The number 3,900 means little on its own. What matters is whether the conditions that made 3,900 a support are getting stronger or weaker. In the week of Nov. 13, spot gold briefly broke above 3,900 and then gave it back. The weekly loss was 4.9%. The intraweek low was $3,852, the lowest level since October 2025. That was not an ordinary pullback probe. It was the first rebound after the monthly breakdown, and it failed above 3,900.
The monthly breakdown already happened
As of November 2026, spot gold had fallen to $3,898.58 on the monthly chart. That is a 30.4% drawdown from the early-2026 high of $5,597.94, over more than 10 months. Price had broken below the monthly Bollinger Band midline and the 20-period moving average. Both sat near $4,072.55. The upper band was $5,185.01. The lower band was $2,960.09. Price was trading between the midline and the lower band, in the zone where bears control the monthly trend.
A technical bear market usually needs three things: a drawdown of at least 20% from the cycle high, a decline lasting at least two months, and a rollover in medium-term moving averages with the trend flipping from bullish to bearish. Gold met all three. The drawdown was 30.4%. The decline had run 10 months. Monthly candles were printing high-volume losses, rebound highs were falling, and short-term averages had turned down.
From 2020 to early 2026, gold completed a five-wave advance from the $1,500 area to $5,597.94. It is now in a larger ABC correction after that advance. In Elliott Wave terms, it may be late in the A-wave decline or close to a B-wave rebound. If a B-wave rebound cannot reclaim $4,500, a C-wave decline follows. That is the main technical path for a deeper bear market in 2027.
Using the Fibonacci retracement from the $1,810.24 low to the $5,597.94 high, three zones matter. The $3,800 to $4,000 zone covers the congestion between the 0.382 retracement at $4,151 and the 0.5 retracement at $3,704. The $3,500 area sits between the 0.5 retracement and the 0.618 retracement at $3,257. The $3,000 threshold overlaps with the 0.618 retracement and the monthly Bollinger lower band near $2,960.
The “Siegfried Line” is not defined by whether 3,900 holds on a closing basis. It is defined by whether $3,800 to $4,000 produces volume-backed stabilization and higher rebound highs. That has not happened. The mid-November rebound failed above 3,900. The first friction layer inside the defense zone is being consumed. If $3,830 to $3,850 breaks in later tests, $3,500 stops being a distant target and enters the pricing range.
Real yields are moving too fast
Gold’s long-term anchor is the real yield, which is the nominal yield minus inflation expectations. Many investors watch the level of rates. The speed of the move matters more for flows. In mid-November, the 10-year TIPS yield rose from 2.66% to 2.86% in three trading sessions. For an asset with no coupon, a faster rise in carrying cost creates more selling pressure than a high but stable yield.
The Fed comments in November made the problem worse. Gursky, Hammack, and Daly each stressed that inflation was moving the wrong way. That eroded the most optimistic scenario, in which the Fed tightens but soon eases. October core PPI was 5.0% year on year. Core CPI rose 0.3% month on month and beat expectations. Unemployment fell to 4.1%. That mix gives the Fed little reason to pivot soon.
On Nov. 6, Goldman Sachs cut its end-2026 gold target from $4,650 to $4,150. It cut its end-2027 target from $5,400 to $4,800. The bank blamed a more hawkish Fed. Goldman was not alone. JPMorgan cut first in July. HSBC, BofA, Morgan Stanley, Standard Chartered, and UBS followed between August and October. The cuts point to a repricing of the rate path. They do not reverse the long-term arguments about central-bank buying and de-dollarization. The wording on those long-term supports has barely changed.
The rate pressure on gold is cyclical and reversible. It depends on one condition: when the Fed’s hawkish path cracks. If U.S. inflation stays above target in 2027 and rates stay high for more than six months, or if hikes restart, gold’s valuation center keeps sliding. If the rate path turns, the suppressed valuation can repair quickly.
Central banks are not one buyer
From 2022 to 2025, central banks bought more than 1,000 tonnes of gold a year on average. That was a major backstop for the bull market. But following central banks into gold has a logic problem. Your cost basis, your cycle tolerance, and your ability to shift risk through your own currency are not the same as a central bank’s. Central-bank buying serves monetary policy. It is not an investment.
The important feature of 2026 central-bank buying is not the total. It is who is buying and who is selling. In May 2026, central banks net bought 41 tonnes. China added 10 tonnes, its 20th straight month of net purchases. Poland added 18 tonnes, its fourth straight month of large buying, for a year-to-date total of 64 tonnes. Russia sold 6 tonnes net and had sold 34 tonnes year to date. Turkey sold 3 tonnes net and had cut 81 tonnes year to date. July data kept the same split. China and Poland kept buying. Russia’s year-to-date sales rose to 50 tonnes. Turkey’s cumulative sales reached 85 tonnes.
