Gold has lost a third of its price in ten months, and I say the fall is nearly over. On Friday it fell to US$3,502, within three dollars of its peak of April 2025, and it held there. That price capped gold for four months last year and I say it is the floor now. A close below US$3,475 proves me wrong, and then US$3,187 is next. The bull market that began in 2022 is not finished. Central banks bought 675 tonnes in nine months and bought more in October, when the price fell 7 per cent. American inflation is 3.7 per cent against a policy rate of 4, and Washington borrowed two trillion dollars in twelve months. I expect gold above US$4,500 by the end of 2027 and at a new record in 2028. .By Dr Duarte F da Silva.Exhibit A · The FallI. A third of the price is gone, and the fall is nearly over Gold settled at US$5,318.40 an ounce in New York on 29 January 2026 and at US$3,515.40 on Friday 20 November. It has lost 34 per cent, or US$1,803 an ounce, in under ten months. Friday's was the lowest close since August 2025, and it ended a week in which gold lost 9.5 per cent, its worst week since March. The fall came in three breaks. On 30 January, the day after the record, gold lost 10.8 per cent in one session. Between 13 and 23 March it lost 12.9 per cent in six sessions. It then slid to US$3,992 in July, recovered to US$4,698 in August and broke for the third time in November, after a bad inflation number, a harder Federal Reserve and a dollar at its highest since April 2025. From April to August 2025 gold traded between US$3,187 and US$3,491. It left that range in September 2025 and rose by more than half in five months to reach the record. The whole of that rise has now been given back, and Friday's close is US$24 above the top of the old range. I say the rise from US$3,500 to US$5,300 was speculation laid on top of a bull market that began in 2022, and that the speculation has now been sold out of the price. The bull market underneath it is intact. The rest of this note sets out who sold, who did not, and why I hold that view. .Figure 1 shows what drove the fall. The real yield, which is the yield on a ten-year American inflation-protected bond, averaged 1.82 per cent in February and stood at 2.81 per cent on Friday, a third of a point below the record of 3.15 per cent set in November 2008. Gold pays no interest, so the real yield is the income a holder gives up to own it. The real yield has risen by one percentage point since February, and gold has fallen 30 per cent from its February average. The Federal Reserve is the cause. Under its chairman, Kevin Warsh, it raised its policy rate on 16 September to a range of 3.75 to 4 per cent and held it there in October by ten votes to two. The minutes of that meeting, published on 18 November, record that nearly all the officials expect to raise again at the meeting of 8 and 9 December, and two of them raised the possibility of a half point. The futures market puts the chance of a rise at about 95 per cent. Goldman Sachs expects another in March and no cut before November 2027. Gold has given back the whole of its run from September 2025 to January 2026. Friday's close is US$24 above the range it left in September 2025. Exhibit B · Who Sold, And Who Did NotII. The tourists sold. The central banks bought Two kinds of money own gold. The first buys because the price is rising. I call it the tourist money: the futures traders and the Western funds that arrive after a rise and leave when it stops. The second buys gold as a reserve, and that is the central banks, whose buying has driven this market since 2022. The tourists are the sellers. The World Gold Council's figures for the third quarter show that gold-backed funds bought 211 tonnes between July and September, for a record US$31 billion, and took their holdings to a record 4,256 tonnes. They paid about US$4,300 an ounce on the average price of the quarter, and they are now sitting on a loss of 18 per cent. The first weekly outflow came in the week to 23 October, and US$1.4 billion has been withdrawn from the largest American fund in the last thirty days. On Friday a senior metals trader told Reuters that the fall in technology shares was forcing funds to sell gold to cover losses elsewhere. The S&P 500 lost 5.9 per cent in the last three sessions of the week. The central banks bought throughout. They took 321 tonnes in the third quarter, 42 per cent more than a year earlier, and 675 tonnes in the first nine months. The buyers themselves reported only 120 of the quarter's 321 tonnes, and the rest was bought without announcement. In October gold fell 7 per cent and ended the month below US$3,900, and the buying increased. Poland bought 27 tonnes and reached the 700 tonnes it set as its target in January. China bought 26 tonnes, its largest month in three years and its twenty-fourth month of buying in a row. Uzbekistan bought 7 tonnes and the Czech Republic 2.5. Those four banks