August has two numbers. Goldman Sachs estimates 79 tonnes. The World Gold Council counted 39 tonnes that central banks reported. The Dutch bars that changed cities were not new demand. A mining share does not collect a tonne. It collects the gold price minus the cost, if the ounces show up.
This is not a recommendation to buy gold, a gold-mining stock, a fund, or a bar. An estimate is not a purchase order. A reported tonne is not a price target. A transfer between vaults is not a bid. Nothing here is personal advice, and nothing here is a promise that gold mining stocks will rally.
The sentence that will travel is the large one. Goldman Sachs tracks 79 tonnes of central bank purchases in August. The sentence underneath it is smaller, and it has a method on it. InvestingLive, reporting a Goldman note on Monday, October 19, 2026, said the bank estimates that central banks bought 79 tonnes that month. The bank builds that gauge from trade and flow data. It does not wait for every central bank to publish. Those estimates often run above the reported totals. They can be revised when more data arrive. Treat 79 as a model of flows. Do not treat it as a scale ticket.
Put the two August numbers side by side
On October 6, 2026, the World Gold Council published the other number. Marissa Salim’s note said reported net buying in August totalled 39 tonnes. Year to date, reported purchases were 170 tonnes. Reported means what central banks and the IMF data showed, where the numbers were in. It is a count of disclosures. It is not a claim that no other ounce moved.
Seventy-nine minus 39 is 40 tonnes. That gap is the whole argument, and it is not a scandal by itself. The two figures are not measuring the same drawer. Goldman is estimating purchases it thinks the trade data imply, including buying that disclosures miss or delay. The Council is adding up what was reported, net of the banks that sold. One can be high. The other can be incomplete. Using either number as “the” August total is how a headline gets ahead of the metal.
At about $4,125 an ounce, which is where spot sat on October 20, 2026, one tonne of gold is worth about $133 million. Seventy-nine tonnes is about $10.5 billion. Thirty-nine tonnes is about $5.2 billion. The 40-tonne gap is about $5.3 billion. Those are paper values of a weight. They are not cash that arrived in a miner’s account in August. They are not a cheque written to a Toronto share. A gold price forecast that starts from $10.5 billion of estimated official buying and jumps to a rally has skipped the step where the ounce has to be bid in the market, and the further step where a company has to produce an ounce below that price.
Goldman said the 79 tonnes was far above the pre-2022 average of 17 tonnes a month. Seventy-nine divided by 17 is about 4.6. More than four times the old monthly pace, on the bank’s own estimate. The same note put Goldman’s July estimate at 44 tonnes. So the bank’s own series stepped up from 44 to 79. That is a change inside one model. It is not proof the world bought 35 more tonnes than it bought in July. The Council’s July reported net figure, published earlier, was 23 tonnes. July also has two numbers. August also has two numbers. Keep the pairs intact.
Who actually showed up in the count
The Council’s August list is specific. The People’s Bank of China added 20 tonnes. That was its 22nd straight month of reported buying. China’s reported additions for the year stood at 80 tonnes. Official gold reserves were about 2,387 tonnes, or 9 percent of total reserves.
The National Bank of Poland bought 8 tonnes in August. Poland still led the year, with 98 tonnes bought so far and holdings of 648 tonnes. Its stated target is 700 tonnes. The Central Bank of Uzbekistan bought 8 tonnes. Its year-to-date buying was close to 50 tonnes. Holdings were about 439 tonnes, or 90 percent of reserves. The National Bank of Kazakhstan bought 7 tonnes. Year to date was close to 36 tonnes. Holdings were about 377 tonnes, or 79 percent of reserves.
Smaller reported buyers included the Czech National Bank, at 2 tonnes, its 42nd straight month, and the Central Bank of the Republic of Turkey, at 3 tonnes. Bolivia and Ghana added about a tonne each. The net figure is a net. Russia sold 6 tonnes in August and was down 56 tonnes for the year, with holdings of 2,271 tonnes. Jordan sold 3 tonnes in the month, though it was still up 2 tonnes for the year, at 75 tonnes. Turkey’s 3-tonne buy did not repair its year. The Council still had Turkey down 82 tonnes year to date.