Buyers such as China and Poland are allocating reserves. They are less sensitive to short-term price. Sellers such as Russia and Turkey are managing liquidity and rebalancing reserves. They have less tolerance for falling prices. If gold keeps sliding, sellers may sell more. Buyers may not speed up enough to offset them. A World Gold Council survey found that 89% of respondents expected global gold reserves to rise over the next year, and 45% of central banks planned to add to their own holdings. Long-term support and immediate price support are different things.
In the 2013 to 2015 bear cycle, central banks kept buying on net, and gold still fell more than 40%. Central-bank buying can slow a decline. It cannot reverse a bear trend. Reading central-bank buying as an iron floor misreads how central banks behave. It is a force that slows declines, not a counter-trend force.
Geopolitical premium is harder to price now
Another driver of this bull market was the safe-haven premium from geopolitical risk, especially the 2022 Russia-Ukraine conflict. When the same logic played out in the US-Iran conflict, gold could not rally. Rates explain part of that. The other part is that the market can no longer reliably price how much geopolitical risk belongs in gold.
A safe-haven premium can push prices up fast. It is hard to measure and hard to remove cleanly when conflicts ease. If major conflicts cool in 2027, safe-haven money leaves gold, and the risk premium in its valuation gets squeezed out. That adds pressure. The bigger problem is that gold lacks the equivalent of crude oil’s inventories, production, and crack spreads. Gold pays no interest. Once it has done its job of fixing monetary value, it reduces liquidity activity, lowers inflation, suppresses investment, compresses consumption in a one-time way, and freezes market prices.
Those effects cut both ways. Gold can be a base position. It is hard to use as a precise anchor. Most professional securities investors hold gold as a base position rather than an anchor because its current price is distorted and exposed to both rate risk and price risk. After a major bull market, even with this year’s correction, gold still cannot price market risk in a clean way. That helps explain why a second monthly close below the Bollinger midline and a recent new low produced a strong correction signal.
AI credit and liquidity pipes
Three large liquidity pipes feed gold’s pool: the AI trade, international oil prices, and crypto volatility. If central-bank support slows, the $3,800 to $4,000 zone has resistance above and the nearest support band behind it near $3,500. The zone is gold’s “Siegfried Line.” If it holds, the market waits for the liquidity rotation that follows an AI bubble bust or a debt crisis. If it fails without a major positive catalyst, price looks for volume support along $3,800, $3,500, $3,300, and $3,000. The faster the drop, the more forced selling is created, and the sharper the rebound.
The second half of 2026 forces a timing correction to that framework. By November, AI-related tech giants had already lost large market value. Alphabet had erased about $1.62 trillion from its 2026 high. Nvidia had erased about $1.49 trillion. Broadcom had erased about $965 billion. TSMC had erased about $700 billion. CDS spreads for Microsoft, Nvidia, Meta, and Amazon had widened sharply since Q3 2025, in some cases by multiples. CoreWeave’s five-year CDS once hit a record 1,194 basis points, implying a five-year default probability above 60%.
The open question is whether this is already a liquidity crisis. Early-November data showed failed rebounds in gold and silver alongside declines in Bitcoin and equities. That was synchronized de-risking, not a panic liquidity squeeze. An AI credit crisis can support gold, but the buying may be slow. Several sell-side notes argued that even if the AI credit crisis worsens, gold’s safe-haven bid “is expected to rise slowly and not in a one-way move.” High rates are still pressing on gold. The net effect of safe-haven demand and rate pressure is unclear.
Waiting for the liquidity-rotation node after an AI bust or debt crisis may be right in direction and wrong in timing. In 2008 and 2020, the pattern was crisis, liquidity squeeze, gold sold first, liquidity rotation, gold surge. If AI credit fractures slowly under rate pressure instead of breaking all at once, gold’s “sold first” phase may be shorter. The post-rotation rally may become a slow valuation repair instead of a V-shaped reversal. A rebound after the defense line holds may come from a lower carrying cost as the rate path turns, not from one dramatic liquidity event.
The dollar tightening cycle still matters
Gold is priced in dollars. Global dollar liquidity drives its medium-term direction. If the Fed keeps rates high in 2027, the dollar index stays strong, offshore dollar liquidity keeps tightening, and capital returns to the U.S. for higher-yielding dollar assets. Gold pays no yield, so institutions keep cutting it, and gold ETF holdings trend lower.
