Gold prices today are lower again, and the reason is not a mystery. Spot gold fell 1.6 percent to $4,096.13 an ounce by 9:20 a.m. Eastern time on Wednesday, October 7, 2026. That was the lowest print since August 5. December futures fell 1.6 percent to $4,121.70. The dollar index was up 0.7 percent. The 10-year Treasury yield was at a high not seen in more than twenty years. The minutes of the Federal Reserve's September meeting were due at 2 p.m. Eastern. The metal did not wait for them.
Here is the only idea that matters. The Federal Reserve minutes will tell you how the committee talked in September. The gold price will tell you whether that talk still has a bid. A hawkish paragraph can push a test. It cannot draw the lines. The lines that change this gold correction are already on the chart. One is the shelf near $4,100 that failed during the morning. One is $4,000, and the June and July troughs just under it. One is $4,225, the pivot a major bank says gold must retake before the decline is allowed to stop. Above that sit the moving averages the metal has lost. Watch the breaks. Do not watch the adjectives.
Nothing in this piece is a recommendation to buy or sell gold, a gold fund, or a gold mining stock. Levels cited below are other people's maps. A map is not an order.
What the tape did before the minutes
The day was not one price. It was a slide. Earlier, around 6:27 a.m. Greenwich time, Reuters had spot gold down 0.8 percent at $4,130.37, with U.S. futures at $4,157. A retail snapshot near 6:41 a.m. Eastern had the price around $4,138 after December futures opened at $4,195. By midmorning in New York the spot market had gone through $4,100 and printed $4,096. The decline extended. It did not reverse into the minutes.
That path matters more than the round number. Gold had been living in a rough range near $4,100 to $4,200 for about two weeks, after a rebound off a low near $4,100. On Tuesday it had bounced. On Wednesday it gave the bounce back and then gave the floor a shove. FXStreet described the metal as a sell-on-bounce trade while it stayed under its main averages. Société Générale said the pullback was drifting toward an interim level near $4,095. The morning tape agreed with the map. Agreement is not fate. It is the condition the minutes walked into.
Gold prices today are still high against any year before this one. They are not high against this year. London spot set a record above $5,500 in January. The late-summer area was near $4,700. September 28 was near $4,110. A print of $4,096 is a new low for the pullback, not a new low for the decade. The gold correction is a drawdown inside a year that already made a record. Both facts can be true. Investors who keep only one of them will misread the minutes.
Why are gold prices declining
Why is gold price falling on a day when nothing in the mine supply changed? Because gold pays no interest, and the things that compete with it paid more this morning.
The dollar rose. A stronger dollar makes an ounce priced in dollars more expensive for a buyer who earns euros, yen, or rupees. That buyer steps back. The price in dollars slips. This is not a theory of empire. It is the arithmetic of one quote.
Yields rose with it. Reuters put the 10-year at a more-than-twenty-year high on Wednesday morning. The higher the yield, the higher the pay for holding a Treasury instead of a bar. Gold investment in the metal is a bet that the bar will do a job cash cannot do. When cash pays a large real return, that job gets harder to justify on a one-month horizon. It does not get harder on a ten-year horizon if the buyer is a central bank. Those are different buyers. The morning tape was the first kind.
Fed interest rates are the root of both moves. The Federal Reserve raised the policy range in September for the first time in more than three years, to 3.75 to 4.00 percent. Officials have not sounded finished. San Francisco Fed President Mary Daly said further hikes depend on whether the forces lifting inflation fade or persist. Kansas City Fed President Jeff Schmid said rates still need to rise to bring inflation down. Markets heard them. By 9:20 a.m. Eastern, the CME FedWatch tool implied an 84 percent chance of a hike in December, Reuters reported. Earlier the same morning a separate snapshot had December closer to 69 percent, with more than 78 percent odds that the October 28 meeting would be a hold. The odds moved during the day. That movement is the story. A live repricing of Fed rate expectations is a direct tax on a non-yielding metal.
Oil added a side wind. Prices rose on Middle East supply risk and on a storm heading for U.S. producing regions. Dearer oil can mean stickier inflation. Stickier inflation, in this Fed's language, means the hiking talk stays alive. Gold can rise with oil when the fear is war and scarce fuel. It can fall with oil's inflation threat when the fear is the next rate move. Wednesday was the second kind.