Add the named August buys and you are already above 39 before you subtract the sales. China, Poland, Uzbekistan, Kazakhstan, the Czech Republic, Turkey, Bolivia, and Ghana come to about 50 tonnes if the last two are taken at about a tonne each. Russia and Jordan sold 9. That named net is about 41 tonnes, not 39. About 2 tonnes sit in smaller lines the Council did not itemize. The list is not a complete balance sheet. That is what “net” means. A headline that says central banks bought 39 tonnes is true as a net reported figure. It hides a larger gross buy, a real sell, and a small remainder. A headline that says they bought 79 tonnes is true only as Goldman’s estimate. It hides the method. Both headlines are usable. Neither is the whole month.
China is where the two methods touch. Goldman estimated 25 tonnes of Chinese buying in August. The People’s Bank reported about 20. The gap is 5 tonnes, not 40. InvestingLive said Goldman reads that gap as accumulation larger than the disclosure, in line with the bank’s view that Chinese buying has run above official numbers this year. Five tonnes is a specific claim about one buyer. It is still an estimate. It is not a warehouse receipt you can read. The other 35 tonnes in the 79-versus-39 gap are not itemized in the public recap of the note. Do not give those 35 tonnes a flag, a vault, or a motive they have not been given in the source.
The bars that changed cities
On September 2, 2026, De Nederlandsche Bank said it had moved gold. Between March and August, about 86 tonnes left the combined pile of about 313 tonnes it held in the United States and Canada, and the position in London rose. The stated reasons were crisis preparedness, liquidity, and tradability. London is a major physical trading centre. The bank said reserves in New York and Ottawa cannot be used as quickly in a crisis, and that a more balanced spread helps.
The locations, before and after, are on the bank’s own table. Zeist stayed at 30.8 percent. London rose from 18.1 percent to 32.1 percent. New York fell from 31.3 percent to 18.5 percent. Ottawa fell from 19.7 percent to 18.5 percent. The home pile did not grow. The foreign piles were rearranged.
The method matters more than the map. About 59 tonnes were sold in New York and bought again in London, in bars that meet the international market standard. More than 27 tonnes were shipped from the United States and Canada to Zeist, and a similar quantity that already met the standard moved from Zeist to London, so the bars did not have to be remelted. Read that again. Fifty-nine tonnes were sold and replaced. They were not added. The Dutch central bank did not become a 59-tonne buyer of new reserve gold. It changed the city of a claim it already had. The physical leg is a shipment, not a mine.
This is the mistake that turns a custody story into a demand story. A sale in New York and a buy in London can show up in trade data. A model that watches flows can see two prints. The reserve did not grow by two prints. If some of Goldman’s 79 tonnes is a relocation wearing the clothes of a purchase, the estimate is not “wrong” so much as it is answering a different question. The question “where did the bar go” is not the question “how many new bars did official buyers take off the market.” The Council’s 39 tonnes is an attempt at the second question, limited to what was reported. Goldman’s 79 is an attempt that includes what the flows suggest. The Dutch 86 is mostly the first question. Do not add 86 to 79. You would be counting a move as a purchase, and then counting it again.
How central bank gold buying affects gold prices
Official buying can support the gold price. It does not set it. The mechanism is dull, which is why it gets dressed up.
A central bank that wants more gold has to get it from a miner, a holder, or the existing stock in a trading centre. If it buys in the market and holds, those ounces are no longer offered to the next buyer. A steady bid, repeated every month, matters more than one loud month. Goldman’s longer view, in a September note, was that official buying near 90 tonnes a month, against that same 17-tonne pre-2022 average, is the structural piece. InvestingLive, reporting that note, described this demand as less sensitive to price than investor demand, and as a bid that can still be there when exchange-traded funds are selling. That wording is the reporter’s account of the note, not a sentence in quotation marks. The point, if the account is fair, is still the useful one. A buyer who does not flinch at a high price is a different buyer from a fund that redeems when the price falls.
Less sensitive is not insensitive. Central banks have targets, budgets, and politics. Poland is buying toward 700 tonnes and is already at 648. A target that is close is a different bid from a target that is far. Uzbekistan already holds about 90 percent of its reserves in gold. A bank that is almost all gold is not the same incremental buyer as a bank whose gold is 9 percent of reserves, which is where China reported itself. China’s pile is huge in tonnes and still modest as a share. That is why strategists stare at China. A small change in the share is a large change in tonnes. It is also why the 5-tonne gap between 20 and 25 gets more attention than it would in a smaller country. Attention is not confirmation.