bought 62 tonnes between them, which is more than every reporting central bank bought in September after deducting the sellers, and September's 61 tonnes had been the highest month of the year. .Figure 2 converts the buying into dollars, which is how a treasury counts it. The tonnes have fallen since the peak of 2022 but the money spent has kept rising. Central banks spent about US$63 billion on gold in 2022 and about US$95 billion in 2025. In nine months of 2026 they spent about US$99 billion, which on my arithmetic is more than in any full year before it, and October's 62 tonnes cost the four buyers another US$8 billion. The 675 tonnes bought to September are worth US$76 billion at Friday's price, so the central banks are US$23 billion down on those purchases and they are still buying. A fund that is 18 per cent down has a risk committee and clients who want their money back. A central bank that is 23 per cent down has neither, because it did not buy gold for a return by December. It bought because in 2022 it watched US$300 billion of Russia's reserves frozen by the governments that had issued them, and it decided to hold part of its reserves in an asset that no other government can freeze. The fall in the price this year has not changed that decision. Central banks spent more on gold in nine months of 2026 than in any full year before it. They are US$23 billion down on those purchases and they are still buying. Exhibit C · The SupportIII. US$3,500 is the support, and it has not broken My first call is that the low of this fall is in or within sight, at the support between US$3,475 and US$3,500. The next four months turn on that level, so I set out exactly where it comes from. On 22 April 2025 spot gold touched US$3,500.05 and turned back. That was the record at the time, and for the next nineteen weeks gold could not hold above it. It tried in June and again in August, and the highest COMEX close of the whole period was US$3,491, on 8 August. It broke through on 2 September 2025, and from that day until this week it never traded below US$3,574. A price that capped the market for four months and then became the base of a 52 per cent rise is the strongest support on the chart.A second measure gives the same level. From the low of US$1,631 on 3 November 2022 to the record of US$5,318 on 29 January 2026 gold rose US$3,687. If half of that rise is given back the price is US$3,475. The old ceiling and the halfway point of the whole bull market are within US$25 of each other, and that is why I put the support at US$3,475 to US$3,500. The market tested it on Friday. Spot gold fell to US$3,502 and the COMEX contract to US$3,502.90. Both stopped within three dollars of the April 2025 peak, and the contract settled at US$3,515.40. Two other readings show how far the selling has gone. The fourteen-day relative strength index, which measures how hard a market has been sold, closed at 22 on Friday. That is its lowest reading since October 2023 and lower than at the lows of March and June. The price is also 21 per cent below its average of the last two hundred sessions. The support can still break, and I do not say otherwise. Four lesser supports have gone in three weeks: US$4,000, where gold turned in June and July; US$3,890, the lows of 30 October and 13 November; the US$3,800 that UBS had named; and on Friday the US$3,605 that JPMorgan had named. The Federal Reserve is expected to raise rates on 9 December. The funds held a record 4,256 tonnes at the end of September, worth US$481 billion at Friday's price, and they have sold very little of it so far. One day at the support does not make a low. The test is a daily close below US$3,475. If that happens the support is broken and I am wrong on the low. A support of this size breaks violently when it goes. From December 2011 gold held US$1,537 on a closing basis for sixteen months and through three tests. It closed below that level on 12 April 2013, lost 9 per cent in the next session, and by 27 June stood at US$1,212, which was 21 per cent below the old support. If US$3,475 goes, the next levels are US$3,187, the bottom of the 2025 range, and US$3,040, where gold would have given back 61.8 per cent of the bull market. For as long as gold closes above US$3,475 I read every fall towards it as another test of the same support. The South African reader must do this sum in rand, because a mine here sells its gold for dollars and pays its costs in rand. The rand has fallen with gold. It was R15.82 to the dollar on the day of the record and R17.59 on Friday, so gold has lost 34 per cent in dollars and 26 per cent in rand. Friday's close is R61,800 an ounce, or R1.99 million a kilogram, against R2.70 million at the record. Harmony received an average of R2.07 million a kilogram in its year to June 2026, a year of record revenue, and Friday's price is 4 per cent below that. Its all-in sustaining cost for that year was R1.19 million a