The 2013 to 2015 taper and hike cycle sent gold into a three-year bear market. The current policy mix looks similar. The question is whether the bubble gets pricked. Nvidia’s case suggests the truth is getting closer.
If the global economy hard-lands in 2027, or a sovereign-debt or corporate-debt crisis erupts, the market enters a cash-is-king squeeze. Institutions sell all assets, including gold, to get dollars. Gold falls faster. That logic played out in the 2008 subprime crisis and in the early days of the 2020 pandemic. It is one trigger for gold to test $3,000.
CFTC data show non-commercial net-long gold futures positions falling from their early-2026 high. Speculative funds have been closing longs and building shorts. If net-long positioning falls to historic lows in 2027 or turns net short, the bear trend reinforces itself. After a 30% drawdown, retail dip-buying is still strong. That is usually not a bottom signal. Bear-market bottoms tend to come with retail panic selling and a freeze in sentiment. If retail dip buyers keep getting trapped in 2027 and sentiment flips from optimism to panic, a final capitulation leg can complete the bottom.
Three paths for 2027
Scenario 1: rate pressure persists and the defense zone is consumed
This is the base case. The Fed keeps rates high in the first half of 2027. The 10-year TIPS yield trades between 2.5% and 3.0%. Real yields do not fall quickly. Central-bank buying stays split. China and Poland do not buy enough to offset Russia, Turkey, and ETF outflows. Gold tests $3,800 to $4,000 repeatedly. The $3,830 to $3,780 zone erodes. $3,500 becomes the next volume-supported level that matters.
Watch two things: whether the 10-year TIPS yield stays above 2.5%, and whether the monthly close is below 3,830 for two straight months.
Scenario 2: the rate path cracks and the defense zone becomes a bottom
U.S. inflation or employment data weaken enough to force a repricing of the Fed’s turn. The 10-year TIPS yield falls from its high. Rate-suppressed valuation repair plus central-bank buying lets gold stabilize on volume in the $3,800 to $4,000 zone. Rebound highs rise. The monthly close gets back above $4,072.
This path does not need an AI crisis. It needs a change in the direction of rates. Watch whether TIPS yields fall for several weeks, whether the monthly gold close gets back above 4,070, and whether Russia and Turkey slow their selling in monthly central-bank data.
Scenario 3: rates stay high and AI credit fractures
AI credit conditions worsen under rate pressure. Tech stocks and credit markets deleverage in a chain. Gold may first fall with risk assets in a liquidity squeeze. Then the Fed is forced to intervene, or rate expectations turn sharply, and gold reprices upward very fast. In this path, the $3,800 to $4,000 zone matters less. If gold breaks 3,800 during the squeeze, the liquidity-rotation rebound can recover the loss just as fast. If no squeeze happens, gold returns to Scenario 1.
Watch AI giant CDS spreads, whether the Fed releases liquidity support before inflation is back at target, and whether gold decouples from U.S. real yields. That decoupling has historically been a leading signal of liquidity rotation.
What to watch instead of the round number
Three facts define the setup. The monthly breakdown has happened. Real yields are rising fast enough to press on gold. Central-bank buying is split, and AI safe-haven demand may arrive slowly. Those forces give Scenario 1 the highest probability right now. Scenario 2 needs a clear turn in the rate path. Scenario 3 needs a credit fracture that is not yet fully visible in the data.
Stop asking whether 3,900 is an iron floor. Ask three questions instead. Is the rise in the 10-year TIPS yield slowing? Is the $3,830 to $3,850 test producing volume-backed stabilization? Are Russia and Turkey slowing their gold sales in central-bank data? The answers will tell you earlier than any round number whether $3,800 to $4,000 is turning into a bottom structure or becoming a rest stop on the next leg down.
Gold pays no interest. Once it has fixed monetary value, it reduces liquidity activity, lowers inflation, suppresses investment, compresses consumption in a one-time way, and freezes market prices. Those effects cut both ways. Most professional investors hold gold as a base position rather than an anchor because its current price is distorted and exposed to both rate risk and price risk. After a major bull market, even with this year’s correction, gold still cannot price market risk cleanly. The market cannot calculate gold’s supply-demand and sentiment mix the way it can for crude oil.
The work now is to reduce iron-floor faith to conditions, speed, and structure. Reduce sentiment calls to indicators you can check. The $3,800 to $4,000 zone is one marker in the gold story. It is not the ending.