None of this repeals the longer bid. China's central bank added to its gold reserves again, Reuters noted, without the early wire pinning a tonne count to the morning. Central banks do not trade the minutes. They trade a decade. The gold price outlook for an afternoon and the gold price outlook for 2027 are allowed to point in different directions. The error is forcing them to be one sentence.
What the Federal Reserve minutes can change
The Federal Reserve minutes are a record of a meeting that already happened. The September decision is not a secret. The hike is in the rate. What the minutes can change is the market's guess about how many officials wanted to go further, how many wanted to wait, and which risks they ranked first. Frank Walbaum at Naga, speaking to Reuters before the release, said the minutes would show the degree of support for more increases and could reshape hike odds. Moves in yields, the dollar, or oil would then amplify whatever the minutes did to gold. That is a fair description of a transmission. It is not a forecast of the sign.
Read the minutes for three things, and ignore the rest.
First, the balance of the committee. A hike that passed with a crowd of dissents is a different hike from one that passed in a hush. Dissents toward patience would soften December. Dissents toward a larger move would harden it. The gold market has been trading the second risk all morning.
Second, the word on inflation's causes. Daly's public line was that the next step depends on whether the pressures fade. If the minutes show the committee thinks the pressures are sticky, the path stays tight. If they show a committee that hiked once and expects to watch, the path is a pause. Gold does not need the Fed to cut. It needs the Fed to stop adding reasons to hold cash.
Third, anything that surprises the odds already in the price. An 84 percent chance of a December hike is a crowded bet. Minutes that merely confirm it will not do much. Minutes that push the chance toward a sure thing can still hurt, because the last 16 points have a price. Minutes that knock the chance down can lift gold even if the prose is stern. The level of hawkishness is not the variable. The gap between the prose and the odds is the variable.
The minutes cannot do three other things. They cannot repeal January's record or Wednesday's low. They cannot make a moving average move. They cannot tell you whether $4,096 holds. That is gold technical analysis, and it is a separate document from the Fed's.
The rate path under the gold forecast
Fed rate expectations, as of Wednesday morning, were a hold in late October and a hike in December. That is not a promise. It is a probability that moved between breakfast and midmorning. A gold price prediction that treats December as settled will be wrong if the minutes, or the next inflation print, move the odds again. A gold price prediction that ignores December will be wrong if the odds stay high and yields stay at multi-decade highs.
The useful gold price forecast 2026 is a stack, not a point. Goldman Sachs, in notes cited earlier this week, cut its year-end 2026 fair value to $4,650 an ounce from $4,900 and kept a December 2027 figure of $5,400. The bank said central-bank buying did nearly all of the expected gain, and it raised its assumed official demand to about 60 tonnes a month. A separate nowcast had the recent pace nearer 91 tonnes. Keep those apart. Sixty is a model. Ninety-one is a recent run rate. Neither is a bid at 2 p.m.
Société Générale's technical work is the near map, and it is less kind. Gold failed to hold above the 200-day average, which that desk put near $4,510 to $4,540. It was drifting toward $4,095. A failure there, they said, opens $4,000 and the June and July troughs at $3,960 to $3,940, if gold also fails to retake about $4,225. FXStreet's daily chart on Wednesday put the 50-day average near $4,332, the 100-day near $4,268, and the 200-day near $4,531, with the relative strength index around 39. The metal was under all three averages. Momentum was weak but not yet washed out.
Those two documents can both be honest. A model that ends 2026 at $4,650 is a statement about official buying and a slower path. A chart that is under its averages, with $4,095 under pressure, is a statement about the last two months. The gold forecast 2026 that an investor can use is the one that says which document is on the clock. Into the minutes, the chart is on the clock. Into 2027, the official buyer may be. Mixing the clocks is how people buy a correction because they remember a target.
Gold support levels that would change the correction
Support is not a trampoline. It is a price where selling paused before. It fails when the reason for the pause is gone. These are the gold support levels in play on October 7, with the source of each line named, so they are not mistaken for a house view.
The first is the zone around $4,100 to $4,125. Orbex had support at $4,100 to $4,125, then $4,080, then $4,045. A separate intraday map had a band near $4,135 to $4,143 and another near $4,115. Société Générale's interim projection was about $4,095. The morning low at $4,096 sat on that projection. This is the shelf the rebound of the last two weeks was built on. A daily close back above it says the break was a spike. A daily close under it says the range gave way. One print through the level, during a session, is not a close. The minutes land in the same session. Do not let the first hour after 2 p.m. write the verdict.
