SA’s gold didn’t run out. The owners did – Duarte da Silva

Duarte da Silva’s Gold Papers came down to one line: we didn’t run out of gold, we ran out of conviction and courage. His new series, The Money Metal, follows the money: the deals, the mining houses, the central banks, the savers. This opening paper audits six gold companies on a single day. Two left. AngloGold sold Mponeng, the world’s deepest mine, for $200m in cash and moved to Denver. The mine has since paid Harmony $1.5bn. Four stayed: Harmony, Sibanye, DRDGold and Pan African. Guess which ones the market rates higher? It’s praise, with measured criticism for the boards, and for a state that can’t count its own gold..By Duarte da Silva*.South Africa declares ninety tonnes of gold a year and produces, on the evidence of other countries' customs books, a third to a half more. Four listed companies dig three-quarters of the declared ounces: Harmony, built over thirty years from mines the majors sold; Sibanye-Stillwater, given the mines Gold Fields did not want and now a platinum company with a gold division; DRDGold, the oldest name on the exchange, which gave up the reef for the dumps and has paid a dividend every year since; and Pan African, which refused to choose between the two. Between them Harmony and Sibanye hold every deep mine the majors were prepared to close, and the shafts, towns and pumps that go with them. Two others left. Gold Fields keeps one mine and a Sandton address. AngloGold Ashanti sold the deepest mine on earth for two hundred million dollars in cash and moved to Denver; the mine has since returned its buyer one and a half billion dollars. Here all six are audited on the same day: what each is, what it earns, what it sold or bought, what a decade of the market says about it. The findings are five, and they come first. .Company figures are from the latest audited or reviewed results of each company: Harmony, DRDGold and Pan African for the year to 30 June 2026, with Harmony's reserve statement and reporting suite of 23 October 2026, Sibanye-Stillwater, Gold Fields and AngloGold Ashanti for the six months to 30 June 2026. Market values are at the closes of Friday 30 October 2026, the day gold fell to a one-year low: gold about US$3,875 an ounce and the rand at 16.71 to the dollar, about R2.08 million a kilogram; rand figures from the companies' June accounts are converted at R16.48, the rate of 16 October on which the audit was first struck, unless the text says otherwise; the ten-year history is at each year-end. The shadow-ounce arithmetic, the leavers' ledger, the sum-of-parts, the per-ounce ratings, the ranks and the judgements are mine. .THE ARGUMENT 01 What this paper finds Six companies mine, or mined, South Africa's gold. Four stayed on the Witwatersrand: Harmony, Sibanye-Stillwater, DRDGold and Pan African. Two left: Gold Fields and AngloGold Ashanti. This paper audits all six on one day, 30 October 2026. It finds five things, and makes one count the market has not made. Each is stated here with its number and the section that carries the evidence; nothing in the body of the paper adds a verdict these four do not already contain. First, the leavers did not merely sell cheap; they left. AngloGold took US$200 million in cash for Mponeng in 2020, moved its address to Denver three years later and now has one employee in a hundred in South Africa; the mine has since paid Harmony US$1.5 billion of free cash, US$573 million of it last year, and its reserve life is twenty years, not the seven AngloGold declared at the sale. Gold Fields unbundled Kloof, Driefontein and Beatrix in 2013 at a value of US$1.07 billion and kept the Sandton address; the three mines have since paid Sibanye's shareholders US$2.7 billion. Both are children of Anglo American, whose founder wrote in 1954 that the group's profits were to be earned so as to make "a real and permanent contribution" to the people and the development of Southern Africa, and whose boards sold their Gold Fields stake in 2004 and their last AngloGold shares to John Paulson at US$32 in the month the markets bottomed in 2009. That verdict is on the ledger, not on opinion. Section XI carries the ledger and Figure 5 shows Mponeng year by year. .Second, the leavers are, today, the better businesses on the measures a buyer of ounces uses, and this paper does not dispute it. Their cost is lower, about US$1,950 an ounce all-in against the stayers' US$2,360 weighted by output; their margins are wider, US$2,650 to US$2,785 an ounce against US$1,490 to US$2,370; their reserves are larger, their balance sheets stronger, and over ten years they have made their holders fourteen and fifteen times their money. The market rates them for it, at US$12,000 and US$13,300 per annual ounce, and at 6.9 and 10.7 times earnings. Sections IX, X and XIII. Third, two of the four who stayed match the leavers on quality and are rated below them, and the other two are rated correctly. Pan African is the best company of the six on the seven measures ranked in Section XIV, and at US$1,867 an ounce the lowest-cost producer in the book; DRDGold is second. Both are 88 to 100 per cent South African, and both are rated within four per cent of Gold Fields per annual ounce and at 7.9 and 7.7 times earnings, a point above it. Harmony and Sibanye are the two highest-cost producers of the six with the two thinnest margins, and the discount on them, at US$6,400 and US$2,700 an annual ounce, is explained: a hedge book that cost R9.65 billion, a copper programme that takes the next four years of cash, and a gold division buried inside a platinum group. It is also the price of having stayed. The two carry between them every deep mine on the Witwatersrand that a major was prepared to close, Harmony's 47,000 employees and contractors and the 70,000 of Sibanye's South African region, gold and platinum together, against AngloGold's one employee in a hundred. Had they not bought, the mines would have gone the way of Blyvooruitzicht and Buffelsfontein: liquidated, unpumped and handed to the illegal trade. Sections V to VIII. Fourth, the discount is depth, group structure and size, not the flag, and every part of it is a board's to close. A South African surface ounce is rated with an Australian open-pit ounce; what the market discounts is four kilometres of rock, gold held under palladium, and companies too small for the index flows that tripled the leavers in 2025, whatever exchange they are listed on. Section XV sets out the three, and Section XVIII what each board should do about them. The discount is also a currency: a company rated at a fifth of its peer cannot buy anyone with its shares, which is why the four who stayed were built in cash, buying at the bottom what the sellers had written off, and why over the decade DRDGold and Harmony have beaten every one of the world's ten largest gold producers and Pan African all but one. Figure 9 sets them against the world. Fifth, the leavers took the balance sheets that explored the basin, and nobody has replaced them. AngloGold and Gold Fields spent about US$574 million on exploration in 2025, thirteen times the whole country's exploration budget of about US$44 million, and none of it in South Africa. The four who stayed drill their own fences: conversion at Harmony, Burnstone at Sibanye, Royal Sheba and Poplar at Pan African, nothing at DRDGold. The ground beyond, the richest on earth, has not been drilled in a generation, and the right to drill it sits in a queue the state has promised to clear five times since 2023. Section XVII. And the count. South Africa produces a third to a half more gold than it declares, on the evidence of London's refiners and Dubai's customs books, and the listed four pay to police the producer that files no accounts. Section IV. The five findings are restated with their numbers in Section XIX. Three propositions hold throughout. A reserve is a line drawn through an orebody at a price; the line moves when the price does, the gold does not. The industry's decline was decided in buildings, not discovered in rock. The pessimists were right about the mines and wrong about the gold. The leavers sold cheap, left, and run the better businesses today. Two of the four who stayed match them on quality and are rated below them; the other two carry the deep mines that would otherwise have closed, and are rated for it. The discount is depth, structure and size, not the flag, and a board can close it. The next ounces are nobody's job. THE FRAME 02 Ninety tonnes, and who digs them South Africa produced about a thousand tonnes of gold in 1970 and about ninety in 2025, eleventh in the world, and the Minerals Council expects it to stay near ninety next year. That is the whole industry: ninety tonnes, some 2.9 million ounces, worth about eleven billion dollars at the price of 30 October 2026. The four who stayed mine about 65 tonnes of it, Harmony alone 38.6, and South Deep another ten. The rest is a dozen smaller producers, among them Qala Shallows, which poured its first gold in March 2026, the first new underground gold mine the country had opened in fifteen years, and the illegal trade that leaks into the official count through the scrap market. Figure 1 has the split. By market value on 30 October 2026 the four who stayed are worth about 21 billion dollars, and 16 billion counting only their gold, because Sibanye's 6.1 billion is a group three-fifths platinum. The two who left are worth 70 billion, with one South African mine between them. That ratio has never made sense to me, and at this gold price it makes less. The market pays 70 billion dollars for the two companies that took their capital out of the Witwatersrand and 16 billion for the gold of the four that kept theirs in it, at a price which, even after the fall of 30 October to a one-year low, is higher than any of the six received in any year before 2025, and at which every one of them is making more money than at any time in its history. The values are in Table 14; Sibanye's gold-only value is worked in Section VI. .FIGURE 1 Gold mined in South Africa by company, tonnes, latest reported year: Harmony year to June 2026; Sibanye-Stillwater's own mines at the midpoint of 2026 guidance; South Deep 2025; Pan African year to June 2026 excluding Australia; DRDGold year to June 2026; the balance to a national ninety tonnes. Sources in the references. .FIGURE 2 Gold mined in South Africa, tonnes, latest reported year, against market value in US$bn at the closes of 30 October 2026; Sibanye's value shown as the group with its gold share in solid colour. Sources in the references; author's arithmetic. .Four companies, 21 billion dollars, 16 of it gold, three-quarters of the country's gold. Two companies, 70 billion dollars, one mine. EXHIBIT A · THE SIX ON THE RECORD 03 The six, before the audit This paper does not start from a blank page. The Gold Papers, the thirty-five papers that precede it, judged these six companies one by one, and the judgements are on the record. They are restated here, plainly, so that the audit can test them. The frame was a charge sheet against the industry, not against any one company. South Africa produced about a thousand tonnes of gold in 1970 and ninety in 2025, a fall of 91 per cent, while the price rose twenty-two times, from US$255 at the 1999 low to the US$5,589 record of January 2026, and the country slid from first in the world to eleventh. The gold did not run out. The Witwatersrand's remaining endowment, about 48,100 tonnes on the published estimate, outweighs every other gold province on earth combined. The basin has been declared dead four times, in the 1890s, the 1940s, 1999 and 2009, and each time the obituarist measured a flow and reported a stock. One sentence carried the whole argument. We did not run out of gold. We ran out of conviction and courage. On the six companies my verdicts were specific, and two things should be said about them before the audit. In places they were softer on the two that left than this paper will be. And they were harder on the two that carried the deep mines than the evidence below justifies. Each verdict is given as it stood: what it said, and what it missed. Harmony I called the buyer of last resort: the company that took the mines the majors discarded. Bernard Swanepoel built it from one Free State mine into the world's fifth-largest gold company and took the share from R20 to R170; beside Oppenheimer he is the model of the builder who stayed. Peter Steenkamp wrote two cheques to AngloGold, about R3.6 billion in 2018 and R4.4 billion in 2020, for mines now worth twenty times what the seller was glad to take. My score was hedged: a collar book that sells away the gold price its shareholders bought the company for, and a copper programme the operator had not yet justified, because the largest cheque in its modern history went to a mine eleven thousand kilometres away when its own dumps offered a second, re-rated metal at home. What that score gave too little weight to is what Harmony carried while the sellers left. Every one of its nine underground mines was bought from a seller that had written it down or was preparing to close it, and Welkom, Carletonville and Orkney kept their shafts because Harmony took them. Mponeng, where it mines the Ventersdorp Contact Reef deeper than Kloof ever reached, is now the best asset it owns. And its surface business, the largest surface gold producer in South Africa, runs under Mr Nel almost unremarked, and unpriced. Sibanye's verdict turned on one mine, Kloof. Kloof's declared life of one year is the product of a reserve-price convention, not of the rock: its resource fell from thirty-three million ounces to four because the assumptions moved, not the gold. I argued that Kloof was not an exception being managed but the template, and that the template showed a gold business funding everything except its own future. The word I used was run-off. Neal Froneman had his due. He took three mines that Gold Fields had stamped end of life in 2013, took the label off, turned R10 billion of market value into R139 billion and returned more than R46 billion to shareholders. My score was exposed: the inheritor, betting that Burnstone and the gold price would rescue it. The score stands as arithmetic. What it left out is that the three mines have been kept open, pumped and paying for thirteen years by the company that was handed them, and that the one new deep gold mine a listed company has approved in this country in a decade, Burnstone, at a 36.1 per cent return, is Sibanye's. I called Richard Stewart a swing vote: a chief executive who will not buy gold reserves at this price, whose own capital-allocation rules rank the development of existing reserves ahead of growth, and who entered Burnstone in that contest and never entered Kloof. One fact stood against the company's green credentials: Cooke's tailings, at 9,980 kilograms of carbon dioxide a troy ounce, are the most carbon-intensive gold in a survey of 329 mines. Dr Stewart has his credit regardless, for Burnstone, the first new gold mine the company has built. DRDGold I described as an environmental clean-up company that generates gold: the largest pure-play tailings retreatment company in the world, though not the largest retreater of tailings, because Harmony's surface business, with Mine Waste Solutions inside it, is bigger and sits unpriced inside a deep-level group. The model is a leveraged bet on the gold price laid over a fixed stock of waste: a magnificent way to harvest what has already been won, and not, on its own, a future for the basin. Three episodes from its record mattered. In 2014 it tried fine-grinding, the mechanical key to the gold locked in the sand, at the wrong price, and put it down; it was the one operator with the sand, the plant and the cash to try. It refused to run uranium through its gold circuit, and was right. And in 2012 it sold Blyvooruitzicht to Village Main Reef for about nine million dollars of shares and a nominal sum, with a rehabilitation fund of about two million dollars against a requirement of nearly five: the mine changed hands cheap because the clean-up it carried was worth more than the gold left in it. The market paid about US$310 for a DRDGold reserve ounce, and I said I would not bid for its ounces at that price; the market had done that work. By title it is probably the largest private landowner in the middle of Johannesburg, and Niël Pretorius has turned sand at 0.193 grams a tonne into the most profitable gold operation on the basin. Pan African was the hybrid by design, and it was the hybrid first: surface ounces built to carry the underground ones. Harmony became a hybrid by accretion, seven years behind; the mid-tier was the pathfinder and the country's largest gold miner the follower. Cobus Loots is the thesis of these papers made flesh, and one number proves it. Mintails, 2.36 million ounces of ground west of Soweto, cost him about three million dollars at today's rate, about US$1.30 a resource ounce, the lowest rung on the ladder, against the US$310 the market pays for a declared DRDGold ounce. Fairview at Barberton, in continuous production since 1886, commercialised biological oxidation for the world in 1986; the process is now owned in Helsinki, and I warned that the royalty on the Witwatersrand's second act will flow there unless the operators here build the reference plant first. My score was a winner: the London-listed scaler. Gold Fields and AngloGold Ashanti I called the absentees. Nick Holland's unbundling of Gold Fields' deep mines into Sibanye, announced in November 2012 and done in February 2013, was what it looked like: within