year. Nothing to store, nothing to insure, and a 1099 that shows the annual expense-sale gains you would otherwise forget.

PPLT in a Roth IRA

The same $31,826 with no tax: a 53.4% gain, 4.37% a year, against the metal’s 5.00%. The only leakage is the fee, and the collectibles rate never applies. This is the wrapper for a platinum holding that is meant to sit for a decade.

If the price is flat

At $1,838 in 2036 the Maple Leafs return $17,012 after the buyback discount and the box, a loss of 18.0%; the Eagles lose 24.2%; PPLT returns $19,538, a loss of 5.8%, all of it the fee. A flat decade in platinum is not hypothetical: the 2008 buyer had one, and then another.

The order is gold’s and the gaps are wider, because platinum’s coin premiums are five to ten times a Gold Eagle’s. A reader who wants metal in hand should buy the cheapest recognised bar, not the Eagle, and should not expect it to beat the ETF.

A PGM holder can lose everything rather than a percentage in five ways, and the list differs from gold’s in one respect: the market itself can do it.

The market

Palladium fell 74.5% from March 7, 2022 to April 7, 2025 with no fraud, no default and no change in the geology, because its one buyer changed its recipe. Platinum lost 67% in eight months in 2008 and 45.7% in five months in 2026. Rhodium lost more than 80% in two years from 2021. A metal with one industrial buyer and no monetary bid can halve on a forecast revision, and a position sized like a gold position will be a mistake you feel for a decade. Nothing else in this section is as likely as this.

The dealer

In September 2020 the CFTC and thirty state regulators charged TMTE Inc., trading as Metals.com, Chase Metals and Barrick Capital, with a fraud that had taken more than $185 million from at least 1,600 mostly elderly customers since September 2017, over $140 million of it retirement savings, by selling bullion at prices 100% to more than 300% above the market (CFTC release 8254-20; NASAA). Safeguard Metals and its principal were charged in February 2022 over a similar $68 million scheme, settled by consent order in October 2023, and in October 2025 the court’s final judgment ordered $25.6 million of restitution and a $25.6 million penalty, more than $51 million in all, for at least 450 customers (CFTC releases 8489-22, 8812-23 and 9139-25). Both sold overpriced coins into self-directed IRAs by telephone, and the platinum and palladium products, with their thin markets and unfamiliar premiums, are where a 100% mark-up is easiest to hide. The test is the melt value: if a dealer will not quote the premium over spot as a number, hang up.

The pool

The flagship covers the unallocated trap: Northwest Territorial Mint in 2016 and Bullion Direct in 2015 both collapsed holding a fraction of the metal their storage customers believed was theirs. PGM storage programs are rarer and smaller, which makes the test more important, not less: a bar list with serial numbers, a third-party vault, insurance in your name and the right to take delivery, or you own a promise.

Authenticity and form

Platinum and palladium coins are hard to fake by density, but plated and underweight fakes exist, and a dealer’s X-ray fluorescence (XRF) gun or a Sigma-style conductivity tester is the check, as for gold. The form risk is subtler: industrial sponge is not a retail product, and any offer of “industrial-grade” metal at a discount is an offer of something you cannot resell. Buy only good-delivery-refiner bars and sovereign coins, and keep the assay packaging.

The single point of failure

Supply has one, and it cuts both ways. A Bushveld strike or a Nornickel accident is a price spike for the metal and a disaster for the miner you own; a return of Eskom’s blackouts is the same. If your PGM exposure is a South African miner, you own the country’s grid, its labour relations and its currency alongside the metal, and the R37.9 billion of dividends the three majors declared between July 29 and September 3, 2026 is what the good half of that looks like.

For a reader who has read the record and still wants the exposure, the sequence is short, because the first two steps remove most of the mistakes.