The other side of the scale is still there. Funds, coin buyers, jewellers, miners, and traders set the day. If investors are selling more than officials are buying, the price falls while the official bid is “strong.” That is not a contradiction. It is arithmetic. August’s estimated 79 tonnes, or the reported 39, did not stop gold from sitting near $4,125 in late October. GoldPrice.com puts the all-time closing high at $5,405 on January 29, 2026. From that close to about $4,125 is a drop of roughly 24 percent. The same site’s Monday, October 19 close, $4,140.81, was 23.4 percent under $5,405. Other daily series that week closed nearer $5,500 to $5,560, and the spot high was about $5,560. The 24 percent is the drop from the $5,405 close, not from that spike. The tonnes did not repeal the drop. They also did not cause it. They sat inside a year when the price had already spiked and given a large part of the spike back.
So how does the buying affect the price, in a sentence a person can use? It is a bid that can lean against selling. It is not a floor you can see on a chart. The size of the lean depends on whether the tonnes were new demand or a change of city, whether they were reported or inferred, and whether the rest of the market was buying or leaving. A model that holds the official bid near 90 tonnes a month can still miss the next $200 move if the fund trade goes the other way. A model that ignores the official bid can miss why the dips have a buyer at all.
Can those purchases trigger another gold price rally
Can central bank gold purchases trigger another gold rally? They can help. They cannot pull the trigger by themselves, and August is the demonstration.
A rally is a rise from here, not a memory of January. The metal already rallied, hard, and then fell back. Official buying was part of the story on the way up, in the telling of the banks that forecast it. It did not stop the way down. If 79 tonnes in a single month, or 39 reported, were enough to start the next leg, the chart in October would not still be arguing with the January high. The bid is real. The switch is not.
Goldman’s own gold price forecast, from the September note reported by InvestingLive and eFXdata, separates timing from destination. Lina Thomas cut the bank’s end-2026 fair value to $4,650 an ounce, from $4,900. The bank kept an end-2027 target of $5,400. The argument was that Federal Reserve tightening would slow the climb through fund demand, not cancel the later level. Economists at the bank, in that note, still looked for cuts later, and they treated official buying as the piece that offsets the drag from higher rates. That is a gold price forecast with a date and a condition. It is not a statement that August’s 79 tonnes already did the work.
Read the levels against the spot you can trade. About $4,125 in late October is under the $4,650 year-end figure. It is much further under $5,400. The gap is the forecast. It is not money you have been paid. Between here and $5,400 sits a Fed path, a fund flow, a mine supply, and whatever the next monthly estimate says when it is revised. Goldman has already revised a year-end number once, down by $250, inside 2026. A target that was cut can be cut again. A target that was held can be held for reasons that do not arrive on time.
There is a further limit inside the September math. InvestingLive’s account of the note said official buying was driving nearly all of the gain the bank expected through 2027, at around 90 tonnes a month. August’s 79 is close to 90. It is one month. A year of 90-tonne months is about 1,080 tonnes. The Council’s reported pace through August was 170 tonnes in eight months, or about 21 tonnes a month of reported net buying. If the true pace is the reported one, the 90-tonne assumption is far too high. If the true pace is Goldman’s, the reported tally is missing most of the bid. The gold price rally in the forecast depends on which of those worlds you are in. The note does not let you live in both without saying so.
What a tonne is worth to a share
Are gold stocks set for another rally? Not because a bank counted tonnes. A gold mining stock is a claim on a business. The business sells ounces. The share moves when investors change what they will pay for the next ounce of cash flow. Central bank gold demand can lift that ounce’s price. It does not drill the hole, staff the mill, or get the permit.
The link, when it works, is simple. A higher gold price, if costs stay put, widens the margin. A wider margin, if the ounces are produced, becomes cash. Cash can pay debt, fund a dividend, or sit on the balance sheet until someone bids for the company. None of those steps is automatic. A mine with an all-in sustaining cost near the spot price does not collect a rally. It collects a remainder that can vanish if the price slips or the cost rises. A mine with a low cost collects more of each dollar. The same 79 tonnes, if they ever showed up in the price, would not be shared equally.
