Neither, until you say what the rally is supposed to pay you. A gold-backed ETF that holds bars will rise and fall with the gold price, minus a fee. A gold mining stock will rise and fall with the gap between that price and the cost of getting an ounce out of the ground. Those are not two ways to own the same thing. They are two different claims. One comparison, as of October 9, 2026, makes the split obvious. Over the full period since 2006, a large bullion fund compounded at about 9 percent a year and a large miners fund at about 5 percent. Over the last three years, the miners fund compounded faster. The next gold price rally does not get to be both winners. You have to pick the claim. This is not a recommendation to buy or sell a fund, a miner, a royalty, or a bar.
Here is the idea, and it is the only one. The ETF owns the ounce. The miner owns the gap. Best way to invest in gold is a phrase that hides that choice. If you wanted the price of gold, you already have an answer, and the rest of the work is the fee, the spread, and whether the shares trade near the metal. If you wanted a business that makes more money when the price rises, you wanted the gap, and the rest of the work is the cost, the country, and the new shares the company may print. Gold stocks vs physical gold is not a contest with one champion. It is a pair of jobs. People who want one job and buy the other tool will call the tool a disappointment. The tool did its job.
What the windows actually did
Use one dated comparison, and do not sand the edges. A Quantlake table of adjusted closes, as of October 9, 2026, put SPDR Gold Shares, the bullion fund known as GLD, next to VanEck Gold Miners, the equity fund known as GDX. One week: GLD up about 1.2 percent, GDX up about 1.7 percent. One month: GLD down about 3 percent, GDX down about 7 percent. Three months: GLD up about 4.8 percent, GDX up about 21.7 percent. One year: GLD up about 3.3 percent, GDX up about 13.6 percent. Three years, annualized: GLD about 31.6 percent, GDX about 52.1 percent. Five years, annualized: GLD about 18.4 percent, GDX about 26.4 percent. Since 2006: GLD about 9.1 percent a year, GDX about 5.2 percent.
Read that as a sequence, not as a winner. In the last month, both fell, and the miner fell harder. In the last three months, the miner rose much more. Over two decades, the ounce won. That is how gold investment performance actually looks. The exciting window and the long window disagree. Anyone selling you "better exposure" from a single window is selling you the window.
The risk lines match the gap. Over three years, the same table put GLD's volatility near 21 percent and GDX's near 40 percent. The max drawdown in that span was about 26 percent for the bullion fund and about 39 percent for the miners fund. GDX's beta to GLD was about 1.55. The daily correlation was about 0.84. They move together. They do not move by the same amount. A 1.55 beta means the miners fund has, in that sample, swung about one and a half times the bullion fund. It does not mean the next month will be 1.55 times. Beta is a history. Gold price volatility in the metal is already large. In the equities it is larger, because a small change in the price is a large change in what is left after costs.
Both were below their own intermediate averages in that snapshot. GLD was about 22 percent under its January 29, 2026 high. GDX was about 23 percent under its February 27, 2026 high. The rally people are asking about is a rally off a drawdown, not a rally off a fresh high. Gold in this week's other reports has been a little under $4,200, after a January peak that took the metal far above $5,000. A fund 22 percent under its high can still be a large winner over three years. Those two facts live in the same chart. Do not let one erase the other.
The fee is real, and it is the small argument
GLD's sponsor fee is 0.40 percent a year. Fund pages print that figure. One comparison site prints 0.42 percent. GDX's net expense ratio is about 0.51 percent on fund-data pages, and 0.50 percent on the comparison above. On $10,000, 0.40 percent is $40 a year. Half a percent is $50. The gap between the two fees is about $10 a year on that sum. Ten dollars did not produce a 21 percent three-month gap, or a 39 percent drawdown, or a two-decade shortfall. Gold ETF expense ratio debates matter when you are choosing between two bullion funds. They are a rounding error when you are choosing between a bar and a mine.
Inside the bullion aisle, the fee does matter. GLD's cheaper sibling, GLDM, is listed at 0.10 percent. That is $10 a year on $10,000, against $40 for GLD. Over ten years, ignoring compounding, that is $100 against $400. Compounding makes the gap a bit wider. GLD is the deep, old, huge fund. One market page put its assets near $142 billion in October 2026. You pay the extra fee for the size and the habit of the market. You do not pay it for a purer ounce. Both funds are trying to own gold bars. Investing in gold ETFs starts with that sentence. The ounce is the holding. The fee is the leak. A gold ETF investment that ignores the leak will still track the metal for a year. Over a working life, the leak is part of the result.