kilogram, 13 per cent more than the year before, which leaves a margin of about R800,000 a kilogram at Friday's price. The dollar cost of a South African mine is not a fixed number, and it is a mistake to treat it as one. Harmony's R1.19 million a kilogram was US$2,343 an ounce at the exchange rate of 29 January and is US$2,107 at Friday's. The company did not spend one rand less. The currency cut its dollar cost by 10 per cent. The same applies if gold falls further. At US$3,187 and Friday's exchange rate gold is R1.80 million a kilogram, and at that dollar price every rand the currency loses against the dollar adds about R100,000 a kilogram. In the week in which gold fell 9.5 per cent the rand fell nearly 4 per cent. The rand has taken more than a fifth of this year's fall away from the South African producer. The rest of the industry has less protection. Its average all-in sustaining cost was a record US$1,930 an ounce in the second quarter, so the margin on an ounce was US$3,076 in the first quarter and is US$1,585 at Friday's price. The dollar price is down a third and the dollar margin is down by half. That US$1,930 was struck at the exchange rates of the second quarter, and the stronger dollar will lower it for every producer that pays its costs in another currency. US$3,500 capped gold for four months in 2025. On Friday it was tested as support, to within three dollars, and it held. Exhibit D · The Safe HavenIV. Gold failed as a safe haven last week, as it does in every liquidation Gold did not protect anyone last week. In the five sessions to 20 November the S&P 500 fell 8.6 per cent and gold fell 9.5 per cent. The money that left shares went to the dollar, which rose to its highest since April 2025, and to Treasuries, where the ten-year yield fell by a quarter of a point. Gold was sold with the shares, to pay for the losses on them. Gold did the same in the first weeks of the last two liquidations, which Table 1 sets beside this one. .TABLE 1 · Gold in three liquidations. S&P 500 closes and COMEX gold settlements. 2008 is measured from 12 September, the last close before Lehman Brothers failed, to gold's low on 13 November; 2020 from 9 to 18 March. The author's arithmetic. In 2008 gold rose first. In the week after Lehman Brothers failed it went from US$765 to US$909, a rise of 19 per cent. The selling reached it after that, and over the next seven weeks it fell 22 per cent, to US$705 on 13 November, while the S&P 500 fell 27 per cent from its level before Lehman. In March 2020 the selling came at once, and gold lost 12 per cent in seven sessions while shares lost 13 per cent. The Federal Reserve ended both falls. In 2008 it began buying bonds on 25 November and cut its rate to zero on 16 December, and gold was back at US$1,002 on 20 February 2009, 42 per cent above its low in three months. In 2020 it cut to zero on 15 March and promised unlimited bond purchases on 23 March. Gold regained its old level in thirteen sessions and set a record of US$2,069 on 6 August, 40 per cent above its low. My reading of that record is that gold protects its holder against what the Federal Reserve does after a crash. During the crash it is one of the few assets that can still be sold at a fair price, and so it is sold. That is the danger to my first call. The Fed is raising rates into this fall, and for as long as it does the funds can go on selling. A full quarter of central bank buying, 321 tonnes, is worth about US$36 billion at Friday's price, and the funds hold US$481 billion. Central banks set the level at which a fall ends, but the funds and the Fed set the date. I still expect the support to hold, and the same record is my reason. A share market that loses 8.6 per cent in a week changes the Federal Reserve's sums, and the ten-year yield has already fallen a quarter of a point on it. The 2008 liquidation lasted seven weeks and the 2020 one seven sessions, and each ended when the Fed turned. Gold has now fallen for three weeks, by 15 per cent from its October average, to the strongest support on its chart, and the share market the Fed watches is breaking. If the Fed holds on 9 December, or says that December's rise is its last, I expect the low to be in on that day. Gold protects its holder against what the Federal Reserve does after a crash. During the crash it is sold with everything else. Exhibit E · The PrecedentsV. This fall is 2008, not 1980 The bears say this is 1980 again. I say it is 2008, and Table 2 is my evidence. It sets five of the large falls in gold since 1971 beside this one and records two things about each: how far the policy rate stood above inflation when the fall ended, and whether governments and central banks were buying gold or selling it. .TABLE 2 · Six falls in the gold price of a fifth or more since 1971. London prices for the first five, rounded; COMEX settlements for 2026. The policy rate is the federal funds rate less the latest annual