Costs are not the only leak. A company can miss ounces. A pit wall, a mill, a power cut, a government, a grade that was not there. The missed ounce is an ounce the central bank did not buy from you, because you did not pour it. The official bid cannot replace it. Investors who buy the sector because “central banks are buying” are buying a story about the metal and hoping every company is a clean pipe to that metal. Some are. Some are a pipe with a hole.
There is also the timing insult. August’s buying, whether 39 or 79 tonnes, is already in the past. The price since then has not started a new high. If you buy a miner today because of August, you are late for August. You are early only for the next month, and the next month has not been estimated yet. October’s Chinese figures, InvestingLive noted, were due around November 7. Until they print, the streak is a streak, not a guarantee. September has already printed. Bloomberg and the Global Times reported an official addition of 740,000 ounces, about 23 tonnes, and the 23rd straight month. InvestingLive called it the largest monthly add in about three years. Keep the ounces. One outlet’s “21 tonnes” for that month divides by the wrong kind of ounce. Twenty-three tonnes of reported Chinese buying is not Goldman’s 25, and it is not a price for your share.
Canadian gold stocks are a cost curve, not a list of winners
The place this question lands, for a lot of readers, is Canadian gold stocks and the TSX gold stocks that trade with them. The Toronto market is where a large share of the world’s listed gold miners live, from senior producers to single-asset stories. That fact makes Canada a window. It does not make every window the same glass.
Gold stocks to watch, if you are going to watch any, are the ones where you can answer five questions without a slogan. What did it cost to produce the last ounce? How many ounces is the company actually guiding, and has that guide been cut? How much of the margin is spoken for by debt, a hedge, or a government take? What has to go right in the next two quarters that is not “the gold price”? What happens to the share if gold does nothing for a year?
A senior producer with several mines can lose a pit and still pour. A single-asset miner cannot. A company that has already told you the ounces will be lower is not a secret. The cut is the news. The rally case has to survive the cut. A company that is still hedging a slice of production has sold some of the upside the central-bank story is supposed to deliver. You are not paid the full spot on a hedged ounce. You are paid the hedge.
None of this is a list of buys. A gold mining stock can be cheap against a $5,400 target and expensive against $4,125, which is the price you can get today. Analysts who anchor on the target are valuing a year that has not happened. The share does not settle in 2027. It settles on the next trade. If you need Goldman’s end-2027 number to make the stock work, you are underwriting the bank’s note, not the mine.
The forecast is a path, and the path has already been cut
Keep the Goldman Sachs gold forecast in its own box, away from the 79 tonnes. The September box said $4,650 by the end of 2026 and $5,400 by the end of 2027. The October 19 box, as reported, said 79 tonnes in August, 25 of them China on the bank’s estimate, 44 tonnes in the bank’s July estimate, and a structural bid. The public recap of the October note did not restate $5,400. Do not staple the two notes together and call it one call. A bank can believe in the tonnes and still have cut the near target. Goldman did that in September. The cut is the honest part. The September note, as reported, said higher rates were already showing up in softer exchange-traded-fund demand, and that the year-end fair value had to come down even if the later target stayed.
That is also the answer to anyone who treats every official-buying headline as a gold price rally in waiting. The same firm that tracks the tonnes reduced what it thought gold was worth at the end of this year. The destination and the next quarter are different objects. Investors who only hear the destination will be surprised by the quarter. Investors who only hear the quarter will call the official bid a failure every time the price dips. Both will be using half the note.
A fair reading is this. Central bank gold purchases, if they are truly new and if they stay near the pace Goldman assumes, are a reason the bank did not throw out $5,400. They are not a reason the price must leave $4,125 this month. The metal can grind. It can also drop through a level everyone called support, while the official buyers keep to their own schedule. Their schedule is not your stop-loss.
What would actually count as a trigger
If you want a test, make it small enough to fail.
First, the next reported month. If China’s official add slows, and the Council’s net figure fades toward the old 17-tonne world, the “structural” bid is weaker than the August headline. If the reported net figure holds near the summer pace, the bid is at least not a one-month story. You do not need a new adjective. You need the next row of the table.