1. Decide which metal and why, in one sentence. “Platinum, because the ratio to gold is at the low end of its range and the lease rate says the deficit is real” is a thesis with an exit. “Palladium, because Russia” was a thesis in March 2022 and it lost 74.5%. If the sentence does not contain a number you will watch, do not buy.
2. Size it as a trade, not a store of value. One to two percent of financial assets, at most a fifth of whatever you hold in precious metals, and an amount you would be content to see halve.
3. Choose the wrapper by account. In an IRA or Roth: PPLT (or PALL, if step 1 said palladium), at 0.60% with no collectibles problem. In a taxable account: PPLT, accepting the 28% rate, or NYMEX futures if the position is above $100,000 and you can post several times the margin, for the 26.8% blended rate. For a view on the cycle rather than the metal: the miners’ ADRs, sized as an equity.
4. If you want coins, buy the cheapest recognised unit and write the premium down. A Platinum Maple Leaf or a refiner’s one-ounce bar at the lowest premium a dealer-comparison site shows that day, never a current-year Eagle at 21% over, never a “collector” issue, never a minor PGM bar. Keep the assay card and the invoice with the premium shown; it is your basis.
5. Store by value. Under about $5,000, a home safe; above that, a safe-deposit box or an allocated vault with a bar list, and confirm the vault will take platinum, since not all bullion programs do.
6. Enter on the ratio and the lease rate, not the headline. Buy platinum below 0.35 ounces of gold with a three-month lease rate above 5%; add while both hold; sell when the ratio crosses 0.50. Buy nothing the day after a 10% up move: on our tape four of the five best sessions in two years fell inside the five weeks before the top.

These are the readings that would change our view, with thresholds and the dates they arrive. Everything here is as of September 9, 2026 and is the part of the guide to refresh first.

• The platinum/gold ratio, daily on our tape. 0.414 on September 8, 2026. Below 0.35 platinum is cheap enough to buy against gold; above 0.50 it has run further than the fleet justifies and we would rotate back into gold. The range in the window was 0.280 (April 22, 2025) to 0.562 (January 26, 2026).
• The three-month lease rate, from the WPIC’s quarterly and the trade press. Above 5% the physical market is tight and a deficit is real; the 2025 peak was 22.7% in June. A rate back near the 2024 norm of 1–3% while the WPIC still reports a deficit means the deficit is an investment-flow artefact.
• WPIC stock cover, quarterly (next editions in November 2026 and March 2027). Forecast at just under three months for end-2026. A print below two months with industrial demand still growing would be the genuine squeeze; a print back above four months says the 2025 deficit was borrowed from investors and is being repaid.
• Johnson Matthey’s PGM Market Report, May 2027. Palladium was forecast in a 214 koz surplus for 2026; a second surplus year confirms the substitution flow and keeps us out of palladium. A platinum fabrication deficit above 500 koz on JM’s count, which excludes investment, would be the number that upgrades platinum from a trade to a holding.
• The car mix. A year in which hybrids’ share of global sales falls while battery-electric share rises is a year of lower PGM loadings; the reverse extends the tail. The IEA’s Global EV Outlook each spring is the reference.
• South Africa and Russia. Eskom’s run without load-shedding dates from May 2025; a return of stage-4 or worse blackouts is a supply shock for the metal and a cost shock for the miners, and a wage strike at any major is the same in miniature. The US anti-dumping and countervailing duties on Russian palladium set in April 2026 take effect only on an International Trade Commission injury finding; that finding, or any G7 sanction, matters for palladium first.
• Our tape. The June 30, 2026 trough of $1,550.20 is the line for platinum: a close below it says the unwind of the 2025 run is not finished. A close above the January 30, 2026 liquidation-day close of $2,102.80 would say the leveraged excess has been worked off. For palladium the equivalents are $1,161.30 (June 24, 2026) and $1,688.30 (January 30, 2026).
• The Platinum Eagle premium. Above 25% over spot, retail is chasing and a coin buyer should wait; below 10%, a buyer who insists on Eagles should do all their buying then. Check the comparison sites, not one dealer.