rate of consumer price inflation at the date of the low. The author's reading of the record. Only the fall of 1980 ended a bull market for a generation. Paul Volcker took the policy rate to seven points above inflation and held it there, gold fell 65 per cent, and it did not see its old high again for twenty-eight years. The bears point to a hawkish chairman, a rising real yield and a price that had more than tripled in little over three years, and they call it the same thing. I reject the comparison. Volcker paid savers seven per cent more than inflation to hold dollars. Mr Warsh pays them 0.2 per cent. The policy rate is 3.75 to 4 per cent and consumer prices rose 3.7 per cent in the year to September. After two more rises the rate will be 4.25 to 4.5 per cent, and unless inflation falls the real rate will still be under one point. It took seven points to break gold in 1980, and this Federal Reserve is more than six points short of that. The falls that resemble this one are 2008 and 1974. In 2008 funds sold gold to meet losses elsewhere, as they did last week, and the price fell 30 per cent in seven months and made a new record eleven months later. Between 1974 and 1976 gold fell 47 per cent while governments auctioned it, and then rose eightfold in forty months. In both cases the real policy rate at the low was at or below zero, as it is today.The fall of 2011 to 2015 is the case against me. Rates were at zero, central banks were buying, and gold still fell for four years. Three things were different then. Inflation was between zero and two per cent and falling, the American deficit was shrinking from nearly seven per cent of national income to under three, and central banks were buying a third less than they buy now. None of those three conditions holds today. Exhibit F · The Next Two YearsVI. Why this Federal Reserve cannot do what Volcker did My second call is that gold is above US$4,500 by the end of 2027, and my third is that it makes a new record, above US$5,318, in 2028. Both rest on one judgement: the Federal Reserve cannot keep real rates high for long, because the United States Treasury cannot afford it. The numbers are the Congressional Budget Office's. In the year to 30 September 2026 Washington ran a deficit of US$1.99 trillion, which is 6.2 per cent of national income, in an economy that was growing. Its debt reached US$40.2 trillion. Net interest on that debt cost US$1.14 trillion, 11 per cent more than the year before and more than a fifth of everything the government collected in tax. Each further point on the average rate the Treasury pays adds about US$400 billion a year to that bill once the debt is refinanced. When Volcker raised rates in 1980 the federal debt was about a third of national income, and he could take the policy rate to nineteen per cent because the borrower could carry it. Today's borrower owes well over a year's national income and is adding two trillion dollars a year. Every rise in rates now adds to the deficit of the Fed's own government. The thirty-year Treasury yield is already 5.33 per cent. I expect the bond market to force an end to this tightening before the Fed chooses one. The bond market is also pricing something I do not believe. The ten-year Treasury yields 4.92 per cent and the ten-year real yield is 2.81 per cent. The difference, 2.1 per cent, is the inflation the market expects over the next ten years. Inflation is 3.7 per cent in America and 4.0 per cent in the euro area, its highest there since 2023. Unless inflation falls by nearly half, the true return on a Treasury is closer to one per cent than to 2.81. Gold is priced today as if the bond market is right, and I think the bond market is wrong. The central banks do the rest. At 75 tonnes a month, their pace in 2026, they take about a quarter of each year's mine supply out of the market and keep it. They bought at US$5,000 and they are buying at US$3,500. Once the funds have finished selling, the buyers who remain set the price, and those buyers have been adding gold for four years. The banks agree with me on direction and differ on the path. Goldman Sachs has cut its forecast twice this month and still sees US$4,650 at the end of 2027. UBS sees US$5,000 and Morgan Stanley US$5,250. Goldman and UBS also have gold at US$4,150 and US$4,500 at the end of this year, five weeks from now, and I think both are too high. My call needs only that the support holds. From Friday's close a price of US$4,500 at the end of 2027 is a rise of 28 per cent in thirteen months. After the low of 2008 gold rose 34 per cent in nine months, and after the low of 2022 it rose 19 per cent, so the call is within what gold has done before. .TABLE 3 · The three calls, what each rests on, and what would prove each wrong. Prices are COMEX settlements in US dollars an ounce. The author's judgements at the close of 20 November 2026. Exhibit G · The CeilingVII. The debt ceiling will lift