GDX also pays a dividend. The comparison put the trailing yield near 0.7 percent. Another page has printed a higher trailing figure. Use the small one and the point still holds. The bullion fund pays you nothing, because a bar pays nothing. The miners fund pays you a slice of what the companies pay. Interest rates and gold meet here. A higher real yield is a higher coupon you skip when you hold the bar. It is also a higher discount rate on a mine's future ounces, and a reason the multiple on gold mining equities can fall even while the companies print cash. Both tools feel the yield. The ETF feels it as an opportunity cost with no offset. The miner feels it as a multiple, with earnings as a partial offset. Neither feeling is a promise that the yield will fall.
How a rally pays the ETF
How gold ETFs perform when gold prices rise is the clean case. If the metal rises 10 percent and the fund holds bars, the shares should rise about 10 percent, minus the fee for the period, and plus or minus a small gap between the share price and the value of the metal. That gap is usually tight in a huge fund and can widen in a scare. You do not get the miner's torque. You do not get the miner's strike, the flooded heading, or the surprise tax. You get the quote. Gold-backed ETFs are a claim on metal, not a claim on a management team. That is the feature. People who are bored by it wanted a different product.
The same cleanliness works on the way down. A 10 percent fall in gold is about a 10 percent fall in the bullion fund. There is no cost gap to cushion you, because you do not have costs. You have a bar. In the last month of that October 9 table, GLD was down about 3 percent and GDX about 7. The bar lost less. In a gold market that can still test lower prices if real yields stay high, losing less is a form of exposure too. Exposure is not only the size of the gain. It is the size of the loss you agreed to sit through.
Central bank gold buying does not change this math. Official buying can lift the ounce, and the ETF will follow the ounce. It will not follow it extra. A reserve manager adding tonnes is a reason the quote might rise. It is not a reason your bullion shares will beat the quote. Gold market trends that start in a vault and end in a futures pit still arrive at the ETF as a price. Gold portfolio diversification, if that is the job, is what the bar is for. A slice of metal that does not share the stock market's full mood. The comparison put GLD's correlation to the S&P 500 fund near 0.18 over three years, and GDX's near 0.32. Lower is not zero. In a panic, correlations jump. A diversifier that jumps with everything, on the worst day, was never a vault. It was a ticker.
How a rally pays the miner
A miner does not own the ounce in the vault. It owns the right to produce an ounce at a cost. Take a sustaining cost of $1,500, which sits inside more than one large producer's 2026 band, and a gold price near $4,200. The rough gap is about $2,700 an ounce. If gold rises 10 percent, to about $4,620, the gap becomes about $3,120. The price rose 10 percent. The gap rose about 16 percent. That is the torque. If gold falls 10 percent, to about $3,780, the gap falls to about $2,280, a decline of about 16 percent. The stock can move more than that, because the market prices a stream of years, not one year, and because it changes its mind about the multiple. Gold stock valuation is the gap times a mood. The mood is not in the cost guide.
The torque is larger when the cost is closer to the price, and smaller when the cost is far below it. A mine at $1,150 of sustaining cost has a fatter gap. A 10 percent gold move is a smaller percent of that fat gap. A mine at $1,850 has a thinner gap. The same 10 percent is a larger percent of what remains. High-cost mines are not "better exposure" in the sense of a better business. They are larger swings. Low-cost mines are better businesses and duller swings. People who want fireworks often buy the worse business. People who want the business often complain that the stock did not double. Both complaints are the idea, refused.
The swing can also be zero. If the mill stops, the gap is a gap on no ounces. A rally in the quote does not put rock on the conveyor. Investing in gold mining stocks is a bet that the ounces will be produced, that the cost will be near the guide, and that the country will allow the work, and only then a bet on the gold price. Skip the first three and you do not have exposure to the rally. You have a story about it.
Five gaps, not five tips
The names below are not a list of best Canadian gold stocks. They are five published gaps, so the word exposure has to become a number. Canadian gold producers are a large part of the miners fund already. They are not a separate miracle.
Agnico Eagle has guided 2026 payable gold of 3.3 million to 3.5 million ounces, and all-in sustaining costs of $1,400 to $1,550. The second quarter was 855,816 ounces at a sustaining cost of $1,459, sold near $4,483. At a spot price a little under $4,200, the gap is still more than $2,500 an ounce. A 10 percent rally in gold widens a gap that was already wide. It does not repair a ground issue the company has flagged. The ounces have to show up. As of a June 30, 2026 holdings list, Agnico was about 10.7 percent of GDX, the largest weight. When you buy the miners fund, you are already buying a large slice of this gap, plus the other gaps the fund holds.