three years the cast-off company had overtaken its parent. The past he gave away thrived in a builder's hands; the future he kept, South Deep, devoured capital for a decade and then worked, a white elephant turned into an 83-year mine, the way that takes twenty years of nerve. The skill did not vanish. It emigrated, to St Ives and Agnew in Australia. On AngloGold I was unsparing. The relay of chief executives, Cutifani, Venkatakrishnan, Dushnisky, Calderón, was congratulated on the tidiness of the exit. Before them, Cynthia Carroll sold Anglo American's last 11.3 per cent of the company to John Paulson for US$1.28 billion in March 2009, at US$900 gold. The Dushnisky board sold Mponeng in 2020. The reef-boring programme, US$30 to 40 million spent on the technology that might have mined the deep reef without men, was let lapse; and the company still owns the largest share of Rand Refinery, the refinery it is leaving to idle. When the leavers' shares were at their highs the market's verdict was plain, and I gave it without flinching: AngloGold a winner, the clean exit; Gold Fields a winner, left early, by structure; and the most valuable South African gold company in the world is the one that stopped being South African. But the test this paper applies was already on the record. AngloGold did not reprice its gold. It moved its address. And the discount does not vanish when a company escapes it; whoever stays inherits it. What I did not say plainly enough is that the exit was a desertion. The company that mined nearly half the country's gold at its birth kept the Rand Refinery stake, the silicosis book and a Johannesburg office, and took the rest to Denver. The Oppenheimer rule, profits earned so as to make a real and permanent contribution to Southern Africa, was the charter of the house that made it. The sale was dressed as portfolio focus. It was the abandonment of a home, for two hundred million dollars in cash. That is where the record stood. This paper adds the audit the record could not yet make: the leavers' ledger, in which the assets they sold are valued by what they have since earned their buyers; ten years of the market's verdict in one chart; the six ranked against one another on one day; and a count of the gold the official figure leaves out. The scorecard of share prices stands. The ledger here is a scorecard of assets, and on it the leavers lose. Two men were left as swing votes, Mr Nel at Harmony and Dr Stewart at Sibanye, and the year since has moved both toward the builders: Dr Stewart with Burnstone, Mr Nel with eleven years of guidance met, the largest surface business in the country and ten million ounces converted to reserve. The hedge and the copper are argued below as decisions I would not have taken; they are argued against a board that has earned the right to be judged on results. On share prices the record scored the leavers winners. On the ground, the audit scores them sellers who sold cheap and left; and on the business they run today, it scores them first. EXHIBIT B · THE SHADOW OUNCE 04 The fifth producer, which nobody lists The ninety tonnes is the official count. I do not accept it as the country's output. Other people's books show why, since our own are no use. Ninety tonnes is the floor. The choice is one line, and it has not changed in four years: the gold is leaving the country either way, and the only choice left is whether it leaves taxed or stolen. And the man at the bottom is in his place: the zama zama is not the crime, the zama zama is the verdict. South Africa's gold statistics are compiled from what the licensed mines declare and what Rand Refinery refines, and the two agree: the refinery took 90.0 tonnes from South African mines in its 2024 year and 87 of the 217 tonnes it refined in 2025, about forty per cent of its intake, and the Minerals Council counts about ninety tonnes a year. Everything else that leaves the ground, and a good deal of what leaves the country, is outside both books. The Minerals Council itself says in its 2025 Facts and Figures that most gold exports are "unclassified" in the revenue service's trade data and are excluded from its tables. The state that collects a royalty on gold cannot see its own gold exports. It says so itself. The London Bullion Market Association's own refiners record where their gold comes from and what kind it is. Recycled gold is scrap: jewellery, dental alloy, industrial waste, old bars. South Africa's legitimate scrap market, on Metals Focus's count, was 25 to 30 tonnes in 2018 and 2019 and about ten in every other year, and Rand Refinery, which handles almost all of the country's mined gold, says that "very little secondary material is sourced from within South Africa" and puts the true local scrap market at two to five tonnes a year. In 2019 the accredited refiners booked 64.8 tonnes of South African "scrap" against a declared mine output of 105 tonnes. Sixty-two per cent of the country's official production arrived in the vaults of London's own refiners labelled as something it was not. Rand Refinery's chief executive said in 2021 that "much of this scrap may contain material from illegal miners", and that smaller refiners' scrap exports "exceed tens of tons every year". Then the line collapsed: the Swiss intake fell from 49.4 tonnes in 2019 to 0.2 in 2021. Nothing happened to the dumps, the shafts or the men underground in those two years that cut the shadow trade by nine-tenths. The gold moved. In the same years the United Arab Emirates recorded 43, 41 and 31 tonnes of gold imports from South Africa against Rand Refinery's declared exports to the Emirates of eight, nil and nil, and an Emirati trader told SWISSAID in 2022 that most of that metal was artisanal gold, dug by zama zamas or smuggled in from a neighbour. Table 1 sets the two mirrors against the declared output, year by year. .TABLE 1 Two mirrors the country's own trade statistics cannot see, against the declared output. LBMA country-of-origin data 2018–24 (2023 includes 7.4 t of "legacy stocks"); UAE imports as charted by SWISSAID from Comtrade, 2018–23 (the 2024 return was posted on 31 October 2025 and withdrawn). Sources in the references. .Some of what the mirrors carry is legitimate: ten tonnes of real scrap, a few tonnes of Krugerrands bought free of value-added tax and exported as coin, and some African doré that came into South Africa for a small refiner and went out again. Some of it is double-counted, because a bar that goes to Dubai can come to Zurich later. And some of it is South African gold that never saw a royalty, a tax return or a reserve statement. The published estimates of that last part run from seven tonnes to fifty-one; Table 2 lists them. I do not believe the ninety tonnes. Take the published range, take the mirrors, 33 to 75 tonnes a year in two channels alone, and the only reading that fits both is that the official count understates the gold that leaves South African ground by between a third and a half. The country produces, on that reading, 120 to 135 tonnes of gold a year. The fifth producer, the one that files no reserve statement, is the size of Sibanye's gold division and may be the size of Harmony's underground mines, and it pays no royalty, no tax, no wages that appear in any survey, and no security bill. .TABLE 2 Estimates of South Africa's undeclared gold. Sources in the references; the last row is the author's reading of the mirrors and the published range. .FIGURE 3 South Africa's declared mine output to 2025, tonnes, against two mirrors the country's own trade statistics cannot see: "recycled" gold from South Africa recorded by all LBMA-accredited refiners, from the association's country-of-origin data (2018 to 2024; 2023 includes 7.4 tonnes of "legacy stocks"), and gold imports from South Africa as reported by the United Arab Emirates, published to 2023 (the UAE's 2024 return was posted to Comtrade on 31 October 2025 and withdrawn). Sources in the references. .Who digs it is known, and so is who owns it. The Institute for Security Studies' 2019 study described four tiers. At the bottom, some 30,000 illegal miners, paid as little as a quarter of the spot price for amalgam about sixty per cent pure, earning four and a half to seven and a half thousand rand a month and paying shaft-access fees of 22,000 to 66,000 rand. Above them, the syndicate bosses who recruit, arm and supply them. Then the buyers: licensed scrap dealers among them, and some of the 182 refining licences issued in the ten years to 2023, more than eighty per cent of them in Gauteng. At the top, the holders of export permits who move the metal to Dubai and, through the scrap books, to Zurich. The listed four know it in their own accounts, and Table 3 shows what it costs them. The convictions on the record in the year to October 2026 are five foreign nationals sentenced in Welkom, and two police generals and a Durban businessman charged in May 2026 and presumed innocent. No dealer and no exporter has been convicted of anything. .TABLE 3 The shadow producer in the listed companies' own accounts. Sources: company reports and announcements as listed in the references, Harmony's security figures from the Integrated Annual Report 2026 filed with its Form 20-F on 30 October 2026; Arrive Alive on the Welkom sentences. .Why a shareholder in the listed four should care is in four numbers. The listed four are producing in a country whose true gold output is a third to a half larger than the number the market uses, and the increment is produced by competitors that pay none of the four's costs. The four are paying to police that competitor, Harmony alone more than R750 million a year in South Africa, and every rand of it is a transfer from the listed shareholder to the state's failure to police its own ground. The shadow ounce is also the raw material of the surface business the four are building, because the dumps the zama zamas rework by hand at the edge are the same dumps DRDGold and Harmony reclaim by pipeline, and the surface company that licenses the artisanal tier of the trade, as Ghana licensed its small-scale miners, converts a competitor into a supplier. And a country that undercounts its own gold output by a third does not know what its own reserve is, what its own royalty base is, or what the listed companies that mine it are worth. It knows what they declare. The LBMA's refiners know more. The country declares ninety tonnes. London's own refiners took sixty-five tonnes of South African "scrap" in one year, and Dubai forty-three in another. The fifth producer pays no royalty, no tax and no security bill, and the four who stayed pay its policing. EXHIBIT C · THE CONSOLIDATOR 05 Harmony: built from what the others sold The verdict on Harmony is in two lines. It is the buyer that proved the sellers wrong, the best operator of deep South African ground alive, the company that turned a US$200 million purchase into a mine earning US$570 million a year, and the reason the deep Witwatersrand is still a working goldfield rather than a chain of flooded shafts. It is also, today, the second-highest-cost producer of the six, hedged, and committing the next four years of the Witwatersrand's cash to copper in Australia. The market rates it at 5.6 times earnings and US$6,400 an annual ounce, little more than half DRDGold's, and most of that discount is explained by decisions the board can revisit. The evidence for both lines follows. Harmony was registered on 25 August 1950 to work a single lease in the Free State, a Rand Mines subsidiary that spent its first forty-five years as one shaft among many. Its second life began in 1995, when Bernard Swanepoel took a company valued at a few hundred million rand and set about buying what the mining houses were disposing of: Evander from Gencor in 1998, Kalgold's open pit in 1999, the Free State shafts the majors called marginal, and in 2003, in the transaction that brought Patrice Motsepe to the chair, the ARMgold mines. The method was the same every time: take a mine the seller had written down, strip its overhead, run it lean, and let the rand price do the rest. Swanepoel left in 2007. The method stayed. Graham Briggs built Hidden Valley in Papua New Guinea. Peter Steenkamp bought Moab Khotsong and Great Noligwa from AngloGold Ashanti in March 2018 for about R3.6 billion, and Mponeng, TauTona and Mine Waste Solutions from the same seller in October 2020 for about R4.4 billion more. Beyers Nel, chief executive since January 2025, has begun spending the proceeds of that thirty-year purchase on copper: CSA in New South Wales, about R16.5 billion, in October 2025, and a month later the go-ahead for Eva Copper in Queensland. Behind them waits half of Wafi-Golpu in Papua New Guinea, one of the world's great undeveloped copper-gold deposits, still without its lease. On 15 October the board took back Alwyn Pretorius, its chief operating officer until 2015, after a decade at Newmont. The result is the largest gold producer in the country by a distance, and the only one of the four that is a company of the first rank by world standards: 1,429,551 ounces in the year to June 2026, within guidance for the eleventh consecutive year, from nine underground mines, six surface sources and Hidden Valley. The year was the best in its history on every line that matters: adjusted free cash flow R17.1 billion, up 54 per cent, the highest dividend it has ever paid, and net debt of R852 million after paying for CSA. Table 5 has the figures. Where the cash came from matters more than the total. Mponeng, the deepest mine on earth, generated R9.4 billion of free cash, more than half the group's. Mponeng and Moab Khotsong together, the two high-grade mines bought from AngloGold, ran a 38 per cent free-cash margin; the seven optimised underground mines ran 25, and that is where the grade is going: the underground recovered grade fell from 6.27 to 5.83 grams in the year, and Target 1 cost more than the price of the metal. Mponeng's eleven-gram grade should be modelled back toward reserve grade as the high-grade zones normalise, and FY2026 is the first year of that normalising. The surface operations produced 221,199 ounces and R6.6 billion of free cash at a 46 per cent margin: the biggest surface gold producer in the country, bigger than DRDGold, inside a company the market rates as a deep-level miner. And Hidden Valley, the Papua New Guinea open pit that Harmony took to full ownership in 2016 for a nominal sum, ran a 68 per cent margin, the best in the group, on five years of reserve. Harmony is the receptacle for what the majors discarded, the reef's buyer of last resort, and the figures above are what that role has earned. I rate its management the best operators of deep South African ground alive. No South African gold company has met guidance eleven years running. None has twice bought a mine from AngloGold for about R4 billion and made it one of the best it owns. None has a cleaner capital account at the end of a year in which it spent R17.1 billion, R4.6 billion of it on a copper project that has yet to produce a tonne. The reserve statement as at 30 June 2026, published on 23 October, carries 27.4 million ounces of gold, up on the year after some ten million ounces were converted at Moab, Tshepong North and Doornkop at about R3,000 an ounce. The price behind those ounces is US$2,915, about a quarter below the gold price of the day and within a hundred dollars of the convention Sibanye used to give Kloof a year. The lives by shaft, in Table 4, say where the future is: the two AngloGold mines carry 7.7 of the 12.6 million underground reserve ounces and the two longest lives, and the Free State is short-dated at this reserve price. Copper reserves of 4.0 million tonnes contained, up 71 per cent, now sit beside the gold on the first page of the statement. The company employed 34,350 people and 12,761 contractors in the year to June 2025, paid them R20.2 billion, and paid R6 billion in taxes and royalties in South Africa; the FY2026 dividend was R8.1 billion. Every one of those numbers was, at the moment of each purchase, a number the seller had decided to walk away from. The alternative to Harmony's cheque was closure, and the country knows what closure looks like: Blyvooruitzicht, liquidated in 2013 with about two million dollars in its rehabilitation fund against nearly five required, its village and shafts given over to the illegal trade; Grootvlei, where the pumps stopped in 2011 and the Eastern Basin filled; Buffelsfontein at Stilfontein, closed in 2013, from whose shaft seventy-eight bodies were recovered in January 2025. The state's own count is about 6,100 derelict and ownerless mines and a R14 billion rehabilitation need against a budget of R124 million in 2022. Harmony's nine underground mines are nine entries not on that list. The company says its South African gold business “is starting to show the world that gold mining in South Africa is a good business”. The accounts bear it out. .TABLE 4 Underground reserve 12.6 Moz at 6.5 g/t; resource 76.8 Moz at 9.8 g/t; reserve price US$2,915/oz (R1,650,000/kg at R17.60), resources at US$3,061. Source: Harmony Mineral Resources and Mineral Reserves 2026 and Operational Report 2026, 23 October 2026. .Against that I hold four things. The first three are the board's own doing. All-in sustaining cost rose 13 per cent in the year, the highest of the six after Sibanye's deep mines, and FY2027 guidance adds another 9 to 17 per cent on up to 9 per cent fewer ounces, as Moab Khotsong moves into an ore gap its own plans show running from 2027 to 2031 while the Zaaiplaats extension is built: R4.5 billion committed in August 2021 for