CHAPTER 01
Price action: the round number was never reached, and 4150 gave way twice
Asia open 4160.72, high 4170.35, low 4141.00, 4143.20 at 17:55 GMT+8. A fourth consecutive session that failed at the round number - and this one did not even reach 4200. Asia and Europe only; New York opens at 21:30 GMT+8.
Today's tape is smaller than yesterday's and worse in character. Yesterday was a 41 dollar round trip built on a Fed interview. Wednesday opened at 4160.72, popped to 4170.35 in early European hours, and then declined without interruption to 4141.00 and 4143.20 at 17:55 GMT+8. Four sessions running now without a hold above the round number, and today the market never even bid 4200 - the ceiling is no longer being tested, it is simply being inherited. That is a quieter and more durable form of resistance than a rejection.
“4142.70; the session low is 4141.00”
Levels are cross-verified timestamped anchors, not a tick series. Asia open 4160.72 and previous close 4161.13 (Cngold, refreshed 17:53 GMT+8). Early European high 4170.35 (Cngold, single source but timestamped). European-afternoon reads: 4155 (FX678, 10:25 GMT+8) -> 4152.66 at 17:32 (Tonghuashun) -> 4143.53 at 17:53 -> 4143.20 (Cngold live). Session low 4141.00 (Cngold); Tonghuashun's 17:32 note has a low of 4142.70 - a one-and-a-half dollar difference on a falling tape, i.e. a snapshot interval. The dashed reference is the previous close of 4161.13. Range for the window is 29.35 dollars, all of it made downward after the opening pop. This chart stops at 17:55 GMT+8 = 05:55 ET. No New York close is plotted, because none exists at this hour - US cash equities open at 09:30 ET and Treasuries at 08:00 ET.
4150 has been tested and given back twice today, and it is now the whole debate. The morning's low was 4142.70; the session low is 4141.00. Two independent Chinese desks mark support at 4150, then 4100, with the October 7 low at 4066 (Guangzhou Jinkong), while FX678's technical note puts the next line at 4120 and the round number at 4000. Same zone, different granularity - 4100-4120 is where two desks and a technical note converge, and it is 20 dollars below where gold trades right now. A break of 4150 on a soft CPI would put the disagreement between the desks directly into play.
The reason is not a headline, and that is the finding. There is no geopolitical event, no surprise data release and no single Fed headline driving Wednesday's European session - only Hammack's already-delivered NABE speech being properly digested, and CME October pricing moving from 20.5% to 22.7%. A gold market that can be pushed 18 dollars by repricing already-public information has not found a new seller; it has run out of buyers who were willing to pay up while the story was still developing. Tonghuashun puts the resulting shape plainly: speculative length will not chase above 4200 until yields soften, while central bank buying and ETF inflows solder the floor between 4100 and 4150. That is a market with a ceiling, a floor, and no catalyst - and the catalyst is 20:30 GMT+8.
Position of the range, stated plainly. Gold has now spent four sessions in 4141-4200 and has not closed above 4161.13 once. The 20-day London average sits at 4200-4220 as resistance with 4250 next, and the 100-day near 4250-4260 depending on source. FX678's read is that daily structure is still bearish - price below the 100-day and below the Bollinger midline near 4220, RSI around 43, MACD histogram still negative but not extending. RSI near 43 with no bearish exhaustion is a market that can keep drifting down without warning, and it is also a market where a single soft print produces an unusually large mechanical move. Both of those are true and they are the same fact seen from two sides.
CHAPTER 02
Fed: the count was 3-1, not 2-1 - and one of the three was missed
Hammack's NABE speech was carried overnight with an unconditional tightening call and this morning's edition logged it as 'unchanged'. She is a 2026 voter. October hike odds have moved 18.3% -> 20.5% -> 22.7% across three consecutive editions.
The correction, stated plainly because readers will compare editions. This morning's edition wrote: "one dovish voice against two hawkish ones is not a pivot - it is a 2-1 vote that sets the terminal rate without changing the next one." Hammack is a 2026 voting member and she was arguing for a restrictive stance as the default position. The count is three voting hawks (Hammack, Barkin, Goolsbee) against one dovish vice chair (Williams), with Waller the swing vote and publicly non-committal on October. The morning edition's conclusion - that the hawks control the terminal rate - survives the correction. But it was reached from a 2-1 count that did not exist. Correcting this in the body rather than quietly rewriting the number is the point: a reader comparing 08:00 and 18:00 editions needs to see the difference, not be left to wonder why the vote count changed overnight.
“18.3% two editions ago -> 20.5%”
CME has now moved against gold on this for three consecutive editions, and the pace is accelerating in the wrong direction for the bulls. October hike odds: 18.3% two editions ago -> 20.5% this morning -> 22.7% now. December at-least-one: 84.2% -> 87.6% -> 87.8%. Two relays agree on direction though not on level - FX678 carries a slightly earlier snapshot at 19-20% for October and 85-88% for December. But the desk on the other side is making the better argument, and this edition has to say so: Tonghuashun's futures desk reads the same table as exhausted, stating that the market has priced at least three further hikes through end-2027 and the room to revise higher is small. An asymmetric repricing setup is the honest reading - hawkish repricing is exhausted, dovish repricing is not.