Lundin Gold's year guide is $1,110 to $1,170 of sustaining cost, on 475,000 to 525,000 ounces. The second quarter printed $1,176. The third quarter was a record 153,736 ounces, at 10.1 grams per tonne. The gap per ounce is the fattest of this set. The map is one mine, Fruta del Norte, in Ecuador. A rally pays this gap almost in full, per ounce, and pays nothing if the mill stops. Country risk is not in the gold price. It is in the multiple people put on the gap.
Wheaton Precious Metals does not dig. In the second quarter of 2026, the cash cost was $568 per gold-equivalent ounce, on 209,115 equivalent ounces, and revenue was $929 million. The 2026 guide is 400,000 to 430,000 ounces of gold and 27 million to 29 million ounces of silver. Net debt at June 30 was about $1.87 billion. A rally in gold lifts the gold half. A rally in silver lifts the other half. They do not have to arrive together. The debt stays if they do not. Wheaton was about 5.6 percent of GDX on that June list. A stream is in the miners fund. It is not a mine.
Alamos Gold cut 2026 production to 510,000 to 560,000 ounces, from 570,000 to 650,000, after a seismic event at Young-Davidson. It lifted sustaining costs to $1,775 to $1,875, from $1,500 to $1,600. The gap is still wide at $4,200 gold, and it is thinner than Agnico's. Thinner means a 10 percent gold move is a bigger percent of the margin. That is more exposure to the next rally, and more exposure to the next fade. Torque is not quality. The rocks in Canada and Mexico did not change their address. The year's math did.
Franco-Nevada is the royalty. Second-quarter revenue was $580.9 million, on 132,405 gold-equivalent ounces. The 2026 guide is 510,000 to 570,000 equivalent ounces. A royalty's revenue moves close to the price, on the ounces delivered, not like a thin margin. It was about 4.9 percent of GDX. If you wanted the ounce, the royalty is a step away, because an operator can miss. If you wanted the wildest gap, the royalty will bore you. Boredom is the product.
The basket is a third tool
GDX is not "the miners" as a pure idea. It is a fee on a list. On the June 30, 2026 holdings page used here, the top two names, Agnico and Newmont, were each a bit over 10 percent. Barrick was about 8 percent. Wheaton, AngloGold, Franco-Nevada, Kinross, and Gold Fields followed. The top ten were about 59 percent of the fund. You are not buying a hundred equal bets. You are buying a handful of large gaps, several of them Canadian, and a tail. Gold mining equities in a basket will not track your favorite junior. They will track the large weights.
That concentration is a kindness and a limit. A kindness, because the large names publish costs and produce ounces. A limit, because a discovery on a side street will not move the fund. Junior gold mining stocks live in a different fund and a different failure mode. One data page, as of October 12, 2026, showed the VanEck junior miners fund up a fraction of a percent for the year, while the senior miners fund was up a few percent. Vendor year-to-date figures disagree by small amounts. The direction of the lesson does not. The junior basket is not a louder version of the senior basket. It is a basket of stories, many of them not yet mines. TSXV gold stocks inside that world can double on a hole and halve on a financing. A bullion ETF cannot do either. Calling all three "gold exposure" is how a portfolio becomes a surprise.
Gold exploration companies are the far end. They have no gap, because they sell no ounces. They have a treasury and a calendar. A gold price rally helps them raise money and helps the story. It does not become revenue this quarter. If you buy an explorer for the next rally in the quote, you are early by a mine. The rally you will actually feel is the rally in the appetite for paper. That appetite dies faster than the gold price. Canadian gold mining stocks that already pour metal are a different instrument from a drill program. Do not average them.
Which one, then?
Which offers better exposure to the next gold rally depends on the rally you mean. If you mean a higher quote, the bullion ETF is the cleaner exposure. It will pay you a percent close to the metal's percent. It will also charge you the percent on the way down. The fee is the part you can control. A 0.40 percent fund and a 0.10 percent fund own the same idea. The cheaper one keeps more of a flat year. In a year the metal rises 20 percent, the fee is a small bite. In a year the metal is flat, the fee is the whole result, and it is a loss.
If you mean a wider gap, the miner is the exposure, and only the miner whose cost you have read. A low-cost producer gives you a smaller percent change and a thicker cushion if the rally fails. A high-cost producer gives you a larger percent change and a thinner cushion. A stream gives you the price with a contract cost and a debt line. A royalty gives you the price with someone else's operating risk. A junior gives you the calendar. Gold investment options are this list. They are not a single "gold" bucket with a best door.