a twin decline four hundred metres deeper on the same reef, 225,000 ounces a year over twenty-four years at nine grams. That is the mine that closes the gap, and the proof that the company builds on its own cash flow. The hedge book is the one I would argue against hardest. It cost R9.65 billion of realised losses in the year, a tenth of revenue handed to a bank for insurance against the one thing a gold company's shareholders bought it for, and it still carries 618,000 ounces of rand-gold collars capped about a sixth above the spot price of 30 October. Harmony's price received was a tenth below what DRDGold, which does not hedge, received for the same metal. The illegal-mining bill is the largest in the industry: R780 million in the year to June 2026, R751 million of it in South Africa, up from R748 million, with arrests at its working shafts up 255 per cent as the syndicates followed the gold price, and the police operation that produced most of its arrests, Vala Umgodi, closed in December 2025 for want of state funding. The fourth is that the free cash is going to Australia, and the balance sheet is being arranged to send it. Net debt of R852 million at June understates what is coming: R9.5 billion of the borrowings were current at that date, and on 29 September the company sold US$500 million of convertible bonds due 2031 at a coupon of 1.5 per cent, convertible at R418.60, 40 per cent above the share price of the day and 70 per cent above the close of 30 October; the shares that would answer them are 3 per cent of the register. FY2027 capital is R14.4 billion for gold, R2 billion for CSA and about R11 billion for Eva, which the company will build to 60,000 tonnes of copper a year for first production at the end of 2028, subject to a federal approval delayed by the discovery of a protected skink. Mr Nel told the Mining Forum Americas on 28 September that Australasia will be 30 per cent of the group within a decade and that the cash inflection comes beyond 2030. His other line was the better one: "it's the gold business in South Africa, where it all began, that's starting to show the world that gold mining in South Africa is a good business". Eleven years of guidance met is the record behind it. A shareholder who bought Harmony for the Witwatersrand is being asked to wait four years for copper, bought at the top of a copper cycle by a board whose entire record is in gold. Offered a re-rated second metal in its own dumps and one eleven thousand kilometres away, the country's largest gold house bought the distant one. I would not have done it. The share price, down 27 per cent this year with gold down ten, says the market would not have either. The judgement is against a board whose record entitles it to be wrong once: it bought Mponeng against the same market's advice. .TABLE 5 Harmony's year to June 2026 and its engines. Sources: results of 27 August 2026 and the results booklet; operational report of 23 October 2026; market data of 30 October 2026; the per-ounce rating is the author's arithmetic. Rand figures are the companies' own; US dollar counterparts are at R16.48 (accounts) and R16.71 (market), 32.1507 oz/kg, and differ by a few per cent from the companies' dollar figures struck at period-average rates. .The best year in its history, on the deepest mine on earth, bought for about four and a half billion rand. The next four years of cash go to copper. EXHIBIT D · THE DIVERSIFIER 06 Sibanye-Stillwater: the gold inside a platinum company The verdict on Sibanye is that it found the right strategy for its ground and has not yet built the company that would be paid for it. The strategy, surface ounces first and a new shallow mine, is the one this paper argues for, and Burnstone and the DRDGold stake are its proof. The company kept open three deep mines their owner had stamped end of life, and has run them for thirteen years for the dividend, the wage bill and the pumps. It is a platinum group in which gold is 28 per cent of earnings, whose deep mines are the highest-cost and thinnest-margin ounces in the book, and whose gold is rated at about US$2,700 an annual ounce because nobody can own it on its own. That is the one discount in this paper a single board decision removes. Sibanye was listed on 11 February 2013 as the part of Gold Fields that Gold Fields no longer wanted: the Kloof-Driefontein Complex and Beatrix, deep, labour-intensive, unmechanised, with the words end of life stamped on them and a dividend promise attached. The founding story of the company is a man who looked at those words and took the label off. Neal Froneman took the unbundled mines, cut their overheads, kept the shafts and the pumps running and paid the promised dividends; the South African region he built around them employed about 81,500 people and contractors at the end of 2022 and about 70,000 after the restructuring of 2023 and 2024, gold and platinum together, with forced retrenchments held, on the company's own count, to 8 per cent of those affected. Within four years he had used them as the balance sheet for everything else: Cooke from Gold One in 2013, Wits Gold and its Burnstone project in 2014, Aquarius Platinum and Anglo American Platinum's Rustenburg mines in 2016, Stillwater in Montana for about R29 billion in 2017, 38 per cent of DRDGold in 2018 for a parcel of dumps and the remaining 12.1 per cent in 2020 for R1.09 billion in cash, Lonmin in 2019, a Finnish lithium project, a French nickel refinery, an American recycler. By 2025 the company that began as two gold complexes was a platinum-group-metals producer on two continents with a gold division attached. Froneman retired on 30 September 2025. Richard Stewart, the geologist who ran its southern African operations, took over the next day. Where it is today is a company whose half-year to June 2026 was its best ever, for a reason that has nothing to do with gold. Adjusted EBITDA doubled to R31.8 billion. Sixty per cent of it came from the South African platinum mines on a basket price up 67 per cent. Gold was 28 per cent, and two-fifths of the gold came from DRDGold, which Sibanye consolidates in full and owns half of. Net debt halved to R9.7 billion and the interim dividend, R5.7 billion, was paid at the top of its policy. Table 6 has the figures. The gold division, read on its own, is the highest-cost set of mines in this paper, and the ground under them is not the reason. Kloof's resource grade of 6.7 grams is above South Deep's reserve grade. The one-year life is the gold that did not move under assumptions that did. Kloof's reserve was cut in February 2026 from 1.6 million ounces to 100,000, three weeks after gold printed its record, and the company's own words are that the mine has no long-life reserves and will be renewed as the gold price allows. Its margin in the half was about US$148 an ounce, the thinnest in the company's South African gold portfolio. Dr Stewart said in October 2025 that the three deep mines had four to ten years between them, Driefontein the longest, Beatrix three or four. The resource was cut from 40.2 million ounces to 27.5, Kloof's alone from 14.5 to 4.0, at a reserve price raised to US$2,837 an ounce. The room that price created was spent writing the reserve off, not mining it. Guidance for 2026 puts the cost of the company's own mines within a quarter of the spot price. Two things redeem it. Burnstone, mothballed, once liquidated and twice deferred, was approved on 1 September 2026 as a twenty-five-year mine at a 36.1 per cent return, with a plant that is largely built. It is the second new underground gold mine in the country inside six months, after fifteen years without one. And DRDGold: Sibanye put in about R2 billion of dumps and cash between 2017 and 2020, and its 50.1 per cent was worth R16.5 billion on 30 October, more than eight times, 16 per cent of Sibanye's own market value. The chief executive says he would like more exposure to it, and the Kloof 2 dump went across in December 2025 for 0.52 million ounces and four years of DRDGold's life. The strategy, in the company's words, is a portfolio transition to shallower and surface ounces as the deep mines are phased out, and that is the right strategy for the ground it holds. The credit is not unqualified, and the reason is disclosure. Every reserve statement Sibanye and its peers have published in this cycle discloses the gold price behind it to the dollar; Burnstone's restated study, in anything published so far, does not. The board discloses its price to the ounce at Kloof, to explain one year of life, and withholds it at Burnstone, to justify R6.2 billion of conviction. Rene Hochreiter's reading was that the mine "is not a dripping roast. At current gold prices, it makes a profit, but not much lower". On the study's own numbers the margin is near 57 per cent, and about 13 if costs double, on a faulted, discontinuous orebody that may not carry the tonnages in the plan. That is what broke Great Basin on the same ground in 2012, when the ramp-up met faulting and water and the year's guidance fell from 90,000 ounces to 30,000. The mining right runs to 16 February 2027, within weeks of the start of mining, and the 2025 Bill carries no transitional clause for empowerment credentials earned under the old order; Dr Stewart's own words were that "historically, Burnstone has been an empowered asset". The project came to Sibanye with Wits Gold in 2014, and Wits Gold had taken it out of Great Basin's liquidation for US$7.25 million and the assumption of about US$177 million of debt. A liquidation price is a seller's admission that an asset is worth its liabilities, and on Burnstone, as on Blyvooruitzicht in 2012, the admission has been wrong. Nor was Burnstone the company's first choice of where to spend. It was the gold project that cleared the bar while a platinum project cleared it faster: approved in 2021 at R2.8 billion plus R1.5 billion of pre-production and a 20 per cent return, deferred in November 2023 "to preserve cash" in the year of R47.5 billion of impairments, and restated in 2026 at R6.2 billion and 36.1 per cent. And the arithmetic on Kloof stands. At Burnstone's capital intensity, about R2,300 a reserve ounce, Kloof's surviving 100,000 ounces would cost about R230 million to bring to production, and its four million excluded ounces are a ten-billion-rand question against nine billion of gold EBITDA in six months. Affordability was never the constraint. Burnstone does not lower the bar Kloof has to clear; it raises it, by removing the one excuse that might have covered the decision, that the discipline asked for is beyond this board's reach. Then look at what the market pays for the lot. Sibanye's enterprise value on 30 October was 6.7 billion dollars, 1.7 times its annualised half-year EBITDA, the lowest multiple of the six by a distance. The DRDGold stake alone is about a billion dollars of that. Take the gold division's EBITDA without DRDGold, about 330 million dollars for the half, 655 million annualised, and value it at the group's own multiple: about 1.1 billion dollars for 426,000 ounces a year from Driefontein, Kloof, Beatrix and Cooke. That is about US$2,700 per annual ounce, against US$6,400 at Harmony, US$11,600 at DRDGold and US$13,300 at AngloGold Ashanti. I do not think the market is wrong. It is pricing deep mines whose owner has said they have four to ten years, whose richest reef carries a year, and whose costs are the highest in the book. They sit inside a group whose share has fallen 41 per cent in 2026 on a palladium price that could not hold, an American operation that lost five weeks to a strike in September and October, a lithium refinery still commissioning, and a Section 45X tax credit the IRS is examining. They also sit under a balance sheet the market has not forgiven. Net debt of R9.7 billion is 0.18 times an EBITDA struck at a record platinum basket; gross debt is R32.1 billion, three times the net figure; net debt was R22.1 billion six months earlier; the US$500 million Franco-Nevada stream of February 2025 is repaid in metal over twenty-five years and appears in no debt figure; and the board's stated priority is to halve gross debt within two to three years, which is the cash that would otherwise go to Kloof. The multiple of 1.7 times is the market's verdict that it does not believe the EBITDA will outlast the debt, and the sum-of-parts applies that multiple to mines that carry none of the borrowing. The gold is worth more than three thousand dollars an annual ounce to anyone who owns it on its own. Inside this group nobody can. That is a board decision, not a geological one. The mechanism is Sibanye's own capital-allocation framework. Ore-reserve development ranks ahead of growth. Burnstone was entered and won. Kloof was not entered at all. .FIGURE 4 Sibanye's enterprise value in three parts at the closes of 30 October 2026: the platinum, palladium, lithium, nickel and recycling businesses; the DRDGold stake at market; and the deep gold at the group's own multiple. Author's sum-of-parts. .TABLE 6 Sibanye-Stillwater's South African gold mines in the half-year to June 2026, and the group around them. Sources: results of 1 September 2026; reserve declaration of 17 February 2026; market data of 30 October 2026. The deep-gold rating is the author's sum-of-parts. Rand figures are the companies' own; US dollar counterparts are at R16.48 (accounts) and R16.71 (market), 32.1507 oz/kg, and differ by a few per cent from the companies' dollar figures struck at period-average rates. .Sixty per cent platinum, twenty-eight per cent gold, and the gold rated at under three thousand dollars an annual ounce. The company that kept the three mines Gold Fields stamped end of life, and found the right strategy for them, is the one the market has not yet paid for it. EXHIBIT E · THE SURFACE SPECIALIST 07 DRDGold: the dumps as a business The verdict on DRDGold is the shortest in the paper. It is the second-best company of the six on the measures ranked in Section XIV, the cleanest gold business on the exchange, and the proof that a South African ounce can be rated with the world's seniors without leaving: US$11,600 an annual ounce, within four per cent of Gold Fields. The two things that count against it are not of its making: a fixed stock of dumps, and a parent that owns half. Durban Roodepoort Deep was registered in 1895, one of the first deep mines on the central Rand, and for a century it was what its name said. The underground mines went one by one: Durban Deep itself in 2000, Blyvooruitzicht sold in 2012 and liquidated the year after, ERPM's shafts put up for sale in 2013. What survived was the thing the company had done on the side since it bought Crown Mines in 1998: washing the old dumps of the central Rand through a plant at Crown. In 2007 it took half of ERGO, the East Rand plant Anglo American had built in 1977 to treat slimes, from Mintails, and by January 2010 all of it; from then on the company was a surface business only. Niël Pretorius, chief executive since 2009, has run it as one: no underground, no exploration, a dividend every year, and one argument to the market, that a dump at a fifth of a gram pays when the plant is big enough and the price is right. The 2018 transaction with Sibanye gave it the West Rand dumps and a controlling shareholder; the 2020 option gave Sibanye 50.1 per cent. The year to June 2026 was the best in the company's history: 155,577 ounces from sand at 0.193 grams a tonne, more than the top of its guidance, an operating margin of 58 per cent, no bank debt, and a nineteenth consecutive dividend. The two operations are not alike. Ergo on the East Rand is the volume; Far West Gold Recoveries on the West Rand, the dumps Sibanye vended in, is the better business, at half Ergo's cost and a 76 per cent operating margin. The R10 billion programme the company calls Vision 2028 is built around the West Rand, and its target is six tonnes a year by FY2028 from three million tonnes a month. The first quarter of the new year, reported on 20 October 2026, ran ahead of the record, and guidance for the year stands. Table 7 has the figures. .TABLE 7 DRDGold's record year and the quarter after it. Sources: results of 19 August 2026; operating update of 20 October 2026; reserve statement at 30 June 2026; market data of 30 October 2026. Rand figures are the companies' own; US dollar counterparts are at R16.48 (accounts) and R16.71 (market), 32.1507 oz/kg, and differ by a few per cent from the companies' dollar figures struck at period-average rates. .I admire this business more than any other on the exchange. The reason is a plain one, which is a caution as well as a compliment: it is a leveraged bet on the gold price laid over a fixed stock of waste, a magnificent way to harvest what has already been won, and not, on its own, a future for the basin. No one goes underground, nothing bursts, nothing has to be cooled, and the shift starts at a pump. The gold was brought to surface by a previous generation at a cost in lives DRDGold does not carry, and the company's product, besides the metal, is the removal of dumps from beside townships and the return of the ground. Its reserve is 6.24 million ounces, 3.65 at Ergo and 2.59 at Far West, with a life the company puts at more than twenty years, to 2043 and 2045 on Sibanye's own reckoning. Its operating margin at today's price is the widest of the four who stayed. It has paid nineteen consecutive dividends through a decade in which the gold price fell by half and recovered fourfold. And it has begun, in its own words, to look for tailings partners on two continents, Africa and South America, as a capital partner that