The institutional-risk item is new and nobody has priced it. Waller, in print today: "this administration poses a threat to Fed independence. So far the threat has not truly materialized." A sitting governor publicly identifying political pressure on the Fed as a term-premium channel is materially different from the usual independence rhetoric, because term premium is exactly the mechanism Williams said was setting the long end - supply, fiscal, AI, energy. If independence risk becomes a priced term-premium factor, it pushes real yields up for a reason that has nothing to do with inflation policy, and gold loses the ability to argue that lower inflation would fix it. This is a slow channel, not a tonight channel. It goes on the register rather than into tonight's position.
THE DATA
| Date | Official | Statement | Stance |
|---|---|---|---|
| 10/12 - 10/14 | HammackCleveland Fed - 2026 voting member | The morning edition recorded this as carried-forward and unchanged. That was incorrect and this edition corrects it. NABE keynote, relayed overnight by Xinhua Finance, Cailian Press via Tencent and Guangzhou Jinkong Futures: "The economy needs to tighten further. If monetary policy stands still, the environment in the market will change." On pace: smaller adjustments, made relatively quickly, rather than waiting and risking a larger correction later. On the current stance: a restrictive stance is appropriate to bring inflation back to 2% in a timely manner. Her inflation list explicitly goes beyond energy - tariffs, AI data-centre buildout driving electricity and technology component demand, and the surge in AI-infrastructure bond issuance - and she warns that companies planning price increases beyond input cost growth is an early warning of inflation thinking becoming entrenched. She also notes real inflation has run above target for five-plus years while long-run expectations have stayed anchored, and will not take that for granted. She did not explicitly endorse an October hike - and she did not need to. She is a voter arguing for a restrictive stance as the default. | Bearish |
| 10/13 | WilliamsNew York Fed - FOMC vice chair, voting member | Carried forward unchanged from the morning edition and still the single most important quote on the board: one more hike could be appropriate in 2026 but no reason to do more, with 2027 inflation falling close to 2%. His attribution of long-end yields to bond supply, AI investment, fiscal policy and energy - "not just monetary policy" - is echoed almost verbatim by Hammack's AI-infrastructure inflation list. Williams owns the ceiling; Hammack owns the floor. They are not contradicting each other, they are describing a narrower range than the market prices. | Bullish |
| 10/13 | CollinsBoston Fed - not a 2026 voting member | Prepared remarks: a further modest hike before year-end, with the risk of inflation running above target described as elevated and the Fed's central concern, citing diesel, jet fuel and heating oil. Worth restating the caveat the morning edition correctly attached: Collins does not vote in 2026, so her vote cannot move December. Her influence is as a voter-adjacent voice in the room, not as a ballot. | Bearish |
| 10/13 | BarkinRichmond Fed - voting member | At Howard University: 2% "is not out of reach - it is achievable" but inflation is "proving persistent", the shocks are clearly persisting rather than passing, and action is justified - "I very much support what we are doing." He named the technical transformation explicitly for the first time among Fed officials: you can see the effect of AI investment on US inflation. Two voting members now on the record linking AI capex to current inflation, which is a hawkish framing of the same technology Williams frames as a future productivity offset. | Bearish |
| 10/13 | WallerFed governor - permanent voting member | Continued in print today: inflation is stubborn and further hikes are still needed, but timing is flexible and hikes need not be consecutive. Two additions carried this morning. On AI, he splits the difference in a way that matters: if AI raises productivity, growth can be stronger without excess inflation, but the payoff is years away while the price pressure lasts two to three years - a "cost first, benefit later" cycle. On institutional risk he named it directly: "this administration poses a threat to Fed independence. So far the threat has not truly materialized." A governor flagging a term-premium channel through political pressure on the Fed is a new item on the register. He remains the most likely dissent against an October hike. | Neutral |
| 10/13 | GoolsbeeChicago Fed - voting member | The AI buildout is starting to push prices, inflation remains too high, and if it stays high "what the Fed needs to do is clear." Framing AI as an inflationary force rather than a productivity one. Carried forward unchanged; three voting members now share this framing, which makes it a bloc position rather than an idiosyncratic view. | Bearish |
CHAPTER 03
FX: the dollar stopped falling, and that changed the gold conversation
DXY held above 102 through Europe after closing below it twice - the first session in three where the dollar did not decline. For gold this is a worse configuration than a rising dollar, because the offset was removed without a replacement.
The structure of the FX tape inverted today, and that is the finding. For three consecutive editions the read was zero dispersion with one cause: six majors, one driver, the dollar down on a dovish Fed. Today the dollar stopped falling - holding above 102 through the European session after closing below it on Monday and Tuesday - and that single change removes the automatic offset that has been supporting gold all week. A dollar that is flat is a worse configuration for gold than a dollar that is weak, provided yields stay high: the weakness was doing work, and it has stopped.