If you mean you do not want to pick a mine, the miners fund is the compromise. You pay about half a percent. You accept a beta that has been near 1.5 to the bullion fund. You accept that a bad month in the metal can be a worse month in the stocks, as the last month in that table was. You also accept that a good quarter in the gap can be a much better quarter in the stocks, as the last three months were. Precious metals market outlook language often skips this and says "miners give you leverage." Leverage is the unsigned word for a bigger swing in both directions. Signed, it is a tool. Unsigned, it is a slogan.
The long window is the warning against slogans. Since 2006, the ounce compounded faster than the miners fund. Costs rose. Mines disappointed. Shares were issued. The gap did not stay as wide as the gold chart suggested, for long enough, often enough. The last three years went the other way, because the price ran far ahead of costs and cash finally showed up. Gold bull market chapters do that. They do not repeal the prior chapters. A gold price forecast 2026 that has the metal marked higher, as more than one bank does, is a forecast about the quote. It becomes a forecast about miners only after you subtract the cost inflation those same banks are starting to worry about. UBS, earlier this month, still had a path that allowed a dip toward $4,000 before a later climb. A dip to $4,000 narrows every gap on this page and does not close the low-cost ones. It hits the thinner gaps harder. The ETF just falls by about the same percent as the metal.
A portfolio percent is not a slogan
Say you put 5 percent of a portfolio in gold, because someone called 5 percent a hedge. Five percent in the bullion fund is about 5 percent of the gold move, minus the fee. If gold rises 10 percent, that slice adds about half a percent to the whole portfolio. If gold falls 10 percent, the portfolio gives back about half a percent. You can see it. You can size it. Gold portfolio diversification, done this way, is a small bet on the quote. It will not save a portfolio that is 95 percent something else. It will also not wreck it in an ordinary month.
Five percent in the miners fund is a different bet. With a beta near 1.5, a 10 percent gold move has, in the recent sample, lined up with something closer to a 15 percent move in the miners, sometimes more, sometimes less. On a 5 percent weight, 15 percent is about three-quarters of a percent on the whole portfolio. The extra quarter of a percent is the torque you came for. The same extra arrives, with the other sign, when gold falls. The last month in the table was the other sign. GDX down about 7 percent against GLD down about 3. On a 5 percent weight, that is a bit more pain than the bar, and it is still not a portfolio event. People who feel it as a portfolio event usually own more than 5 percent, or they own a single mine and called it a diversifier.
A single mine can be a portfolio event. Lundin's gap is wide, and the mine is one place. A stoppage does not cost you 5 percent of the gold move. It can cost you a large part of that position, while gold itself does nothing wrong. Alamos already showed a smaller version of this. A seismic event cut the year's ounces and raised the year's cost. The gold price did not do that. The ground did. An ETF of bars cannot have a seismic event. A fund of many miners can have one inside a single weight, and the weight is capped by the index. That cap is the reason the basket exists. It is also the reason the basket will not feel like your best idea on the best day.
So the practical test is three lines, written down before the rally. What percent of the portfolio. What you want that percent to do if gold rises 10 percent. What you can stand if gold falls 10 percent, or if one mine stops. If the answers are "about the same percent as gold" and "I do not want a mine," the ETF is the tool, and the next argument is 0.40 percent versus 0.10 percent. If the answers are "more than gold" and "I have read the cost," a producer or the miners fund is the tool, and the next argument is which gap. If the answers are "a double" and "I will not read a balance sheet," you are not choosing exposure. You are choosing a story. Stories are allowed. They are not a hedge.
What you can underwrite
You can underwrite the split. The ETF owns the ounce. The miner owns the gap. As of the October 9, 2026 comparison, the bullion fund had won since 2006, at about 9 percent a year against about 5 percent, and the miners fund had won the last three years, at about 52 percent a year against about 32 percent. One month, the miners fell harder. Three months, they rose much more. Volatility was about 21 percent against about 40 percent. That is the record. It is not a schedule for the next rally.
You can underwrite the fee as a bullion problem, not as the reason to pick a miner. About 0.40 percent for the large bullion fund. About 0.10 percent for its cheaper sibling. About 0.50 percent for the miners fund. On $10,000, those are tens of dollars a year. The drawdown difference is thousands. Do not choose the 39 percent swing to save $10 of fee.
You can underwrite a business only from its cost. Agnico's sustaining band is $1,400 to $1,550. Lundin's is $1,110 to $1,170, with one mine in Ecuador and a record quarter of 153,736 ounces. Wheaton's cash cost was $568 per equivalent ounce, with about $1.87 billion of net debt and a large silver book. Alamos's new band is $1,775 to $1,875, on a cut to 510,000 to 560,000 ounces. Franco-Nevada did $580.9 million of quarterly revenue on contracts, not on a shovel. A rally pays these five in different sizes. GDX already holds several of them in size. Buying them again on top of the fund is a bigger bet than the label says.