Second, the next revision of the estimate. Goldman has said its gauge can be revised. If 79 becomes 50 once the late reports land, the headline aged badly. If 79 holds or rises, the gap with the Council is a standing fact about hidden or delayed buying, not a one-off. Until the revision, the gap is a hypothesis with a number on it.
Third, the fund tape. Official buying supports the price most when it is not the only bid. If funds keep shedding metal, the official tonnes are a cushion, not a rally. If funds turn and the official bid is still there, then you have two bids. Two bids are how a gold price rally actually starts. One bid is how a decline slows down. Slowing is not the same event.
Fourth, the company’s own numbers. A miner that raises cost guidance in the same season the gold price firms has handed the margin back. A miner that hits its ounces at a steady cost keeps the margin. The central bank did not do either of those things. The mine did. Gold stocks to watch are the ones that just told you the truth about cost and ounces, not the ones attached to the loudest macro paragraph.
Fifth, the city test. If the next “buying” story is a reserve that sold in one vault and bought in another, put it in the Dutch column. Liquidity and crisis preparedness are real policies. They are not incremental demand. The price cares about the net ounce removed from the market. A bar that leaves New York and appears in London is still one bar.
The sanction memory, and its limit
The reason officials give, and the reason Goldman gives for them, is reserve safety. After Russian reserves held in Western currencies were frozen, gold that sits in a bar looked harder to seize than a deposit that sits in a correspondent bank. That lesson is now old enough to be a habit. Emerging-market central banks in particular have been adding. The Council’s year-to-date leaders, Poland and China, fit the habit. So do Uzbekistan and Kazakhstan. The habit is not universal. Russia itself has been a reported seller in 2026. A seller with 2,271 tonnes is not a small footnote. It is a reminder that official holders also reduce.
Diversification away from the dollar is a motive, not a schedule. A motive can run for years and still pause for a quarter. It can also be used to explain any month, which makes it a weak explanation of one month. August does not prove the post-2022 regime. The string of months might. One string is China’s reported streak, 22 months through August and, on the later recap, 23 through September. Another string is Poland’s march toward 700 tonnes. A third is the gap between 17 tonnes a month before 2022 and whatever you believe the pace is now. If you believe the Council’s reported net, the pace is higher than 17 but nowhere near 90. If you believe Goldman’s estimate, the regime shift is still on. You are allowed to say you do not know which description is complete. That is the accurate position until the revisions land.
A share is not a reserve
Do the month in ounces, once, so the dollars stop floating. One tonne is about 32,151 troy ounces. At $4,125, that tonne is about $133 million. The Council’s 39 reported net tonnes are about 1.25 million ounces, or roughly $5.2 billion. Goldman’s 79 estimated tonnes are about 2.54 million ounces, or roughly $10.5 billion. The pre-2022 pace of 17 tonnes is about 547,000 ounces, or roughly $2.3 billion at today’s price. The step-up everyone is arguing about is the difference between those piles, not a new law of the market.
Annualize them and the argument gets loud, which is a reason to be careful. Twelve months of 39 tonnes is about 468 tonnes. Twelve months of 79 tonnes is about 948 tonnes. Twelve months of the 90-tonne pace in the September forecast is about 1,080 tonnes. The Council’s actual reported net for eight months of 2026 was 170 tonnes. Run that rate for a year and you get about 255 tonnes. The gap between 255 and 1,080 is not a rounding error. It is the difference between a firm bid and the bid the $5,400 target is leaning on. If you build a gold price forecast on 1,080 tonnes of official demand, and the year delivers something closer to the reported run rate, the forecast has a hole in it the size of the assumption.
None of those annual figures is a prediction. They are the monthly numbers times twelve, so you can see the scale. Goldman’s own series already moved from an estimated 44 tonnes in July to an estimated 79 in August. A year made of different months will not look like August copied twelve times. Sellers are in the tape too. Russia’s reported 56-tonne year-to-date reduction and Turkey’s 82-tonne year-to-date net sales are not trivia next to Poland’s 98 and China’s 80. The net is what is left after the official sellers. A bull case that quotes only the buyers is a gross case wearing a net costume.