The second is $4,000, and the summer troughs at $3,960 to $3,940. Société Générale called that band the lower edge of a multi-month range. Several desks, in notes earlier this week, have talked about $4,000 as a floor they would rather not see tested. A floor that everyone can name is a crowded place. It can hold because buyers wait there. It can fail because the buyers already bought, and the next seller does not care about a round number. If $4,095 gives way and $4,225 has not been retaken, this is the zone the bank says comes next. It is a scenario. It is not a destination the metal owes anyone.
The third is the stress case that is not the base case. Bank of America's technicians, in a July note, said a major top in 2026 could leave the second half vulnerable and put about $3,315 in play if the year proved to be a top on the order of 1980 or 2011. That number is a long way from $4,096. It is on the page so the near levels are not mistaken for the worst levels. A market can lose $4,000 without going to $3,315. A market that is still inside a central-bank bid is a poor candidate for a clean replay of 2011. Poor is not impossible. Size the thought, do not marry it.
Gold support levels do not know that the minutes are today. If the minutes are hawkish and the dollar adds another leg, the $4,095 area gets a real test before the close. If the minutes are only as hawkish as the 84 percent odds already say, the test can fade and the range can resume. The level is the event. The prose is the excuse.
Gold resistance levels the bounce has to clear
A decline does not end because it is tired. It ends when price reclaims a level that sellers were defending. These are the gold resistance levels overhead, again with the authors named.
The nearest is the top of the recent range, roughly $4,200 to $4,245. Orbex had resistance near $4,200, then $4,245, then $4,315. Other maps put a weekly pivot near $4,177 and a further resistance near $4,244. Société Générale was blunter. If gold cannot get back through about $4,225, the decline can continue. That single line does more work than a page of oscillators. Below it, bounces are rallies inside a correction. Through it, the burden shifts to the seller.
Above the range sit the averages FXStreet listed. The 100-day near $4,268. The 50-day near $4,332. The 200-day near $4,531, which lines up with the zone Société Générale said gold failed to hold. A gold market outlook that calls the long trend intact has to explain a price under all three. The long trend can still be intact as a year-long fact and broken as a month-long fact. Reclaiming the 50-day would be the first serious repair. Reclaiming the 200-day would be the repair that talks to the January high. Neither is a task for one set of minutes.
There is a reason to be suspicious of the first green hour. FXStreet's framing on Wednesday was sell-on-bounce while the averages remain overhead and the index of momentum sits near 39. A bounce into $4,177 or $4,200 that dies is the pattern. A bounce that closes through $4,225 and holds the next day is a different pattern. Investors who buy the first and call it the second are how resistance gets paid.
Will gold prices rebound
Will gold prices rebound? They can, and they have, inside this same range, as recently as Tuesday. A rebound that stops under $4,225 is a trade. A rebound that retakes the 50-day is a change in the correction. The minutes can start either one. They cannot finish either one.
The rebound case is specific. December hike odds fall because the minutes show a committee less eager than Schmid's public line. The dollar gives back the morning's gain. Yields ease. The close is back above $4,100, and the next day does not break it. China's reserve bid is still there as a background buyer. In that world the gold correction is a pause in a bull market, and the $4,650 year-end figure stops looking far away. It is still a figure, not a promise.
The no-rebound case is just as specific. The minutes match the 84 percent odds or push them higher. The dollar holds the gain. The close is under $4,095. The next support the banks have named is $4,000 and then the summer troughs. Official buying continues and does not care about a hundred dollars, which is why the price can fall while the long story stays constructive. That split was the story of September. It can be the story of October.
Anyone selling a gold price prediction 2026 as a single number is selling a clock they do not own. The honest prediction is a pair. Near term, the correction holds until $4,225 is retaken or $4,095 is lost. Further out, the path depends on whether official demand stays near the tonnes the banks have been counting, and on whether the policy rate stops rising. Both conditions are observable. Neither is due at 2 p.m. in full.
What this does to gold mining stocks
Gold mining stocks are not the gold price. They are the gap between the gold price and the cost of an ounce, multiplied by a balance sheet and a mine plan. When gold falls 1.6 percent in a morning, the shares often fall more. In September, New York gold futures fell 6.4 percent and the large gold miners in a global ranking fell 12.7 percent. Not one of those fifteen names rose. The VanEck Gold Miners fund, GDX, went from about $101 in early September to the high $80s by October 6. That is the multiplier. It works on the way down without asking permission.