You can underwrite the job. Gold investment opportunities that are really a diversifier belong in the ounce, where the correlation to stocks has been lower and the drawdown smaller. Jobs that are really a bet on margins belong in a cost you have read. Jobs that are really a bet on a drill belong in a treasury and a date, and they are not a gold ETF in disguise. Gold price outlook 2026 can be right about the metal and wrong about your tool. The tool is the part you choose.
What would make this reading wrong
The reading is wrong if the next rally is so large, and costs stay so still, that the gap compounds faster than the ounce for as long as the ounce did after 2006. That has happened in chapters. It has not been the full-period result in the table above. If costs truly stay flat while gold adds another thousand dollars, the miner's percent gain on the gap will keep outrunning the ETF. Then "the ounce won the long game" will be a fact about the past, not a fact about the path in front of you. You would still need the ounces to be produced. A perfect gold price and a stopped mill is a perfect price for someone else.
The reading is wrong if the bullion fund's share price stops tracking the bar. A large premium or discount, a failure to get metal, a tax change, or a panic in the share count can make the ETF a worse ounce than the ounce. Those are real risks. They are not the usual day. On the usual day, the leak is the fee. If your reason for avoiding the ETF is a fear of the usual day, look at the fee and the asset size before you invent a tracking horror.
The reading is wrong if you needed income and treated the bar as a bond. It is not a bond. The miners fund's yield, under 1 percent in one print, is not a bond either. Interest rates and gold will keep arguing. A higher real yield raises the cost of holding the bar and can compress the miner's multiple at the same time. If yields fall, both can rise, and the miner can rise more. That is the beta you already saw. It is not a new idea. It is the same gap, with the discount rate moving.
The reading is wrong if you pick the three-year window because it flatters the bet you already wanted. The one-month window flatters the bar. The two-decade window flatters the bar. The three-year window flatters the miner. Gold market outlook 2026 pieces that show you only the flattering window are editing the table. The whole table is the argument. Take the window that matches your horizon, and then take the one that does not, so you can see the cost of being early or late.
The reading is also wrong if a single-stock disaster, or a single-stock discovery, is what you meant by exposure. The ETF will not discover a mine. The miners fund will barely notice a small one. The junior might, and it might issue shares until the discovery is a smaller piece of a larger company. Better exposure, in that case, was never the fund. It was a bet you should size as a bet, not as a substitute for gold.
The idea, once
A bullion ETF owns the ounce. A miner owns the gap between the gold price and the cost. As of October 9, 2026, the large bullion fund had compounded at about 9 percent a year since 2006, and the large miners fund at about 5 percent. Over three years the miners fund had compounded faster, near 52 percent against near 32 percent, with volatility near 40 percent against near 21 percent, and a beta near 1.55. The fee difference is about $10 a year on $10,000. The drawdown difference is not. Agnico, Lundin, Wheaton, Alamos, and Franco-Nevada are five different gaps, and the miners fund already concentrates in names like them. The next rally pays the tool you actually hold. If you wanted the quote, buy the ounce and argue about 0.40 percent versus 0.10 percent. If you wanted the gap, read the cost. Do not call those the same exposure.
A note on sources and limits
The head-to-head returns, volatility, drawdowns, beta, correlation, yield, and the distances below the 2026 highs are from a Quantlake comparison of GLD and GDX using adjusted closes as of October 9, 2026, published October 12. Other vendors print different year-to-date and one-year figures for the same funds. The disagreement is a reason to treat any one percent as a snapshot, not as a law. GLD's 0.40 percent fee, its asset size near $142 billion, and the 0.10 percent fee on GLDM are from fund and market pages in October 2026, including etf.net. GDX's expense ratio near 0.51 percent, and the June 30, 2026 top holdings and weights, are from fund-data pages including a PlainIndex holdings list. Company operating figures are from those companies' 2026 guides and quarterly releases: Agnico's production and cost guides and second-quarter results, Lundin's second-quarter cost, third-quarter production, and guide, Wheaton's second-quarter revenue, cost, guides, and June 30 net debt, Alamos's July 29, 2026 guidance cut, and Franco-Nevada's August 2026 quarter. The UBS path that allows a move toward $4,000 is from accounts of an early-October note, used only to show that a lower quote narrows gaps. Nothing here is investment advice, a solicitation, or a personal recommendation. Funds and stocks can fall. Fees, holdings, guides, and tax treatment change. Past windows do not order the next one. You can lose money, including a large loss on a miner or a junior. Do your own work.