brings the process skill and takes a share of the proceeds. A company that started as a major's non-core dumps now has an aspiration of its own. The rating is explained in five words, no reserve cliff to fall off, and that still holds. What should be understood about it, rather than held against it, is that at a fifth of a gram every cost is a function of tonnage and the margin is almost entirely the price: 58 per cent for the year at R2.1 to R2.3 million a kilogram, 45 at the FY2025 average, a third at FY2024's. No company in this paper has more operating leverage to the rand gold price, and the share, down more than a quarter in 2026 against a gold price down ten per cent in dollars, is the market pricing that leverage both ways. The dumps are finite, and the growth after Vision 2028 depends on dumps other people own; the company has begun to look for them on two continents. The one thing I do hold against it is control. Sibanye owns 50.1 per cent, consolidates it, supplies its best feed and has said it wants more. A minority shareholder in DRDGold owns a share of a subsidiary whose parent has said it wants more of it, at a price the minority does not set, and the minority has never asked loudly what the parent intends. It should. A dump at a fifth of a gram, a 58 per cent margin, nineteen dividends and no one underground. The price is doing the work, and the parent owns half. EXHIBIT F · THE HYBRID 08 Pan African: both roads at once The verdict on Pan African is that it is the best company of the six. It is first on the seven measures ranked in Section XIV, the lowest-cost producer in the book at US$1,867 an ounce, the only one of the six growing, unhedged, ungeared, and the hybrid of surface and deep ounces that this paper argues the whole basin should become. The market rates it at 7.9 times earnings and US$11,600 an annual ounce, within four per cent of Gold Fields and nothing like the premium its record would command in Perth or Toronto. The one decision I would reverse is a thirty-million-dollar buy-back. Pan African Resources was incorporated in 2000 as an exploration company with nothing in production. In 2007 it bought Barberton Mines, Fairview, Sheba and Consort, from Metorex: the oldest working goldfield in the country, mined since 1884, narrow, refractory, and the place where the BIOX bacterial process was commercialised in 1986. In 2013 it bought Evander from Harmony for R1.5 billion. The second act was surface: the Barberton tailings plant in 2013, Elikhulu at Evander in 2018 for about R1.7 billion, and in 2022 the Mintails estate west of Soweto from a liquidator for R50 million, built as the Mogale plant for about R2.2 billion in fourteen months and poured on 3 October 2024, under budget; Mr Loots has since committed R5.3 billion more to the West Rand and Soweto. The third act is abroad and back underground: Tennant Creek in Australia's Northern Territory, bought at the end of 2024 and taken to full ownership in July 2026, and Royal Sheba, a new mine on the Barberton fault with its first development blast due in January 2027. Cobus Loots has been chief executive since February 2015, and he is the thesis made flesh: the whole argument of this paper, that the cheap ounces on the surface can carry the hard ounces underground, run as a company for a decade. On his own count the share has returned seventy-fold since 2009 on a fivefold gold price. The comparison that matters for this paper is this: Pan African is the hybrid by design and was the hybrid first; Harmony is a hybrid by accretion, seven years behind; the mid-tier is the pathfinder and the largest deep miner in the country the follower. The year to June 2026 was a record on every line: 272,310 ounces, up 38.6 per cent, at US$1,867 an ounce all-in, the lowest cost of the six, and headline earnings three times the year before. Net debt of US$150 million became net cash of US$186 million after capital and the dividend, with no hedge. The spread is the design: six sources, none above 27 per cent of the ounces. Three tailings plants produced 45 per cent of the group at a weighted cost of about US$1,330, down from about 60 per cent of the group only because Tennant Creek's underground ounces came in. Evander's new sub-vertical shaft lifted its output 68 per cent and its grade from 6.8 to above 11 grams. Tennant Creek came in at 32,124 ounces against a 50,000-ounce plan. Reserves were replaced at 13.04 million ounces at a reserve price raised to US$2,400. Table 8 has the figures..TABLE 8 Pan African's six sources and its record year. Sources: results of 16 September 2026; market data of 30 October 2026. US dollar figures are the companies' own; rand counterparts are at R16.48 (accounts) and R16.71 (market), 32.1507 oz/kg. FY2027 guidance in rand at the company's own R17.00. .It is the cheapest, the least geared, the most diversified and the only one of the four that is growing. Three projects are in the price at nothing, and Table 9 sets them out. Soweto's feasibility study is done, at a 29.55 per cent return on a gold price about a tenth below spot, and the investment decision is due in December 2026 or, on the chief executive's words at the results call, early to mid-2027 once the water-use licence and the servitudes are in hand. Royal Sheba is the country's third new gold mine in a year, at about US$375 per annual ounce of capacity against Burnstone's US$2,900. Poplar is the deep one, inside the Evander mining right, with no decision before 2028. The second half of FY2026 ran at an annualised 290,000 ounces, and guidance for FY2027 is 7 per cent higher again. .TABLE 9 Pan African's three projects. Sources: Soweto definitive feasibility study, SENS of 11 September 2026; results and call of 16 September 2026. .Two things qualify the record, and one I would reverse. The first is the share count: Tennant Creek has cost a fifth more shares, 1,979 million to 2,434 million, for a mine running at US$2,261 an ounce and a ramp-up the company itself calls slower than hoped, and on the full count FY2027 headline earnings at the spot price are about 14 US cents against 17.64, arithmetic rather than deterioration, but arithmetic the market will see in February. The second is that the sector's cost inflation reaches Pan African too, guided up 14.5 per cent in FY2027, though it starts lowest and the group without Sheba and Consort is at US$1,702. The one I would reverse is the buy-back of up to R500 million, about US$30 million, approved from October and not started on the record by 23 October. It is under one per cent of the market value. It sits beside a dividend the announcement puts at 31.8 per cent of discretionary cash against a 40 to 50 per cent policy, in a year with a US$294 million programme to fund. US$30 million into Soweto's first year is a plant; US$30 million into the London order book is a gesture. BlackRock, meanwhile, crossed five per cent of the register on 19 October, fell below it on the 22nd and was back at 5.00 per cent on the 23rd, two per cent of it through securities lending: the index world arriving at the one South African gold company it had not yet needed to own, and hovering at the line. .Lowest cost, no debt, no hedge, growing, and three projects at zero in the price. The one thing I would change is thirty million dollars in the order book. EXHIBIT G · THE TWO WHO LEFT 09 Gold Fields: one mine and a passport The verdict on Gold Fields is that in 2013 it gave away the mines that made it, kept the address, and has run a good business since from everywhere but the Far West Rand. It gave away three mines that have paid someone else US$2.7 billion. It then built, through two impairments and a decade of ridicule, the one mechanised deep mine in the basin that works, and around it the second-lowest-cost producer of the six, with the widest margin in the book and a reserve into the next century. It is the third-best company of the six on the measures ranked in Section XIV. The market has it at 6.9 times, more than a quarter below where it stood before it offered A$38.7 billion of its own paper for Northern Star. Gold Fields of South Africa was Cecil Rhodes's company, registered in London in 1887 to hold claims on the Rand. The modern Gold Fields was assembled in 1998 from its gold assets and Gencor's. For a decade it was the country's second producer, with Driefontein, Kloof and Beatrix. In 2006 it bought the one thing in the basin that was not an old reef: South Deep, the mechanised bulk mine on the Far West Rand, about three billion dollars in all for Barrick's half and Western Areas'. In February 2013 it unbundled everything else into Sibanye and kept South Deep, Ghana, Australia and Peru. The decision has aged two ways. South Deep took a decade longer than planned and absorbed capital and impairments before it worked; it now makes US$350 million of free cash in six months, double the year before, on a reserve of 27.6 million ounces and a life the company states as 83 years. It is twelve per cent of the group's production and fifty-seven per cent of its reserves. The rest of the company became Australian in all but address, with the best mine in the group in Chile, a project waiting on approval in Quebec, and leases in Ghana that expire in April 2027. The nearest thing to an answer came on 22 October, when the minister of lands told Bloomberg that it was not government policy to take mines from foreign companies and that Tarkwa's case differed from Damang's, which Gold Fields had been made to hand back in April; the president's brother's contractor is meanwhile claiming more than US$700 million from the company in arbitration over the two mines. Table 10 has the figures. .TABLE 10 Gold Fields on one page. Sources: interim results of 25 August 2026; reserve supplement, February 2026; the Northern Star announcements of 28 September 2026; market data of 30 October 2026. US dollar figures are the companies' own; rand counterparts are at R16.48 (accounts) and R16.71 (market), 32.1507 oz/kg. .The half-year to June 2026 was the best in Gold Fields' history: adjusted free cash flow up 134 per cent, the interim dividend up 132, net debt down to US$437 million. Then, on 13 September, it sent Northern Star Resources a proposal worth A$38.7 billion, three-quarters of it in Gold Fields paper, which would have made the Sandton-domiciled group the second-largest gold producer on earth with one South African mine in it. Northern Star's board rejected it unanimously as falling well short of fundamental value, and objected to the paper. Gold Fields' shares fell 12 per cent on the day it was disclosed. On 14 October Redwheel, a long-standing holder, published a letter asking the board to drop a bid it called dilutive. On 15 October Mike Fraser, the chief executive, said a higher offer would need reciprocation he was not hearing, and that walking away was a step from where the company stood. The share closed on 30 October more than a quarter below where it stood before the approach. The market's objection is the one it has to every scrip bid from a company on four times EBITDA for one on seven. The acquirer pays with cheap paper for dear assets. The precedent is in the company's own past: Chris Griffith's bid for Yamana in 2022 took as much as 23 per cent off the share on a single failed offer, and cost him the job. South Deep works, and the company that owns it built it, through two impairments and a decade of ridicule, which is more than the rest of the industry managed. It is what commitment bought, the way that takes twenty years of nerve against the way that takes a single August, and Nick Holland is owed fairness on both counts: the past he gave away thrived in a builder's hands, and the future he kept devoured capital before it paid. It is the one mechanised deep mine in the basin, the one with a reserve into the next century, and it is the mine that shows what commitment to the deep reef buys; the men who bedded the recovery down, Martin Preece through 2023 and Mike Fraser from 2024, inherited a cash machine the executive who endured the worst of it had already begun to build. What I cannot forgive is where the cash goes. Twelve per cent of the revenue is South African, and the growth capital of the last two years, Gold Road, Windfall, the Northern Star offer and, on 7 October, a nickel concentrator at Kambalda, has gone everywhere but the Far West Rand. The company has said it will drill 100,000 to 200,000 metres at South Deep over five years, mainly south of the Wrench fault, and that it is studying a step change into the 2030s. It has not put a number on it. A company that keeps its registered office at 150 Helen Road, Sandton, and twelve per cent of its ounces in the country, is a South African company in the way a Mauritian holding company is Mauritian. It is the quieter desertion of the two, and the sleight of hand is the address: South African for the register and the index, Australian for the capital. 10 AngloGold Ashanti: the exit, priced The verdict on AngloGold Ashanti holds two things at once. It is the best balance sheet in this paper, net cash of US$991 million, the largest reserve, 36.5 million ounces, and the dearest share, at 10.7 times earnings and US$13,300 an annual ounce, the New York premium in one number. It is also the author of the worst South African mining transaction of the decade, and of the cleanest desertion: the sale of the deepest mine on earth for US$200 million in cash, a mine that has since returned its buyer seven times that sum, followed three years later by the removal of the company itself to Denver. The second line does not cancel the first, and this paper does not pretend it does. AngloGold was Anglo American's gold division made a company in June 1998, the largest gold producer in the world at its birth and, after the Ashanti merger of April 2004, second only to Newmont. The house that made it had a written rule. Sir Ernest Oppenheimer set it down in 1954: the aims of the group were to earn profits, but to earn them so as to make "a real and permanent contribution" to the well-being of the people and the development of Southern Africa. Anglo American still quotes the sentence on its centenary page, with one edit: the contribution is now to the communities in which it operates, and Southern Africa is gone from it. Its gold child spent sixteen years leaving, and the leaving had a sequence that should be kept straight. Tony Trahar branded the South African mines non-core at Anglo American in the 2000s; Mark Cutifani kept them running, made them dramatically safer and fixed the hedge book, which is the exoneration the relay owes him; the substantive exit from the gold was Kelvin Dushnisky's board's in 2020; the relocation of the address was Alberto Calderón's in 2023, and the two should not be confused. Elandsrand and Deelkraal went to Harmony in 2001; the Free State shafts to ARM and Harmony; Tau Lekoa in 2010; Moab Khotsong and Great Noligwa to Harmony at the end of February 2018 for about US$300 million, and Kopanang to Heaven-Sent for R100 million; and in 2020 Mponeng, TauTona, Savuka and Mine Waste Solutions to Harmony for about US$300 million, US$200 million in cash and up to US$100 million deferred against production above 250,000 ounces a year. On 25 September 2023 it completed its move: a company registered in England, a corporate office in Denver, a primary listing in New York, a secondary one in Johannesburg, and 99 per cent of its employees outside South Africa. What remains in the country is a 42.41 per cent shareholding in Rand Refinery, the post-retirement medical and silicosis obligations of the mines it sold, and what its filings call a substantial corporate office in Johannesburg. The obligations stayed because they could not be sold; the mines went because they could. The company it became is the most valuable in this paper, and the best balance sheet in it: net cash of US$991 million at June after buying back US$666 million of its own bonds, a dividend doubled, and a reserve up 17 per cent to 36.5 million ounces. Its mines are in Ghana, Tanzania, Congo, Egypt, Guinea, Australia, Argentina and Brazil, and its growth project is in Nevada. Shareholders approved a US$2 billion buy-back on 23 July 2026. Its chief executive explained on 29 September why the buy-back approved in July had not started: the company that left South Africa still needed the Reserve Bank's permission to buy its own shares, and the permission came slower than the dividend. Its chief financial officer said in London on 16 October that the buy-back has to be opportunistic and that repurchases made no sense at these values: the right instinct, and one her board has given itself five years of authority to abandon. Table 11 has the figures. .TABLE 11 AngloGold Ashanti on one page. Sources: results for the six months to 30 June 2026, 31 July 2026; reserve report at 31 December 2025; market data of 30 October 2026. US dollar figures are the companies' own; rand counterparts are at R16.48 (accounts) and R16.71 (market), 32.1507 oz/kg. .The balance sheet, the reserve and the dividend are what a well-run major looks like, and there is the Rand Refinery stake, through which the company that left still owns the largest share of the gate every South African ounce goes through. The sale is the thing, and the finding it rests on is that AngloGold did not reprice its gold, it moved its address. Harmony paid about US$300 million, including the deferred tail, for the Mponeng package in 2020. In the year to June 2026 Mponeng alone generated US$570 million of adjusted free cash at today's rate, and Mine Waste Solutions, which came in the same package, is the core