A set of hawkish minutes can widen the move. It does not create a new fact about any one mine. Costs, grades, and guidance are the other document. Kinross cut its 2026 and 2027 ounce guidance in late September for reasons that had nothing to do with the Fed. Weather and recoveries. A falling gold price on top of a cut is two problems. A falling gold price at a miner that has not cut is one problem. The minutes will not tell you which miner is which. The filings will.
If gold rebounds through $4,225 and toward the averages, the intact producers can rise faster than the metal. That is the same multiplier in the other direction. It is not a reason to own them into the print. It is a reason the rebound, if it comes, will look like skill and will have been leverage. Investors who wanted the bar and bought the stock in September learned the difference. The lesson did not expire on Tuesday.
There is a second mining point that the minutes ignore. Some producer states are taxing or fencing gold exports, and China is both restricting its own exports and importing a very large tonnage. That is a medium-term supply story. It did not stop Wednesday's decline. A structural bid and a cyclical selloff fit in the same month. Gold mining stocks feel the selloff first, because the shares reprice in minutes and the structural bid arrives as tonnes over quarters.
A level is a question, not a target
Use the levels as questions you can grade after the close, not as prices you must trade.
Did spot close back above $4,100 to $4,125, or under $4,095? The morning low was $4,096. The close is the grade. A wick is a rumor.
Did the bounce, if there is one, fail under $4,225? If it did, Société Générale's condition for a continued decline is still met. The gold correction is intact.
Did December hike odds rise or fall after the minutes? The pre-minutes print Reuters used was 84 percent. The change is the news. The level of the odds is the background.
Did the dollar hold the 0.7 percent gain? Gold's morning was a dollar and yield story. If those reverse and gold does not, something else is wrong. If gold reverses and those do not, the reversal is thin.
Did any official buyer show a number? A reserve addition without tonnes is a headline. Tonnes are the fact. Do not promote the headline into a floor.
Fail two of those and the day was noise. Pass the $4,095 break on a close, with the dollar still bid, and the gold market outlook for the next few weeks shifts toward $4,000. Retake $4,225 and the shift is the other way. The minutes are allowed to be interesting and still not be the decider.
Spot, futures, and the trap of one number
Even the price of gold was not a single number on Wednesday morning. Spot was $4,096 in the later Reuters print. December futures were $4,122. Earlier, futures had opened at $4,195. A buyer who says gold is at $4,100 and a buyer who says gold is at $4,200 can both be quoting a real screen from the same day. The gold price today depends on the contract and the minute. Levels should be tied to the series you will actually grade. If the map was drawn on spot, grade spot. If it was drawn on the December future, grade that. Mixing them makes a false break.
The gap between spot and the active future is usually a cost-of-carry story. It is not a signal by itself. What is a signal is a future that opens firm and a spot market that spends the morning giving the gain back. That is what Wednesday did. Strength at the open was not confirmed. Confirmation would have been a hold above the shelf, not a print at the open. Gold technical analysis that cites only the open will flatter the metal. Analysis that cites only the low will flatter the bear. The close is the compromise the day is required to make.
There is a related trap in the phrase two-month low. The low since August 5 is real on the Reuters comparison. It is also a low inside a year whose high was above $5,500. A two-month low can be the start of a larger break. It can also be the kind of low a range makes before it goes quiet. You cannot know which from the label. You can know which only if $4,095 holds or fails on a close, and if $4,225 is or is not retaken. The label is a headline. The levels are the test.
One more distinction belongs here, because people ask for a gold price prediction as if the metal were a stock with earnings. Gold has no earnings. It has a competing yield, a dollar price, and a set of buyers who do not mark to the same calendar. A prediction that ignores the yield will miss days like Wednesday. A prediction that ignores the central bank will miss years like the one that ran into January. The Federal Reserve minutes speak to the yield. They are silent on the tonnes. Weight them that way. Do not average the two clocks into one gold price prediction 2026. An average of a bad week and a constructive year is not a plan. It is a way to be wrong on both horizons at once.
The correction inside the year
The gold price 2026 has already done the rare thing. It made a record above $5,500 and then gave back a large piece of it. From that high to a morning low near $4,096 is a decline on the order of a quarter, depending on the exact January print you use. From the late-summer area near $4,700 it is about 13 percent. From the end of August, when futures were near $4,441, September alone took 6.4 percent. This is a correction with a shape. It is not a crash that forgot the starting point, and it is not a dip that never broke anything.