That costume is how a gold price rally gets promised early. A reader sees 79 tonnes, multiplies by a gold price, multiplies by twelve, and feels a shortage. The shortage is conditional. It appears if the estimate is truly new demand, if it repeats, and if the rest of the market is not supplying the ounces by selling coins, funds, or scrap. Mine supply does not vanish because a central bank had a busy August. Recycled metal comes out when the price has been high, which it was. The official bid is one pipe. It is not the only pipe into or out of the stock.
For a mining share, translate the tonne back into a margin and stop. Suppose, only as arithmetic, that the entire 40-tonne gap between 39 and 79 were real new demand and that it lifted the gold price by some amount. The share still does not receive the tonne. It receives the change in price, minus any change in cost, on the ounces it actually sells. A company guiding fewer ounces has already told you the pipe is narrower. A company whose costs are rising has told you the margin is thinner than the headline price. Central bank gold demand cannot repair either sentence. It can only change the price those sentences are multiplied by.
There is a clean way to hold the official bid without turning it into a slogan. Write four lines on a card, and do not add a fifth until the next print.
Line one is the reported net. For August that line reads 39 tonnes, from the World Gold Council on October 6, with China at 20, Poland and Uzbekistan at 8 each, Kazakhstan at 7, and named sellers on the other side of the net. Line two is the estimate. It reads 79 tonnes, Goldman Sachs, note of October 19, built from flows, with China inferred at 25 against a reported 20, and with an explicit warning that the gauge can be revised. Line three is the relocation. It reads 86 tonnes, De Nederlandsche Bank, announced September 2, of which about 59 were sold in New York and replaced in London. Line three does not add to line one or line two. Line four is the price you can actually trade, about $4,125 on October 20, not the $4,650 year-end fair value and not the $5,400 end-2027 target.
If a later note changes line two, change the card. If China’s next reported month changes line one, change the card. If someone describes a shipment as a purchase, it goes on line three or it does not go on the card at all. A gold mining stock does not belong on this card. It belongs on a second card, the one with cost, guided ounces, hedges, and the cash left after those three. Mixing the cards is how August becomes a reason to buy a mine the buyer has not read.
The second card is where Canadian gold stocks and other TSX gold stocks either earn a place or do not. A senior producer can absorb a bad month at one pit. A single-asset company cannot. A low-cost ounce keeps more of a rally. A high-cost ounce needs the rally just to stand still. A hedge gives away the part of the price the official bid is supposed to deliver. None of those facts is improved by 79 tonnes in a vault you do not own. The tonnes can change the gold price. The second card decides whether the share is paid.
Pull the pieces into one place, because the headline wants a yes or a no.
Central bank gold buying can support the gold price when it is new metal, held back from the market, and repeated. Central bank gold purchases of the Dutch kind do not. They relocate. Goldman’s 79 tonnes is an estimate from flows, led, on the bank’s figures, by an inferred 25 tonnes in China against a reported 20. The Council’s 39 tonnes is the net of the reports, in a month when China, Poland, Uzbekistan, and Kazakhstan did most of the disclosed buying and Russia and others sold. The Goldman Sachs gold forecast of $4,650 and then $5,400 is a September path, not a receipt for August. Spot near $4,125 has not confirmed the next gold price rally. It has confirmed that the prior one can give back a quarter and still leave officials buying.
Gold mining stocks are a bet on the margin, not on the tonne. Canadian gold stocks and the wider TSX gold stocks will rise if the price rises and the ounces arrive at a cost that leaves cash. They will not rise, as a group, just because a note said 79. Some will rise for reasons that have nothing to do with August, a drill hole or a permit or a stupidly low cost. Some will fall in a rally because the cost moved faster than the gold. The sector is a crowd. The share is a single set of books.
Are gold stocks set for another rally? They are set for whatever the next price and the next cost report say. The 79 tonnes did not set them. The 39 tonnes did not set them. The 86 Dutch tonnes set a new address for bars the Netherlands already owned. If you want the official bid in your thesis, keep it as a bid, sized at the number you can defend, and do not let it do the work of a mine. If you cannot say whether you believe 39 or 79, you do not yet have a view on central bank gold demand. You have two headlines. Pick the one whose method you can explain. Then leave the share price to the ounces.