of a surface business that generated US$400 million. Two years of the free cash of the assets sold would have paid the purchase price more than three times over, and the deferred consideration AngloGold took for the tail was capped at US$100 million. The company's reasoning in 2020 was that deep South African ounces had lost the contest for capital to shallower, cheaper ones elsewhere. At US$1,500 gold that was arguable. At a price near four thousand dollars the mine it sold is the single largest cash generator in the company that bought it, and the deepest gold on earth is paying a 38 per cent free-cash margin to somebody else. It is the worst South African mining transaction of the decade, and I will keep calling it that until someone shows me a worse one. The sleight of hand was in the telling. The sale was presented as an improvement in the quality of the portfolio; the deferred consideration was capped at US$100 million, so the seller kept none of the upside it now says nobody could have foreseen; and the mine's life was declared at seven years by a board that would not fund the R7.9 billion deepening its buyer has since paid for out of the mine's own cash, and that gives it twenty. Thirty pieces of silver is the old price for a home. This one went for two hundred million dollars, and the seller moved to Denver. The market has forgiven it entirely: it is rated at US$13,300 per annual ounce, twice Harmony. Gold Fields kept the one mine it built and sent the money abroad. AngloGold sold the deepest mine on earth for two hundred million dollars in cash, left the silicosis book behind, and is worth 41 billion in Denver. The market has forgiven both. This paper does not. EXHIBIT H · THE LEAVERS' LEDGER 11 What they sold, what it earns, and what they built instead A company that leaves a goldfield makes two bets at once: that the ground it sells is worth less than the price, and that the capital it takes away will earn more elsewhere. Both bets can be audited now, because the buyers publish accounts and the sellers publish impairments. I have done the audit for the three companies that took capital out of the Witwatersrand and for the one that kept putting it in, and I have tried to be fair to the leavers, because their shareholders have done well. Their shareholders would have done better, and the ground they left would be a liability on the state's books, not an asset on Harmony's and Sibanye's, had nobody stayed. Three boardrooms, three lawful decisions, one pattern. The majors looked at the country's deep gold, called it someone else's problem, and sold it cheap to the people who proved them wrong. AngloGold Ashanti first, because its ledger is the cleanest and the worst. It sold the Vaal River mines, Moab Khotsong and Great Noligwa, to Harmony for US$300 million in 2018. It sold the West Wits, Mponeng, TauTona, Savuka and Mine Waste Solutions, to Harmony in 2020 for US$200 million in cash and a tail the parties expected to be worth about US$100 million. The package is usually quoted at US$300 million. This ledger uses the cash, and the tail is being paid: Mponeng has produced above the 250,000-ounce threshold in both of the last two years. Harmony booked a bargain-purchase gain of about US$70 million on the day it took the assets, which is the accounting term for buying something for less than it is worth. What the assets have earned the buyer is in Harmony's own segment reporting, and Figure 5 sets it out: about US$1.5 billion of free cash from Mponeng alone in five and three-quarter years, on a cash price of US$200 million. Seven times the price, in cash. That is before Mine Waste Solutions, whose dumps now anchor a surface business that generated US$400 million of free cash in the latest year, and before Moab Khotsong, which has returned about US$500 million since 2021 on its US$300 million. The reserve statement tells the same story. Mponeng's life was seven years when AngloGold sold it. It is twenty. .FIGURE 5 Mponeng's adjusted free cash flow by Harmony financial year, US$m at R16.48, against the US$200m cash price paid on 1 October 2020. FY2021 is nine months. The deferred tail, capped near US$100m, is being paid and is excluded. Source: Harmony segment reporting; author's arithmetic. .What AngloGold built with the capital it took home is a good company, and the market says so at US$40.8 billion. But the ledger of what it built is not the ledger of a company that was right to leave. In 2013, the year the exit began in earnest, it impaired up to US$2.6 billion of assets after tax, most of them outside South Africa. It spent some US$500 million redeveloping Obuasi for a design of 350,000 to 400,000 ounces a year, and Obuasi has yet to produce 270,000 in a year, with a sill-pillar collapse in 2021 that killed a man, an illegal-mining incursion in January 2025 in which nine people died, and a two-week suspension after a fatality in April 2026. Put the two side by side. The mine it sold for US$200 million is producing 307,000 ounces a year at US$1,760 all-in with a twenty-year life, and threw off US$573 million in the last year at today's rate. The mine it is building in Nevada for US$3.6 billion will produce 500,000 ounces for nine years. The company that left South Africa because the deep ounces could not compete for capital is paying, per ounce of annual capacity, eleven times what its buyer paid in cash for Mponeng, to build a shallower mine with half the life. Its shareholders have been paid about US$2.4 billion of dividends in thirteen years, more than half of it in 2025. They have done well since 2015 because gold has. They would have done better owning Mponeng. Gold Fields' ledger is subtler, because it did not sell: it gave away. Nick Holland said at the time that the three mines still had "inherent quality" and resource potential; he was right about that, and he did not believe it enough to keep them, and the share fell by roughly sixty per cent in the year of the split. On 11 February 2013 it listed Sibanye, holding Kloof, Driefontein and Beatrix, at a market value of about US$1.07 billion, and distributed the shares to its own holders. What it kept built the one deep mine in the basin that works: about US$1.9 billion into South Deep between 2006 and 2018, US$760 million of it impaired and US$285 million of that reversed in 2025 when the mine produced 308,800 ounces and US$670 million of free cash. What it gave away has paid Sibanye's shareholders US$2.7 billion of dividends and buy-backs since 2013, and the gold mines themselves earned about US$2.2 billion of EBITDA between 2017 and 2025 on Sibanye's segment accounts. The capital Gold Fields kept went abroad, to Chile, Quebec and Western Australia, and twice into bids for other people's companies that failed, Yamana in 2022 and Northern Star in 2026. Its market value is US$29.6 billion against US$8.95 billion before the spin; add the Sibanye its holders were given and the holder who kept both is up about four times. The one asset the group would not part with, South Deep, is the one that proves the deep basin was never the problem. Sibanye's ledger is the one the market has already written. It took the three deep mines, paid their dividends, and spent their cash on platinum, palladium, lithium and nickel, Table 12 lists the purchases, and then wrote some US$4 billion of them off in three years, against a market value today of US$6.1 billion. Its market value went from about US$1.1 billion at listing to US$11.9 billion at the end of 2020, US$2.4 billion at the end of 2024, and US$6.1 billion now. The one transaction in that list that has unambiguously made money is the one that went the other way: about US$120 million of West Rand dumps and cash into DRDGold, worth about US$1.0 billion on 30 October. The company that was given the deep mines and told to run them for cash did so, and then bought everything except more gold; the one thing it bought that was gold, a half-share in a surface company, is its best investment by more than eight times. Harmony's ledger is the mirror of all three. Since 1995 it has bought what the houses and their successors wrote off, Table 12 lists that too, and its market value has gone from US$430 million at the end of 2015 to US$9.2 billion now: twenty-one times in eleven years, the best record of the six, built almost entirely from assets the sellers had decided were not worth keeping. It paid about US$490 million of dividends in the last year alone, more than in the previous twelve together. Above all four stands the parent, and the leavers learned the habit from it. At the closes of 22 October 2026, Anglo American put its gold into AngloGold in 1998, when AngloGold mined nearly half the country's gold. It sold out in three placings, at US$51, US$44 and US$32 a share, the last to John Paulson in the month the markets bottomed in 2009, for US$5.2 billion in all. The same shares were worth US$10.8 billion at that close. It sold a fifth of Gold Fields in 2004 for US$1.2 billion; that fifth is worth US$3.5 billion. It spent the gold money, US$10.5 billion, buying its own shares in 2006–08 at an average of US$49.50, borrowed to finish, suspended the dividend at US$12, and was cut to junk in 2016. The shares it bought are worth less today than it paid. Thirteen decisions scored to a number cost its shareholders US$32 billion net on prices, about US$35 billion with the dividends the sold shares have since paid their new owners. The four chief executives who signed them were paid about £150 million. The company that was worth more than BHP in 1999 is a quarter of BHP now, fourteenth in the world behind a royalty trust. Had its boards simply held what they were given it would be ninth. Had they cancelled the buybacks and bought the rest of AngloGold, the Amplats minorities and Gold Fields at the prices of the day it would be worth about US$156 billion: third in the world, ahead of Rio Tinto, with a gold division the size of Newmont's and the head office still in Johannesburg. Count every dividend, buyback and demerged share and the shareholders got about US$112 billion from the company as it was run. From the house the founder's rules would have left them, about US$187 billion. The three leavers in this section are Anglo's children, and they did what the parent showed them: sell the ground when it is out of favour, and call it focus. The parent had a rule against it, written by its founder in 1954 and still on its website: profits earned so as to make a real and permanent contribution to Southern Africa. The permanent contribution was sold in three placings, the last at the bottom of the market. Harmony, which Anglo never treated as a peer, now earns three-fifths of its profit from mines Anglo found, sank, equipped and sold. .FIGURE 6 What the seller received against what the asset now earns its buyer in a single year, US$ million: the Mponeng package (Harmony's FY2026 adjusted free cash flow at Mponeng plus Mine Waste Solutions' share of the surface business), Moab Khotsong and Great Noligwa (Harmony's FY2026 figure), and the Kloof-Driefontein-Beatrix complex at its 2013 listing value against Sibanye's 2025 South African gold EBITDA excluding DRDGold. Author's arithmetic from company accounts. .TABLE 12 The economic audit of leaving and staying. Sources: Anglo American's annual reports, results and announcements 1999–2026 for the parent's row, priced at the 22 October 2026 close; AngloGold, Gold Fields, Sibanye-Stillwater and Harmony announcements, results and impairment notes as listed in the references; market values from companiesmarketcap.com and the closes of 30 October 2026. Rand figures converted at R16.48 to the dollar for the June accounts and at R16.71 for the market values; historical dollar figures as reported. Author's compilation. .AngloGold sold Mponeng for US$200 million and is spending US$3.6 billion to build half of it in Nevada. Gold Fields gave away mines that have since paid 2.7 billion dollars to someone else. Sibanye wrote off four billion dollars of everything it bought that was not gold. Harmony bought what they sold and is worth twenty-one times what it was. EXHIBIT I · TEN YEARS 12 A decade of the market's verdict The chart below compresses ten years of market value into one measure, what a dollar put into each company at the end of 2015 is worth today, on a scale where a doubling is the same distance for a US$70 million company as for a US$41 billion one, and sets today's earnings multiple beside it. It is the single most useful picture of this industry I know, and it says three things no company's results do. The first is the starting point. At the end of 2015, with gold at US$1,060, the entire listed South African gold industry was worth almost nothing. Harmony, with 1.08 million ounces of production, was worth four hundred dollars per annual ounce, the price of a company the market expected to close; DRDGold was worth US$71 million. The decade since has been a re-rating of the whole industry from the price of scrap to the price of a business, and the question the chart answers is who re-rated most. Table 13 has the values. .TABLE 13 Year-end market values in US dollars, and the close of 30 October 2026. Pan African from sterling at year-end rates. Year-end values as recorded by companiesmarketcap.com; share counts changed over the period (Pan African's by a fifth in 2026), so the 2026 share-price moves in the text differ from the change in value. Sources in the references. .The second is the answer: the companies that stayed. DRDGold is worth twenty-eight times what it was, and paid a dividend every year on the way. Harmony twenty-one times, and nearly five times from the end of 2019, weeks before it agreed to buy Mponeng. Pan African seventeen. AngloGold fifteen and Gold Fields fourteen. Sibanye three times, having been at US$11.9 billion at the end of 2020 and US$2.4 billion at the end of 2024, and that is the group: in 2015 the group was the gold mines and nothing else, and count only its gold today, about US$1.1 billion for the deep mines on the sum-of-parts, and the gold Sibanye was given is worth about half what it was worth in 2015, in a gold price that has risen three and a half times. Gold itself rose a little over three and a half times. Every company that stayed in the Witwatersrand and mined gold outperformed the metal by a wide margin; the one that stayed and bought palladium underperformed it. The leavers did well; the stayers did better; and the stayers did better from a base the market had written down to nothing, which is the argument about depth and jurisdiction in one picture. Set them against the world and the claim holds. Of the world's ten largest gold producers by value, none has matched DRDGold's twenty-eight times or Harmony's twenty-one; Agnico Eagle, the best of the majors outside this paper, is at fifteen, Newmont at twelve, Barrick at seven. Only two mid-caps beat the two South Africans, Lundin Gold, which was a developer with no production in 2015, and Endeavour Mining, and only Northern Star beats Pan African. Figure 9 has the ranking. The third is the shape of the last two years. In 2025, with gold up 65 per cent, AngloGold and Gold Fields more than tripled, which is what a world-scale producer with operating leverage and a New York listing does in a bull market, and what index flows do to the two South African names that are in the world indices; the four who stayed roughly tripled too, from lower bases. Then, in 2026, with gold down about ten per cent, the four who stayed gave back between a sixth and two-fifths of their share prices: Sibanye 41 per cent, DRDGold 27, Harmony 27, Pan African 15. AngloGold is down 6. Gold Fields fell 25, on its own bid. Friday 30 October was the pattern in a day: gold fell 3.6 per cent to its lowest in a year, and all six fell between six and seven per cent, the leavers no less than the stayers. The market treats the South African names as the high-beta end of the trade: bought last in the rally, sold first in the pause. .FIGURE 7 Market value on 30 October 2026 as a multiple of the end-2015 value, log scale, with the gold price over the same period for comparison; and price to trailing earnings on 30 October 2026 (the shaded band is six to nine times). Pan African's sterling values at year-end rates. Sources in the references. .FIGURE 8 Share-price change from 1 January to 30 October 2026 in the listing currency, and gold in dollars. Sources in the references. .FIGURE 9 Market value in October 2026 as a multiple of the end-2015 value, US dollars, log scale: the six South African companies against the world's largest gold producers and the mid-caps that beat them. Year-end values from companiesmarketcap.com; Northern Star's 2015 value is the author's estimate. Author's arithmetic. .The right-hand panel tells the other half: the multiple of trailing earnings the market pays today, read against the year-end multiples since 2020. The first pattern is that the South African names have rarely been rated above ten times earnings at a year-end, and never for long. Harmony at 5.6 times today, Sibanye at 6.1, DRDGold at 7.7 and Pan African at 7.9 are rated where they have always been rated, between five and ten. AngloGold at 10.7 and Gold Fields at 6.9 have moved in a wider band. The second is that the market's multiple has fallen as earnings have risen. Four of the six are on a lower multiple today than at the end of 2025, and Pan African is within a point of where it was. The reason is that the market does not believe gold near four thousand dollars is the base from which to capitalise. That disbelief is the opportunity and the risk in one number. If gold holds, the