Corrections inside bull markets end when the reason for them fades and a broken level is retaken. Corrections that become bear markets end those hopes at a level everyone thought was safe. $4,000 is the level currently nominated for that role. It has not been tested. $4,095 is being tested now. The Federal Reserve minutes sit in the middle of the test like a commentator, not like a player. The players are the dollar, the yield, and the close.
Gold investment that needs a new record this month is in the wrong instrument and the wrong week. Gold investment that can tell a close under $4,095 from a speech about inflation is in the right argument. The metal will choose. The choice will be a price, not a paragraph.
The close
Gold extended its decline on October 7 because the dollar and yields rose into the September minutes, and because the market spent the morning raising the odds of another hike in December. Spot traded down to about $4,096, the lowest since August 5. The point that matters is not a secret meeting. The meeting already happened. What matters is whether the price closes back through the shelf it broke, or goes on toward $4,000 and the summer lows near $3,940 to $3,960.
Why are gold prices declining? Cash got more attractive, and the bar did not. Will gold prices rebound? They will, in a way that counts, if they retake about $4,225 and start repairing the averages overhead, starting with the 100-day near $4,268 and the 50-day near $4,332. They will not, in a way that counts, if every bounce dies under that pivot and the close stays under $4,095. The gold price forecast that says $4,650 by year-end and $5,400 by the end of 2027 can still be someone's base case. It is not a bid. The bid, today, is a level. Watch the level. Grade it on the close, not on the first adjective out of the minutes.
A note on sources and limits
Prices and odds are from October 7, 2026, and they moved during the day. Reuters reported spot gold at $4,130.37, down 0.8 percent, around 6:27 a.m. Greenwich time, with futures at $4,157. A later Reuters report put spot at $4,096.13 at 9:20 a.m. Eastern, down 1.6 percent and the lowest since August 5, with December futures at $4,121.70. The dollar index was up 0.7 percent in that later report, and the 10-year yield was described as an over-two-decade high. A separate morning snapshot had December futures opening at $4,195 and the price near $4,138 at 6:41 a.m. Eastern. Use one timestamp when you compare.
The minutes of the September Federal Open Market Committee meeting were due at 2 p.m. Eastern on October 7. The September policy range of 3.75 to 4.00 percent is the decision those minutes describe. CME FedWatch odds also moved. A morning snapshot put the chance of an unchanged decision on October 28 above 78 percent and a December hike near 69 percent. Reuters at 9:20 a.m. Eastern put the December hike chance at 84 percent. Daly and Schmid's comments are from same-day reporting of their public remarks. Walbaum's comments are from Reuters.
Technical levels are attributed, not adopted. Société Générale's October 6 note, as reported by FXStreet, put the 200-day zone near $4,510 to $4,540, an interim level near $4,095, a pivot near $4,225, and further support at $4,000 and at $3,960 to $3,940. FXStreet's October 7 chart note put the spot price near $4,139 at one point in Asia, the 50-day average near $4,332, the 100-day near $4,268, the 200-day near $4,531, and the 14-day relative strength index near 39. Orbex published a zone with support at $4,100 to $4,125, $4,080, and $4,045, and resistance at $4,200, $4,245, and $4,315. Other intraday maps differed by tens of dollars. Levels are not walls.
Goldman Sachs figures, a year-end 2026 fair value of $4,650 cut from $4,900 and a December 2027 figure of $5,400, are from the bank's notes as cited in reporting earlier in the week of October 6. They are forecasts. The January record above $5,500, the late-summer area near $4,700, and the September 28 print near $4,110 follow that same week's price reporting. September's 6.4 percent drop in New York futures, and the 12.7 percent drop in the large gold miners, are from mining.com's October ranking. GDX's move from about $101 on September 3 to the high $80s by October 6 is from published quotes. The Bank of America technical scenario near $3,315 is from a July note as later cited, and it is a stress case.
China's latest reserve addition was noted by Reuters on October 7 without a tonne count in the early headlines used here. Do not invent one. Official buying can continue while the price falls.
This is not investment advice, not an offer, and not a solicitation to buy or sell any security or commodity. Gold and gold mining stocks can fall further. Minutes can be misread. Support can break. Readers should check the live price, the Fed's own minutes, and a licensed adviser before any decision.