gold in the second half of 2027 The United States will reach its debt ceiling early in 2027, and I expect the fight over it to lift gold in the second half of that year. None of the bank forecasts I have quoted rests on it. The Congress that must raise the limit is not the Congress that raised it last time. The limit is US$41.1 trillion. Debt subject to it stood at US$40.33 trillion on 30 October, which left US$777 billion of room, and the Treasury borrows about US$200 billion a month. The Bipartisan Policy Center expects the limit to be reached in February or March 2027. After that the Treasury lives on its cash and on accounting measures for six to nine months. Bank of America puts the day the money runs out, the X-date, in November 2027, and says it could come as early as August. The politics have changed since the limit was last raised. On the Associated Press count on the morning of 21 November, Democrats had won 257 seats in the House and Republicans 173, with five counted for neither. In the Senate the count was 56 for the Democrats, including two who caucus with them, and 44 for the Republicans. The new Congress sits on 3 January. Every earlier standoff set Republicans in Congress against a Democratic president. On these numbers the roles are reversed. CNBC reported on 4 November that Democrats are preparing to use the increase to seek concessions, though they have announced no demands. One event removes all of this. President Trump has asked Republicans to raise the limit in the weeks before the new Congress sits, without Democratic votes. If they do it, there is no ceiling to fight over in 2027 and this argument falls away. Their own fiscal hawks have resisted for months, and they must also pass a funding bill by 11 December. If the limit is untouched on 3 January, the standoff is coming. Table 4 records what gold, the dollar and the ten-year yield did the last four times the ceiling was reached. .TABLE 4 · Gold, the dollar and the ten-year Treasury yield in the four debt-ceiling episodes since 2011. COMEX gold settlements, the dollar index and the ten-year yield at the closes of the dates shown. The author's arithmetic. Gold rose in three of the four. The largest rise was in 2011, the only standoff that ran to a deadline and a downgrade, when gold gained 26 per cent in ten weeks. It fell in 2013, when it was in a bear market and the Federal Reserve was preparing to withdraw its support. In 2023 it reached a record a month before the deadline and gave back 6 per cent once the deal was signed. The dollar does not collapse in a standoff, and I do not claim that it does. In 2011 the dollar rose and the ten-year yield fell by more than a point, because investors bought the debtor's own bonds while the debtor argued about paying them. The dollar fell hard only in 2025, when there was no standoff and other forces were driving it. The asset a standoff moves is gold. All three rating agencies have now cut the United States, in 2011, 2023 and 2025, and each cut came in a year when the ceiling was in play. The same record gives the timing. The limit was reached on 16 May 2011 and gold did nothing for six weeks: it was US$1,490 on the day and US$1,482 on 1 July. The whole rise came in the five weeks before the deadline and the five weeks after it. On that pattern gold does not respond in February or March, when the limit is reached. It responds in the weeks before the X-date, between August and November 2027. The ceiling therefore leaves my dates where they are and gives the second call a calendar. A rise the size of 2011's, measured from Friday's close, takes gold to US$4,440 without any help from the Fed. It also strengthens the argument of Exhibit F, because Congress cannot refuse to lift the limit while the central bank keeps raising rates and the Treasury runs out of cash. I expect the months that bring the X-date to be the months in which the Federal Reserve has to stop. In 2011 the ceiling was reached in May and gold did nothing for six weeks. Then it rose 26 per cent in ten weeks around the deadline. The 2027 deadline falls between August and November. Exhibit H · The BearVIII. What the bear says, and what I answer The bear has three arguments. I set out each as he would make it, and I reject all three. He says the Fed means it. The minutes point one way, a governor is asking for half a point, and the same governor says the bank's credibility is at stake. I accept that for December, and I have already said it is the danger to my first call. I do not accept it for the next two years. A real policy rate of plus 0.2 is not tight policy. To reach Volcker's seven points the Fed would need a policy rate above ten per cent, and the Treasury's interest bill would more than double. That will not happen, and the bear knows it.He says inflation will fall and take the case for gold with it. Petrol was most of September's number and energy prices have eased, so he may