six are on six to eleven times a record year and the four who stayed are the cheapest of them. If it does not, the multiple was right and the earnings were the anomaly. From US$71 million to US$2.0 billion at DRDGold, from US$430 million to US$9.2 billion at Harmony. The stayers beat the metal and the leavers; the market still rates them at six to eight times a record year.EXHIBIT J · THE LEDGER 13 Six companies on one page Set the six on one page at the same gold price and the same date, and the comparison the market makes every day becomes visible. The table below carries the latest reported period for each, the balance sheet at the same date, and the market's rating on 30 October 2026. Three things stand out. The four who stayed are the higher-cost producers: about US$2,360 an ounce all-in, weighted by output, against about US$1,950 for the two who left, and Harmony and Sibanye carry the two highest costs and the two thinnest margins of the six. Pan African, at US$1,867, is the exception and the lowest cost in the book; Sibanye is the cheapest multiple and Harmony the largest free cash flow in rand among the four. The two who left have the lower cost, the widest margins, the largest reserves, the strongest balance sheets and the highest ratings. And the rating does not follow the cost, the margin or the balance sheet: it follows depth, jurisdiction and liquidity, in that order. .TABLE 14 The six on one page. Sources: company results as dated in the references; closes of 30 October 2026; gold about US$3,875, rand 16.71. Enterprise value per annual ounce uses the latest annual or annualised production; Sibanye's deep-gold figure is the author's sum-of-parts described in Section VI. Sibanye's market value, P/E, yield and EV/EBITDA are the whole group, 60 per cent platinum; the pro-rata gold value is the group's market value in proportion to the gold division's share of adjusted EBITDA (28 per cent, DRDGold included); EV per reserve ounce for Sibanye uses gold reserves only and overstates the gold rating because the PGM business carries most of the enterprise value. All ratings are the author's arithmetic. Rand counterparts are the companies' own where they report in rand (Harmony, Sibanye-Stillwater and DRDGold cost; DRDGold price) and otherwise at R16.48 for the accounts and R16.71 for the market, 32.1507 oz/kg. .FIGURE 10 All-in sustaining cost against the price received, US$ an ounce, latest reported period: Pan African, DRDGold and Harmony for the year to June 2026, Gold Fields, AngloGold Ashanti and Sibanye-Stillwater's South African gold for the six months to June 2026. Harmony's price is net of US$571m of realised hedge losses. Sources in the references. .On one page: the four who stayed mine at a higher cost and a thinner margin than the two who left, about US$2,360 an ounce against US$1,950. The rating follows depth and distance, and it discounts the four by more than the cost gap explains. 14 Ranked against one another One page of figures invites one page of ranks, and the ranks are more revealing than the figures, because they separate what a company is from what the market charges for it. The table below ranks the six, one to six, on seven measures of quality and five of price, and sums each. The quality measures are the cost of an ounce, the margin on it, the balance sheet, the years of reserve at the current rate of production, the growth in output in the latest year, whether the company has sold its price away in a hedge, and what it has done for its owners over ten years. The price measures are the earnings multiple, the enterprise multiple of EBITDA, the enterprise value per annual ounce and per reserve ounce, and the dividend yield. A low total is good on both sides: the best company and the cheapest share. .TABLE 15 The six ranked on seven measures of quality and five of price, with the figure in brackets. Reserve years are reserves divided by the latest annual production; ten-year value is market value at 30 October 2026 over market value at the end of 2015; Harmony and Gold Fields tie on reserve years and share the rank; Sibanye's EV per reserve ounce uses gold reserves against an enterprise value that is mostly platinum, and overstates the rating. Author's compilation and arithmetic. .FIGURE 11 Quality against price: the quality rank total from the table above against price to trailing earnings, 30 October 2026. Lower is better on the quality axis; the shaded band is six to nine times. Author's ranks; multiples at the closes of 30 October 2026. .Read the quality total against the earnings multiple and the market's logic is plain, and it is not a logic about quality. Pan African is the best company of the six on these measures, by a distance, with a quality score of 12 against DRDGold's 17 and Gold Fields' 19, and it is rated at 7.9 times, within a point of DRDGold and Gold Fields and nothing like the premium its quality would command in Toronto or Perth. DRDGold is the second-best company, at 7.7 times. AngloGold is the fourth-best company and the dearest share by a distance, at 10.7 times, the New York premium in one row; Gold Fields, the third-best, has fallen to 6.9 times, which is what a failed bid does to that premium. Harmony is the fifth-best company and, at 5.6 times, the cheapest share, and its quality score is dragged down by three rows the board controls, the hedge, the cost line and the output guidance, not by the ground. And Sibanye ranks last on these measures by a distance, at 6.1 times earnings and the cheapest in the book on every other measure of price: the market has priced the quality correctly and offers no discount beyond it, except the one that applies to the gold inside the group, which is the only mispricing in the table that a single board decision would remove. Set quality against the multiple. Pan African and DRDGold are the two where quality exceeds what the price charges for it. Harmony is cheap for a reason the board can fix. Gold Fields and AngloGold are priced for what they are. Sibanye is cheap for reasons that are real. The ranking does not change if the weights are changed within reason. It changes only if the gold price falls. Then the cost and margin rows decide everything, and the order becomes Gold Fields, Pan African, AngloGold, DRDGold, Harmony, Sibanye, which, Pan African aside, is also the order of distance from the Witwatersrand. The best company in the group is rated as the average one. The two that carry the deep mines are the two cheapest. Only the one that left for New York carries a premium. Quality is not what the market is pricing; depth and distance are. EXHIBIT K · THE RATING 15 What the market pays, and why The chart below is what the market thinks. The market pays, per ounce of annual production, about US$2,700 for Sibanye's deep gold on a sum-of-parts, US$6,400 for Harmony, US$11,600 each for DRDGold and Pan African, US$12,000 for Gold Fields, and US$13,300 for AngloGold Ashanti. The ounces are the same metal, sold into the same market at the same price. What differs is where they come from and what the market believes about their future. .FIGURE 12 Enterprise value per ounce of annual gold production, US$ thousand, at the closes of 30 October 2026. Sibanye's deep gold is the author's sum-of-parts: the gold division's EBITDA excluding DRDGold, valued at the group's own multiple of 1.7 times. Harmony on FY2026 production; DRDGold and Pan African on June 2026 years; Gold Fields and AngloGold on 2026 guidance midpoints. Author's arithmetic. .Three things explain it. The first is depth. The two South African companies rated with the world's seniors, DRDGold and Pan African, are the two whose ounces are shallow: dumps at a fifth of a gram with nobody underground, and a group 45 per cent of whose ounces come from the same source. The market pays about US$11,600 to US$12,000 for an annual ounce that cannot be lost to a rockburst, a seismic event or a strike, and it pays the same for a Pan African ounce as for a Gold Fields ounce from Gruyere. The deep-level ounce is rated at half that at Harmony and a quarter at Sibanye, and the discount is on four kilometres of rock and not on South Africa, on the fatality record, on the grade decline from 6.27 to 5.83 grams in a year, on reserves that the owner itself declares in single digits of years. The deep-level discount is a discount on the owner's own reserve statement, and the owner can change it. I have argued this before, and the only board that has acted on it is Harmony's: Harmony did, with ten million ounces of conversion at about US$180 an ounce in a year, and its rating has not yet followed because the market is waiting for the copper. The second is jurisdiction, and the evidence is that it matters less than the companies say. Gold Fields and AngloGold are rated at US$12,000 and US$13,300 an annual ounce with 88 and 100 per cent of their ounces outside South Africa. Pan African is rated at US$11,600 with 88 per cent of its ounces inside it. The gap between a South African surface ounce and an Australian open-pit ounce is a few hundred dollars per annual ounce, within the noise of a week's trading. Measured at the level of the whole market the jurisdiction discount is real: a South African ounce fetches about 42 cents on the global dollar, price to net asset value some 30 per cent below Australia, enterprise multiples 40 to 80 per cent below. The table in this paper shows the discount has not gone. It shows that two South African companies have escaped it without leaving, by being shallow. What South Africa costs is the deep-level premium on the AngloGold multiple over Harmony's, about US$6,900 per annual ounce, and most of that is depth and cost, not the flag. A company that argues it must leave the country to be rated with the seniors has Pan African's rating to answer. The third is liquidity, and it needs stating carefully, because it is not about where the shares are listed. All six are listed abroad. Harmony and DRDGold have traded on the New York Stock Exchange for decades; Pan African is on London's AIM and, since June 2026, Sydney; Sibanye has its New York receipt. A foreign fund that wants any of the four can buy it in dollars tomorrow, and the JSE, whose listings have fallen from more than 800 in the 1990s to about 280, is not what stands in the way. What stands in the way is size and index weight. AngloGold's primary listing in New York, its dollar reporting and its US$41 billion market value put it in the world gold indices and in the hands of the capital that tracks them; Gold Fields, at US$30 billion, is in the same class. A 2-billion-dollar company, however listed, is not, and the index flows that tripled the two leavers in 2025 do not reach it. The premium that class pays is real, about two points of P/E on this evidence, and it is the strongest argument the leavers have. It is also an argument about scale, not address: the four who stayed have already exported the listing without exporting the mines, and the premium has not followed, because none of them is yet large enough to be carried into the indices by it. The Public Investment Corporation, meanwhile, holds about 14 billion dollars of these six companies' shares, 15 to 20 per cent of each, and not one ounce of bullion: overweight the riskiest slice of gold, flat the safest, and it has never asked a board what it is doing about the discount. Then there is the one company none of the three explains. Sibanye's gold is rated at a quarter of DRDGold's per annual ounce, and depth, jurisdiction and liquidity do not account for the whole of it. Harmony is as deep and as South African and is rated at more than twice as much. What accounts for it is the group, and the group's borrowing. The market cannot buy Sibanye's gold without buying its palladium, its Montana strike, its Finnish lithium, its French nickel and R32 billion of gross debt, and it will not pay a gold multiple for a basket of which gold is 28 per cent. The cure is the one Sibanye applied to its own dumps in 2017: a separately priced vehicle. A South African gold company holding Driefontein, Kloof, Beatrix, Cooke, Burnstone and the 50.1 per cent of DRDGold, listed on its own, with Burnstone's non-recourse debt and none of the group's, would be rated as a gold company with a 25-year new mine, a surface business worth 2 billion dollars and three deep mines run for cash. DRDGold itself, more than eight times in eight years, says the rating would be a multiple of the US$2,700 an annual ounce the gold earns inside the group today. It is the one transaction on the Witwatersrand that would make money for every party to it. The board has done it once. I have yet to hear a reason from Sibanye why it will not do it again. The discount has one more cost, and it changes what a board can do about it. The rating ladder is also a ladder of currencies. A company at the bottom cannot pay for a rival in its own shares, because a target's board reads the ladder the way the market does. A company near the top can buy almost anyone. Gold Fields, at twelve thousand dollars an annual ounce, was told by Northern Star's board in September 2026 that its paper carried a meaningfully higher jurisdictional risk than the target's own. One of its own shareholders, Redwheel, told it in October, for the second time in four years, not to pay that paper away at all. The Public Investment Corporation, holding fifteen per cent, said nothing either time. Harmony, at six thousand four hundred, could not have made the offer. That is what the discount costs a board beyond the share price, and it is why the four who stayed were built in cash. A currency the market will not take cannot be spent. The companies that could not spend theirs bought at the bottom, from sellers who had given up, and were paid for it in the ten-year chart above. The argument holds in Perth and Denver as it does in Sandton. All of this is three sentences, and the section above is those sentences with numbers. The leaver wins an arbitrage, not an orebody: the reward the market pays for escaping a mispriced jurisdiction, not for making gold pay. The market pays a premium to abandon the reef and a penalty to mine it. Close the discount and the calculus inverts: the builder who stays, not the leaver who goes, is the winner. Depth, jurisdiction and liquidity, in that order. Pan African proves the second matters least; Harmony and DRDGold, on the New York board for decades, prove the third is size and index weight, not address; Sibanye proves what a conglomerate costs. THE BEAR, MET 16 What the sceptic rightly takes Most of what the sceptic takes, this paper has already given him in its second finding: the leavers run the lower-cost, wider-margin, better-capitalised businesses, and the four who stayed are the higher-beta exposure to the metal. Every number in this paper is a function of a gold price near four thousand dollars: a one-year low on 30 October, and still 40 per cent above where it stood two years ago. Harmony's US$1,616 margin after hedges and Sibanye's US$1,492 would halve at US$3,400 and vanish at US$2,700; AngloGold and Gold Fields would still be earning about fourteen hundred dollars an ounce at the first and seven hundred at the second, and rand costs are rising 13 to 15 per cent a year at all four. Conceded in full. What the bear should not take is DRDGold's and Pan African's margins, US$2,232 and US$2,368. They are within four hundred dollars of AngloGold's and six hundred of Gold Fields', and they are earned on the Witwatersrand. Four objections remain that the findings do not already answer. The first is the dumps run out. DRDGold's reserve is twenty years and Pan African's three tailings plants have eight, seventeen and six years of their own dumps before Soweto's fifteen. Conceded as a date. A twenty-year reserve at a surface miner is longer than Driefontein's ten and Beatrix's four, and the 1,600 tonnes of gold the Witwatersrand's dumps still hold, on the published estimate, is nearly two centuries of the two companies' combined surface output. The constraint is deposition and water, not ore, and both companies are building the facilities now. The second is control. A DRDGold minority sits under Sibanye, and a Harmony shareholder sits under a board that has said the cash is going to copper. Conceded. Only one of the four is free of both, and the market has already noticed, which is why it is rated at eight times earnings with three projects at nothing and not at six. The third is that the discount is a buyer's absence and not a buyer's view: that nobody bids for deep South African gold because nobody can, and the rating reflects a closed market and not a judgement. Partly conceded; Barrick's African mines, for which no buyer is both willing and allowed, are the same case. But the Witwatersrand has had bidders, and they were the four in this paper: Harmony paid cash for Mponeng, Sibanye took Kloof for nothing, Pan African paid fifty million rand for Mintails, and each has been repaid many times. The market that will not bid is the foreign one. The domestic one has bid at every bottom, and won. The last is the shadow ounce. Mirror statistics are not production: a bar can be counted twice on its way from Dubai to Zurich, foreign doré that came into a small South African refiner leaves as South African, and the Swiss "recycled" line fell after 2019 because the London association began publishing its refiners' country-of-origin returns, not because the gold stopped. Partly conceded, which is why I give a range and not a number. But the Emirati line rose as the Swiss line fell, the state's own ministers have put the loss at sixty to seventy billion rand a year, and the floor of the range, seven tonnes, is still a producer the size of Pan African's South African output that pays no royalty. Reject every mirror and the Minister's figure remains, and it is larger than mine. The bear takes the beta and the cost curve, and the second finding gives him both. He takes the finite dumps, the control and the double-counted bar. He leaves the margins at the surface, the reserve that is a decision, and the US$200 million mine that earns nearly three times that a year. THE FUTURE 17 The hole the leavers left Everything in this paper so far is about ounces already found. The question that decides whether the Witwatersrand has a third act is who is looking for the next ones, and the answer is that almost nobody is, and that the people who used to are the two companies that left. Exploration is the one line in a mining company's accounts that only a strong balance sheet can carry, because it is money spent on ground that may hold nothing, for a return a decade away. For a century that balance sheet, on this basin, belonged to the houses and then to their gold children. Anglo American's geologists found the Free State and the Far West Rand; AngloGold sank Mponeng and Moab and ran a reef-boring programme, US$30 to 40 million of it, that was let lapse when the exit began; Gold Fields found and then built South Deep. When the two left, they took that capacity with them, and no one has replaced it. The numbers are stark. Statistics South Africa puts the country's entire exploration spend, every mineral, every company, at R738 million in 2025 at 2015 prices, about US$44 million, down for a seventh consecutive year and more than 85 per cent in three decades; the Minerals Council's president put the fall at 80 per cent in two. South Africa attracts less than one per cent of the world's exploration budgets, in a year in which gold was the metal the world explored for most. Against that, the two who left spent, in the same year, about US$276 million on exploration at AngloGold Ashanti, US$64 million of it on greenfields ground in Australia, Brazil, Tanzania, Egypt and the United States, and about US$298 million at Gold Fields, tripled on the year and driven by Canada. Two companies born on the Witwatersrand spent thirteen times the entire country's exploration budget looking for gold, and not a dollar of it in South Africa. Gold Fields' one commitment here, 100,000 to 200,000 metres of drilling at South Deep over five years, is brownfields work on a mine it already owns. Table 16 sets the six side by side. .TABLE 16 Who is looking for the next ounces. Company figures from the 2025 exploration and results reports as listed in the references; the national figure from Statistics South Africa as reported 13 March 2026; cadastre dates from the Department's director-general, 9 June 2026. Author's compilation. .That leaves the four. Harmony's exploration is conversion: ten million ounces moved from resource to reserve in a year at about R3,000 an ounce, the Zaaiplaats decline at Moab, the Mponeng deepening, two reclamation studies over six million ounces of surface resource. Sibanye's is Burnstone, a deposit found by someone else and bought out of a liquidation. Pan African's is the only new ground: Royal Sheba, a new mine on the Barberton fault, and Poplar, 6.57 million ounces inside the Evander mining right that it is drilling toward a decision in 2028. DRDGold's is none, by design; its orebody is a map of other people's dumps. All of it is near-mine. None of it is the regional, basin-scale work that found the Free State, and none of the four has the balance sheet for that work the way AngloGold has: Pan African's entire market value is a year and a half of AngloGold's free cash. The obligation to find the Witwatersrand's next ounces has fallen on the four companies least able to carry it, because the two companies most able chose to carry it in Nevada and Quebec instead. That is the hole the leavers left, and it is the part of their ledger that no asset sale measures. There is a second line no asset sale measures, and it runs the other way. The ground the leavers sold came with its water. The Mponeng package carried a rehabilitation provision of R1.86 billion against R1.27 billion of restricted cash, R230 million of it for decant from the tailings; the three mines Gold Fields gave away now carry R4.0 billion of closure liability on Sibanye's books against R3.2 billion of cash, and at Beatrix the cash and guarantees together are R289 million short of the gross figure. Behind both sits the public line: the state's three treatment plants on the Witwatersrand cost R475 million to run in the year to March 2026, and in May 2025 the Central Basin decanted anyway. None of that is on the leavers' balance sheets. Not all of it has left them. AngloGold was named with Harmony on the 2005 Orkney water directive, and the Supreme Court of Appeal held in 2013 that such a directive does not lapse when the land is sold; since June 2023 the law says the same of a closure certificate. The leavers' accounts carry no South African water line, and Gold Fields classifies South Deep's own post-closure water as a contingent liability it cannot yet estimate. The four who stayed carry the provision, and the risk that goes with it: Harmony's Form 20-F of 30 October 2026 discloses a third-party appeal against Mponeng's own water-use licence, provisionally suspended while the parties talk, and R539 million of bank guarantees standing behind its rehabilitation obligations. The two who left carry the tail, unpriced, and the market has not asked for it. It is not a hole the ground made. The basin's remaining endowment, on the published estimate used throughout this series, is about 48,100 tonnes, more than every other gold province on earth combined, and the first new underground gold mines in fifteen years, Qala Shallows in March 2026 and Burnstone in September, were both found decades ago and both opened by companies the majors would not have noticed. What stands between the endowment and the drill is the state. The system that grants a prospecting right, Samrad, has been dysfunctional for more than a decade. Its replacement, the cadastre, was put out to tender in August 2021, promised for March 2023, then for June 2025, then piloted in the Western Cape, and on 9 June 2026 the director-general gave 31 March 2027 as the national date; Bernard Swanepoel, asking him for it from the floor, observed that it was the fifth such promise. The department receives about 2,800 applications a year and processes about 2,500, so the backlog grows by three hundred a year by its own arithmetic. The department's own strategy, published in 2025, targets five per cent of the world's exploration spend from under one; the instrument it has produced is a R100 million fund for juniors, a seventh of what AngloGold spent on greenfields alone. The draft mineral-resources bill of May 2025 and the industrial strategy of June 2026 would add beneficiation conditions to mining rights. South Africa has no equivalent of Canada's flow-through share, the tax instrument that funds most of the world's junior exploration, and has had none proposed. A country that cannot say what its gold output is, Section IV, also cannot say what its prospective ground is, because it has not had a working register of who holds it for ten years. The consequence belongs in the valuation. A reserve of twenty years at Mponeng, forty at DRDGold, forty-eight at Pan African, is what the four who stayed declare today. What they do not declare, and cannot, is what comes after, because the ground beyond their fences has not been drilled in a generation and the right to drill it sits in a queue. The market's discount on the four is in part a discount on that silence. It is one the leavers do not bear, because they bought their next decade abroad, and one the four cannot close alone, because the capital it needs is the capital that left. The programme that follows asks each board for an exploration line in its reserve statement; it asks the state for the register that would make the line worth publishing. The two who left spent US$574 million on exploration in 2025, thirteen times the whole country's budget, and none of it here. The four who stayed drill their own fences. The ground beyond is the richest on earth, and the right to drill it is in a queue the state has been promising to clear since 2023. THE PROGRAMME 18 What each board should do The road map for the boards is short. Redraw the reserve lines at a price within sight of the one you sell at. Price the resource and not the reserve. Apply to Kloof the test you applied to Burnstone before you flood it. Harmony should report its surface business as a segment with its own reserve statement, convert the six million ounces in its two reclamation studies, and give the market a price for a business it now rates at the group's deep-level multiple. DRDGold is the evidence: a listed surface company with Harmony's 221,199 ounces would be rated at nearly twice what Harmony's own paper fetches, and a minority placing at that rating would fund the copper without a hedge book that cost 571 million dollars last year. Sibanye should list its gold: a South African gold company holding the three deep mines, Burnstone, Cooke and the 50.1 per cent of DRDGold, with Sibanye keeping control. Its own 2017 precedent proves the transaction, and it is the only way the Burnstone decision is paid for in the share price instead of buried under palladium. DRDGold should build the regional facility and Withok on time, because its rating is a rating on deposition capacity, and should say plainly what its parent intends. Pan African should cancel the buy-back and put the thirty million dollars into Soweto's first year. All four should put an exploration line in the reserve statement: what was spent beyond the fence, where, and what it found, so that the market can price the next decade and not only the declared one. Gold Fields should put a number on South Deep's step change into the 2030s before it bids twenty-five billion dollars for someone else's mines. AngloGold Ashanti owns 42.41 per cent of Rand Refinery and nothing else in the ground it came from. It should answer the question its 2020 sale leaves open: what the deepest mine on earth was worth at the price it is now paid, whether a board that sold it for US$200 million should be spending US$3.6 billion in Nevada, and what, under the rule its parent wrote in 1954, it still owes the country it was made in. For the funds that own these companies: vote at Sibanye, Harmony and Gold Fields as owners of an endowment, not as holders of a ticker, and ask the PIC why it owns a fifth of a company whose richest reef is being flooded and has never put the question. For the state, one item, because it affects every company in this paper and everything that follows it. Count the gold. The revenue service classifies most of the country's gold exports as "unclassified". London's refiners recorded, in a single year, South African "scrap" equal to sixty per cent of its declared output. The only convictions in the illegal trade are of men at the bottom of it. A country in that position cannot say what its gold industry is worth or who owns it. Reconcile the official output against the mirror data every year and publish the gap. Deliver the cadastre on 31 March 2027, the fifth date promised, and clear the queue behind it; and give exploration a tax instrument, as Canada's flow-through share has for forty years, because a R100 million fund will not find a basin. License the artisanal tier so that the shadow producer becomes a declared one, as the draft mineral-resources bill of May 2025 proposes and the criminal bill of August 2026 does not. Aim the Precious Metals Act, the organised-crime law and the trafficking law at the dealers, the refiners and the export-permit holders who own the shadow ounce, not a thirty-year sentence at the man underground. The listed four pay for the policing of a competitor the state does not count. That is a cost of capital, and it is in every multiple in this paper. THE VERDICT 19 The reef pays whoever stays Four companies kept this industry alive, and none of them is the one the market pays most for. Harmony took the mines the majors wrote off and turned the deepest of them into a mine that throws off 570 million dollars a year on a two-hundred-million-dollar cheque; had it not, Mponeng, Moab Khotsong, Kusasalethu and the Free State shafts would be on the state's list of 6,100 derelict mines beside Blyvooruitzicht and Buffelsfontein, and 47,000 jobs with them. Sibanye took the three mines Gold Fields stamped end of life, kept them open and pumped for thirteen years, paid their shareholders 2.7 billion dollars, found the right strategy for the ground, surface first and a new shallow mine, and has yet to be paid for it because the strategy sits inside a platinum company that has written off four billion dollars of everything else it bought. DRDGold took the dumps nobody wanted and turned them into a company with nineteen consecutive dividends and a 58 per cent margin, worth twenty-eight times what it was ten years ago, and takes the dumps away from the townships as it goes. Pan African took the oldest goldfield in the country, built three surface plants to carry it, and became the cheapest producer in the basin with no debt and three projects in hand. Those four companies mine three-quarters of the country's declared gold and are worth 21 billion dollars, 16 billion of it gold. The two that left mine one mine in it and are worth 70 billion dollars. The five findings, restated with their numbers. The leavers were wrong to sell what they sold at the price they sold it, and wrong to leave: AngloGold took US$200 million for a mine that has returned US$1.5 billion in under six years and is spending eighteen times that sum to build half of it in Nevada; Gold Fields gave away mines that have paid US$2.7 billion to someone else's shareholders, and kept the one deep mine that proves the basin was never the problem; both broke the rule their parent wrote in 1954, that profits were to be earned so as to make a real and permanent contribution to Southern Africa. The leavers are the better businesses today: US$1,950 an ounce against US$2,360, the widest margins, the largest reserves, the strongest balance sheets, and fourteen and fifteen times their value in ten years. Two of the four who stayed match them on quality and are rated below them: Pan African first on seven measures and lowest in cost, DRDGold second, both rated with Gold Fields per ounce and a point above it on earnings, from bases the market had written to nothing; Harmony and Sibanye are the two highest-cost producers in the book and are rated as such, for a hedge book, a copper programme and a platinum group that are each the board's to undo; they are also the two that kept the deep mines open, and the rating is the price of carrying them. And the discount is depth, group structure and size, not the flag: Harmony can close its by declaring and listing its surface business, Sibanye by listing its gold, Pan African by doing nothing differently for another five years. And the leavers took the exploration balance sheet with them: US$574 million spent looking for gold in 2025, none of it here, against a national budget of US$44 million and a prospecting register the state has been promising since 2023. The deep Witwatersrand was not uneconomic. It was unfashionable, and the companies that could not tell the difference sold it for thirty pieces of silver to the ones that could. The four who stayed are rated today at six to eight times a record year because the market does not believe the price. If the market is right, the cost rows decide the order and the leavers win. If it is wrong, the four who stayed are the cheapest gold in the world, and Pan African is the best of them. The ninety tonnes is a floor and not a count. South Africa produces a third to a half more gold than it declares; syndicates, dealers and export-permit holders the state has never convicted own the increment; and the listed four pay for its policing out of their shareholders' cash. Any valuation of the South African gold industry that starts from ninety tonnes starts from the wrong number. The reef pays whoever stays on it, and the test every board in this paper should be marked against is one sentence: we did not run out of gold, we ran out of conviction and courage. In the year to June 2026 it paid the four who stayed about R47 billion of EBITDA and free cash between them, at a price the two who left said it would never see, and it paid a fifth producer, the one with no listing, no royalty and no reserve statement, something between a third and a half as much again. I hold positions in Pan African Resources and DRDGold through one of my funds. This is not advice. .REFERENCES COMPANY REPORTS, RESULTS AND ANNOUNCEMENTS Harmony Gold Mining Company. Results for the financial year ended 30 June 2026, and the FY2026 results booklet (operational and financial tables by mine). Johannesburg, 27 August 2026. Harmony Gold Mining Company. Mineral Resources and Mineral Reserves Report for the year ended 30 June 2025 (HAR-RR25). Johannesburg, October 2025. Harmony Gold Mining Company. Annual Report on Form 20-F for the financial year ended 30 June 2024 (deferred consideration on the Mponeng package; bargain-purchase gain). Johannesburg, 2024. Harmony Gold Mining Company. Annual Report on Form 20-F for the financial year ended 30 June 2022 (final acquisition accounting for the Mponeng package: rehabilitation