be right about the headline rate. But core inflation rose to 2.8 per cent in October, a six-month high, with the Fed already raising rates. If inflation does fall to target, the Fed stops, the real yield stops rising and the pressure on gold lifts. The bear needs inflation to fall while rates stay high for years, and a government with forty trillion dollars of debt cannot afford that combination. He says the central banks will stop. He says they are sensitive to price, and that a buyer who is US$23 billion down on the year will be told to stop by his finance minister. He has some facts, and here they are. Poland bought 150 tonnes this year, more than any other central bank, and it has now reached its target and named no new one. Kazakhstan says gold is 78 per cent of its reserves and that it will cut that to 75 per cent by the year-end. Russia sold 62 tonnes in nine months and Turkey sold 82. I do not accept his conclusion. Central banks bought 321 tonnes in the quarter in which the price fell, the most since late 2024, and that total is after every one of those sales. Poland's governor said on 5 November that he sees no reason to stop buying. China has bought 129 tonnes in ten months, 89 of them since July, and gold is still only 7.9 per cent of its reserves. The Council's own reading is that central banks now respond to price in one direction only, by buying more when it falls, and October bore that out. This is the test I watch most closely, and it is in Table 3. If official buying falls below 150 tonnes a quarter for two quarters running, I am wrong and I will say so. The fourth risk is my own and not the bear's: the support could break. If gold closes below US$3,475 I am wrong on the low and I will say so that day. The next levels are US$3,187 and US$3,040, and April 2013 showed how fast a market travels between such levels once a support has gone. My dates would then move out by some months. My direction would stay the same, because nothing in Exhibits B, F and G depends on where the low is. The VerdictIX. Do not sell the metal. Do not chase it either I make three calls and I have put a number on each. The support at US$3,475 to US$3,500 holds, and the low of this fall is in or within sight. Gold is above US$4,500 by the end of 2027. It makes a new record in 2028. If you hold the metal, keep it. The sellers today are funds with losses to cover and committees to answer to. The buyers are central banks that have bought for four years and have not sold in this fall. A holder who sells at US$3,500 sells a third below the record, at the strongest support on the chart, to the buyers who have been right for four years. I would not do it. If you want to buy, the support tells you how. Gold is a third below its record and at its support. Above US$3,475 I would buy in parts, because I know the exact price at which the view is wrong. If gold closes below US$3,475 I would stop and wait for US$3,187. The Fed meets on 9 December, and that day will test the support harder than Friday did. If you own gold shares, work out the margins before the market does. For the industry a third off the dollar price is half off the dollar margin, and the South African producers are cushioned only for as long as the rand keeps falling with gold. The companies that spent 2026 buying back their shares and bidding for one another at US$4,300 an ounce did it with cash that US$3,500 no longer produces. Their next results will show it, and until they do I would rather own the metal than the miners.To the boards I say keep the cash. In 2026 a gold company could buy back its shares, pay a special dividend and bid for a rival in the same year, and that ended on Friday. The companies that kept their balance sheets will be the buyers in 2027, and the others will have to take the prices they are offered. .US$3,500 CAPPED GOLD FOR FOUR MONTHS IN 2025. ON FRIDAY THE TOURISTS SOLD IT DOWN TO THAT PRICE, AND IT HELD. THE TEST IS A CLOSE BELOW US$3,475. THE CENTRAL BANKS ARE STILL BUYING. A FED THAT PAYS 0.2 PER CENT ABOVE INFLATION, TO A BORROWER THAT OWES FORTY TRILLION DOLLARS, WILL NOT BREAK GOLD. ABOVE US$4,500 BY THE END OF 2027. A NEW RECORD IN 2028. Northbound Processing • Germiston, November 2026. Gold Note — the gold price, short and medium term. This note is the author's opinion on public information and does not constitute investment advice; prices are at the close of 20 November 2026 and the forecasts are the author's judgements, which may prove wrong. .Sign up for your early morning brew of the BizNews Insider to keep you up to speed with the content that matters. The newsletter will land in your inbox every morning on weekdays. Register here.Support South Africa's bastion of independent journalism, offering balanced insights on investments, business, and the political economy, by joining BizNews Premium. Register here. If you prefer WhatsApp for updates, sign up to the BizNews channel here.