provision, restricted cash and investments, and the groundwater liability). Johannesburg, 2022. Harmony Gold Mining Company. Annual Report on Form 20-F for the financial year ended 30 June 2025 (Moab Khotsong ore gap 2027–2031). Johannesburg, 2025. Harmony Gold Mining Company. Annual Report on Form 20-F for the financial year ended 30 June 2026, filed 30 October 2026, with the Integrated Annual Report 2026 as Exhibit 15.1 (security expenditure by region FY2026 and FY2025; illegal-mining arrests at operating and non-operating shafts; Operation Vala Umgodi; shafts sealed since 2008; the US$500 million convertible bonds of 29 September 2026). Johannesburg and Washington, 2026. Harmony Gold Mining Company. Presentation and transcript, Mining Forum Americas, Colorado Springs, 28 September 2026 (Eva Copper capital range; Australasia share of output). Harmony Gold Mining Company. Acquisition of Mponeng, Mine Waste Solutions and related assets from AngloGold Ashanti: announcement of 12 February 2020; completion effective 1 October 2020. Harmony Gold Mining Company. Acquisition of Moab Khotsong and Great Noligwa from AngloGold Ashanti: announcement of 19 October 2017; completed 28 February 2018. Harmony Gold Mining Company. Company history (corporate website), accessed October 2026. Sibanye-Stillwater. Operating and financial results for the six months ended 30 June 2026, including the Burnstone project update. Johannesburg, 1 September 2026. Sibanye-Stillwater. Mineral Resources and Mineral Reserves declaration as at 31 December 2025. Johannesburg, 17 February 2026. Sibanye-Stillwater. Group Impact Supplement 2025 (closure liability, cash funding and guarantees at Kloof, Driefontein and Beatrix). Johannesburg, 2026. Sibanye-Stillwater. Annual Report on Form 20-F for the year ended 31 December 2022 (dividends and buy-backs 2013–2022). Johannesburg, 2023. Sibanye-Stillwater. Annual Report on Form 20-F for the year ended 31 December 2013 (market value at listing and at year-end). Johannesburg, 2014. Sibanye-Stillwater. Reports on Form 6-K carrying the impairment notes for 2023, 2024 and 2025 (US PGM, Sandouville, Keliber, Kloof). Sibanye-Stillwater. Illegal mining and security fact sheet, April 2026. Sibanye-Stillwater. Capital allocation framework (investor presentation). Sibanye Gold. Pre-listing statement and listing on the JSE, 11 February 2013; unbundling from Gold Fields effective 18 February 2013. DRDGold Limited. Results for the year ended 30 June 2026, and the results presentation. Johannesburg, 19 August 2026. DRDGold Limited. Mineral Resources and Mineral Reserves statement as at 30 June 2026 (in the 2026 Integrated Annual Report). THE FOUR WHO STAYED · THE MONEY METAL · THE HINGE PAPER Page 50 DRDGold Limited. Annual Reports on Form 20-F (history: Crown Mines 1998, ErgoGold 2008, Ergo Mining 2010, Blyvooruitzicht disposal FY2012, ERPM held for sale 2013). DRDGold Limited. Announcement of the transaction with Sibanye-Stillwater for the West Rand Tailings Retreatment Project assets, 22 November 2017, and completion, 31 July 2018; exercise of the option to 50.1 per cent, January 2020; transfer of the Kloof 2 dump, December 2025. Pan African Resources plc. Summarised Audited Results for the year ended 30 June 2026. Johannesburg and London, 16 September 2026. Pan African Resources plc. Integrated Annual Report 2026 and Mineral Resources and Mineral Reserves Report 2026. Pan African Resources plc. Soweto Cluster definitive feasibility study: SENS announcement, 11 September 2026. Pan African Resources plc. Results presentation and call transcript for the year ended 30 June 2026 (Soweto decision timing; Royal Sheba; Poplar; Tennant Creek ramp-up), 16 September 2026. Pan African Resources plc. Announcement of the share buy-back programme through Peel Hunt, October 2026. Gold Fields Limited. Interim results for the six months ended 30 June 2026, and Report on Form 6-K. Johannesburg, 25 August 2026. Gold Fields Limited. Mineral Resources and Mineral Reserves Supplement to the Integrated Annual Report 2025 (reserves at 31 December 2025; South Deep). Johannesburg, February 2026. Gold Fields Limited. Annual Report on Form 20-F for the year ended 31 December 2025 (South Deep life of mine). Gold Fields Limited. Results for the year ended 31 December 2025 (South Deep output), 19 February 2026. Gold Fields Limited. Announcement of the rejection of the unsolicited, non-binding proposal from Northern Star Resources, 28 September 2026. Gold Fields Limited. Completion of the acquisition of Gold Road Resources, 26 September 2025. Gold Fields Limited. Impairment notes for 2017 and the first half of 2018 (South Deep), and the 2025 reversal. Gold Fields Limited. Annual Report on Form 20-F for the year ended 31 December 2025, note 39, contingent liabilities (South Deep post-closure water; Cooke 4 and the Cooke 1–3 closure application), and the Sustainability Report 2025 closure-cost table (South Deep US$54 million). Johannesburg, 2026. AngloGold Ashanti plc. Results for the second quarter and six months ended 30 June 2026. Denver, 31 July 2026. AngloGold Ashanti plc. Mineral Resource and Mineral Reserve Report as at 31 December 2025, and the fourth-quarter 2025 results release (Arthur project pre-feasibility). AngloGold Ashanti plc. Annual Report on Form 20-F for the year ended 31 December 2025 (corporate restructuring; Johannesburg office). AngloGold Ashanti Limited. Report on Form 6-K, 25 July 2014 (shareholding in Rand Refinery). AngloGold Ashanti Limited. Sale of Mponeng, Mine Waste Solutions and related assets to Harmony: announcement of 12 February 2020 (Form 6-K). AngloGold Ashanti Limited. Sale of the Vaal River assets to Harmony: announcements of 19 October 2017 and 28 February 2018. AngloGold Ashanti Limited. Annual reports and impairment notes for 2013, 2022 and 2023. Rand Refinery (Pty) Ltd. Shareholding and throughput as reported by the LBMA and BullionStar (below). Anglo American plc. Annual reports 1999 to 2025; announcements of the AngloGold Ashanti placings of 2006, 2007 and March 2009, the Gold Fields sale of 2004, the 2006–08 share buy-back, the Rustenburg sale of 2016, the Thungela demerger of 2021 and the Valterra placing of 4 September 2025; interim results 2026. London and Johannesburg. Anglo American plc. 100 years of Anglo American (the 1954 statement of aims by Sir Ernest Oppenheimer, as the company now renders it), corporate website, accessed October 2026; Oppenheimer, N. See what is right and be brave: farewell address to Anglo American, Brenthurst Foundation Discussion Paper 2011/11 (the 1954 wording), Johannesburg, November 2011. Sibanye-Stillwater. Conclusion of the Section 189A consultation on the restructuring of the South African gold operations and regional services (South African region headcount about 81,500 at end-2022 to about 70,000; forced retrenchments 8 per cent of those affected since January 2023). Johannesburg, July 2024. DRDGold Limited. Operating update for the quarter ended 30 September 2026. Johannesburg, 20 October 2026 (Form 6-K, 21 October 2026). Pan African Resources. Holdings in company (TR-1): BlackRock crosses 5 per cent on 19 October, notified 20 October 2026; below 5 per cent on 22 October, notified 23 October; 5.00 per cent (2.97 per cent in shares, 2.03 per cent through financial instruments) on 23 October, notified 26 October 2026; summarised audited results for the year to 30 June 2026 (the R500 million repurchase programme). Gold Fields Limited. Statement on the Northern Star proposal, 28 September 2026. Northern Star Resources. Response to media speculation (the board's letter of 24 September 2026), 28 September 2026. Redwheel. Open letter to the board of Gold Fields on the Northern Star proposal, 14 October 2026, as reported by Miningmx. Gold Fields Limited. Media release: Gold Fields acquires nickel concentrator and exploration assets in Kambalda, 7 October 2026. Miningmx. Ghana signals Tarkwa may stay in foreign hands (the minister of lands and natural resources to Bloomberg News; the Engineers and Planners arbitration), 22 October 2026. AngloGold Ashanti plc. Remarks of the chief executive at Mining Forum Americas, Colorado Springs, 29 September 2026, as transcribed by Investing.com (South African approval for the buy-back; Arthur; Obuasi). Harmony Gold Mining Company. Appointment of independent non-executive directors (Alwyn Pretorius, James Boardman), SENS, 16 October 2026; pricing of US$500 million convertible bonds due 2031, SENS and Form 6-K, 22 September 2026. AngloGold Ashanti plc. Remarks of the chief financial officer at the Financial Times Mining Summit, London, 16 October 2026, as reported by Miningmx. INDUSTRY BODIES, REGULATORS AND OFFICIAL STATISTICS Minerals Council South Africa. Facts and Figures, editions 2015 to 2025 (national gold output; export classification in SARS data). Johannesburg. Minerals Council South Africa. Gold output expectation for 2026, as reported by Reuters via Miningmx, 24 March 2026 (with exploration spend 2006 and 2025). Statistics South Africa. Mining: production and sales (gold output series, 1970 to date). LBMA. Spotlight on the South African market (country profile), chapter 4: refining licences. London. LBMA. Country of Origin Data, gold and silver, 2018 to 2024 (data download, LBMA-CoO-Data-Download_2018_2024.xlsx; South Africa rows, totals by category). London, 2026. LBMA. Responsible sourcing country-of-origin returns for recycled gold, 2018 to 2022, as compiled by SWISSAID (below). South African Government News Agency. Minister Mantashe on the cost of illegal mining, 23 January 2025. Department of Mineral and Petroleum Resources. Draft Mineral Resources Development Bill, published for comment, May 2025 (artisanal and small-scale mining provisions). Government of South Africa. Draft criminal legislation on illegal mining, August 2026. Trans-Caledon Tunnel Agency. Integrated Annual Report 2025–2026 (acid mine drainage treated in the Western, Central and Eastern Basins; operating costs by basin, Table 6; the May 2025 decant near the Blue Sky shaft). Centurion, 2026. Supreme Court of Appeal of South Africa. Harmony Gold Mining Company Ltd v Regional Director: Free State, Department of Water Affairs [2013] ZASCA 206 (the 2005 Orkney directive under section 19 of the National Water Act; its addressees; survival of the directive after the sale of the land). Bloemfontein, 4 December 2013. Republic of South Africa. National Environmental Management Laws Amendment Act 2 of 2022, section 10 (section 24R of the National Environmental Management Act: environmental responsibility notwithstanding a closure certificate), commenced 30 June 2023 by Proclamation 125 of 2023. Pretoria. Public Investment Corporation. Holdings in listed gold companies, from the companies' shareholder registers and the Corporation's disclosures. South African Revenue Service. Trade statistics, HS 7108 (gold), as cited by the Minerals Council and SWISSAID. UN Comtrade. Trade in HS 7108, United Arab Emirates imports reported from South Africa, 2016 to 2023, as charted by SWISSAID (below). RESEARCH, SURVEYS AND INVESTIGATIONS SWISSAID. African Gold Report: South Africa profile (mirror trade; Rand Refinery exports; UAE-reported imports; LBMA recycled-gold returns; refining licences). Bern, 2024, online at africangoldreport.org. SWISSAID. On the Trail of African Gold: quantifying production and trade to combat illicit flows. Bern, May 2024. ENACT / Institute for Security Studies. Uncovered: the dark world of the zama zamas. Pretoria, 2019. PwC. SA Mine 2017 and SA Mine 2025 (estimates of the cost of illegal mining). Johannesburg. Metals Focus. Gold Focus, editions 2017 to 2024 (South African scrap supply). World Gold Council. Gold Demand Trends and country data, 2019 (gold smuggled out of South Africa, as cited). Handley, M. F. Estimate of the gold remaining in the Witwatersrand's tailings (about 1,600 tonnes), Journal of the Southern African Institute of Mining and Metallurgy, vol. 123, no. 4, April 2023. Carbon-intensity survey of 329 gold mines (Cooke at 9,980 kg CO2e an ounce). Edison Investment Research. Pan African Resources: research notes 2015 to 2021 (market values and multiples). Profile's Stock Exchange Handbook 2026/2. Gold Fields Limited (revenue by country). PRESS AND MARKET DATA Business Day. DRDGold's quarterly revenue and operating profit rise despite higher costs. Johannesburg, 21 October 2026. Miningmx. Gold Fields urged to drop 'dilutive' Northern Star bid. 14 October 2026. Miningmx. AngloGold Ashanti rules out buy-back binge. 16 October 2026. Stockanalysis. JSE, NYSE and ASX closes of 30 October 2026 and 31 December 2025. Trading Economics, USD/ZAR at the close of 30 October 2026. Kitco, spot gold at the New York close of 30 October 2026 (US$3,874.40) and at 31 December 2025 (US$4,325.45). Mining.com, Gold price slumps to one-year low, silver price breaks $50 as mining stocks tank, 30 October 2026. Business Day. Exploration decline threatens the future of South Africa's mining (Dunne, Wanblad, Nzimande; the R100 million junior fund; the dtic strategy of June 2026). Johannesburg, 10 October 2026. Bloomberg via Mining.com. South African mining exploration falls for seventh straight year (Statistics South Africa, 2025 exploration spend at 2015 prices; the Department's 5 per cent target). 13 March 2026. Daily Maverick. SA mining cadastre national rollout kicked to 31 March 2027 (Mbele at the Junior Mining Indaba, 9 June 2026; applications received and processed). 9 June 2026. AngloGold Ashanti plc. Exploration report, fourth quarter and full year 2025 (greenfields and brownfields spend; tenement countries). Denver, February 2026. Gold Fields Limited. Results for the year ended 31 December 2025, 19 February 2026, exploration spend as reported by Mining Reporters, February 2026. Mining Weekly. Harmony to acquire Mponeng and MWS from AngloGold, 12 February 2020; Harmony's high-grade and optimised underground margins, 11 September 2026; Stillwater strike settlement, 7 October 2026; Fraser on Northern Star (Reuters), 15 October 2026; Sibanye listing day, 11 February 2013. Miningmx. Kloof reserves and 2025 cost, 20 February 2026; Stewart on mine lives, 10 October 2025; gold mines "set up to operate until 2035", 13 February 2026; Burnstone capital, 12 October 2026; Pan African security costs and retrenchments, August 2025; Pan African results, 16 September 2026; Reuters on 2026 output expectation, 24 March 2026. Engineering News. Smaller refiners and scrap gold, 9 June 2021. Business Day. South Deep's cost to date, 9 February 2018; DRDGold's tailings partners, 19 August 2026. Daily Investor. DRDGold corporate history, 2024. Arrive Alive. Sentences in the Welkom illegal-mining case, April 2025. BullionStar. Rand Refinery: shareholders and throughput, updated 2025. West Wits Mining. Qala Shallows first gold pour, 17 March 2026. Kitco and Reuters. Spot gold and December futures settlement, 16 October 2026. PoundSterlingLive and Trading Economics. USD/ZAR closing rates, 16 October 2026. Federal Reserve Bank of St Louis (FRED). Year-end gold prices and USD/ZAR, 2015 to 2025. companiesmarketcap.com. Year-end market values for Harmony, Sibanye-Stillwater, DRDGold, Gold Fields and AngloGold Ashanti, 2012 to 2025, and at 30 October 2026 from the closes of that day; and for Newmont, Barrick, Agnico Eagle, Zijin Mining, Kinross, Evolution Mining, Alamos Gold, Endeavour Mining and Lundin Gold, 2015 to 2025 and at 30 October 2026. Stockanalysis, Northern Star Resources market value at the Sydney close of 30 October 2026, converted at the AUD/ZAR and USD/ZAR rates of the day. Stockanalysis, Stockopedia, Investing.com, Google Finance, Simply Wall St, Sharenet, Eulerpool, Halo and macrotrends. Closing prices, share counts, trailing multiples, yields and dividend histories at or to 30 October 2026, as stated beside each figure. HARMONY'S REPORTING SUITE Harmony Gold Mining Company. Reporting suite for the year ended 30 June 2026, published 23 October 2026: Mineral Resources and Mineral Reserves 2026 (gold reserves 27.4 Moz at US$2,915/oz, R1,650,000/kg at R17.60; resources 106.8 Moz at US$3,061; lives by operation; copper reserves 4.0 Mt); Operational Report 2026 (FY2027 guidance and lives by operation); Sustainability Report 2026 (material issues; fatalities); SENS announcement of the publication, the modification to the audited statements and the notice of the electronic AGM of 24 November 2026. Johannesburg. Harmony Gold Mining Company. Integrated and ESG reporting for the year ended 30 June 2025 (34,350 employees and 12,761 contractors; R20.2 billion of wages and benefits; R6 billion of taxes and royalties in South Africa), as reported by Mining Weekly, 28 October 2025. Johannesburg. Northbound Processing • Germiston, October 2026. The Money Metal — The Hinge Paper, The Four Who Stayed. This paper does not constitute investment advice. Company figures are from the published sources listed in the references; the shadow-ounce arithmetic, the leavers' ledger, the sum-of-parts, the per-ounce ratings, the ranks and the judgements are the author's; the mirror-trade figures are SWISSAID's and the LBMA's as cited and are read as a range, not a count. The author holds positions in Pan African Resources and DRDGold through a fund he manages, as disclosed above.

SA’s gold didn’t run out. The owners did – Duarte da Silva

Related Stories

Duarte da Silva: Ghana bought Africa’s shadow gold. SA wants to jail the man who digs it.

Dr Duarte da Silva: The Platinum corollary

Duarte da Silva: SA's richest goldmine is the bin — and almost nobody's mining it

Duarte da Silva: The country we declared explored