“weakened while gold fell”
Why the CHF matters more than its 0.12% move. The franc weakened while gold fell. For two days the standard reading of a soft dollar was that CHF and gold bid together, and this column flagged that divergence explicitly when it closed. Today both legs gave it up at once. A haven bid that leaves CHF and gold on the same day has not disappeared - it has moved, and the destination is the live question ahead of a CPI print. Anyone long gold as a haven instrument rather than a rates instrument should note that the co-movement which justified that framing broke today at a materially higher gold price than it will break at a lower one.
Two levels to carry into 20:30 GMT+8. EUR/USD 1.1150 - Monday's cycle low and the level that gave the euro a bid; it is 30 pips below the current level and the technical note's 1.1100 Bollinger lower band is the next line, with a sustained break there strengthening the bearish case for the pair and by extension for dollar pricing of commodities. USD/JPY 158.75 - the range top sold on every attempt, with 159.80 the intervention-sensitive band, and the BOJ explicitly not hiking in October. Neither is a gold signal directly, but both are where a genuine change in the policy-expectations complex will appear first - and gold has just lost the dollar leg of that complex entirely.
THE DATA
| Pair | Close | Change | Q4 outlook & reasoning |
|---|---|---|---|
| EUR/USD | 1.1180 | ▼ | Indicative European level, single-source. From Tuesday's Xinhua fix of 1.1201, EUR/USD is back near 1.1180. The technical note in circulation this evening is unambiguous on structure: spot is far below the Bollinger midline and the 100-day SMA, sitting at the lower bound of its recent range, with RSI(14) hovering just above oversold - downside momentum still dominant but approaching the point where selling pressure starts to weaken. Resistance is layered at 1.1310 (Bollinger midline), then 1.1480 (100-day) and 1.1520 (upper band) - a dense supply shelf between 1.1310 and 1.1520. Next support is the Bollinger lower band at 1.1100, and a sustained break there strengthens the bearish case. Monday's cycle low of 1.1150 is 30 pips below the current level and remains the level that has to hold. |
| GBP/USD | 1.3230 | ▼ | Indicative, single-source, sitting on the 10-day average. From Tuesday's Xinhua fix of 1.3257, sterling has given back roughly 27 pips. The daily structure is described as a "bearish deadlock": price is pinned to the 20-day Bollinger average with the 100-day at 1.3395 and the upper band at 1.3375 forming dense selling supply directly above. RSI 44-45, MACD below water and crossed down. The 10-day at 1.3230 is the short-term life line - a break targets the lower band at 1.3150 and then 1.3100-1.3044. The 4-hour chart does show stabilisation from the 1.3177 low with a contracting MACD histogram, but 1.3275-1.3300 caps every attempt. Only a decisive break of 1.3395 clears it - which is why this column is not reading sterling strength into a weak-dollar week that has stopped being weak. |
| USD/JPY | 158.10 | ▼ | Indicative, single-source, still pinned above the 100-period average at 157.93. The Japanese story is unchanged from the morning edition and worth restating because it is now the single largest scheduled risk in FX: BOJ officials are inclined not to hike in October, avoiding a rare back-to-back move, while still considering December at the earliest (Bloomberg via Sina; Jin10 independently). Japan's September corporate goods prices were +6.8% y/y, easing at the margin. Structural picture: 158.75 is the range top that has been sold every time it is tested; 159.00-159.55 is the dense resistance band; 159.80 is the upper Bollinger band and the zone where Japanese intervention becomes a live consideration. Support at 157.93, then 157.60 where the lower band sits, then the 50-day at 156.90. RSI near 58 with a flattening slope carries a latent divergence warning - bullish medium-term structure, fading short-term momentum, and one BOJ decision away from a carry unwind. |
| USD/CAD | 1.4255 | ▲ | Indicative, single-source - and a clean expression of the day's one real trend. CAD weakened from Tuesday's Xinhua fix of 1.4215 as WTI probed 95 dollars intraday, up about 0.93%, and Brent recovered to 106.75 (Tonghuashun evening note, 17:32 GMT+8). Note the divergence from Tuesday: yesterday CAD was the best major on oil strength; today oil is still rising and CAD is giving it back. The BoC's Macklem appears at 14:05 GMT+8 on the calendar. Chart shape worth noting: WTI's daily Bollinger bands have gone from closing yesterday to opening upward today, and the market has taken a first run at 95. Support references are 93.1 (daily MA5), 92.5 (4-hour midline and MA120 confluence) and 91.8 (daily MA10) - below 91.8 the bullish structure is broken. The unresolved Brent conflict flagged in this morning's edition - 104.77 against 106.35 and 105.70 - is now moot in practice because Brent has traded above 106 into today; the new cluster is tighter. |
| USD/CHF | 0.8335 | ▲ | Indicative, single-source - and it has flipped. CHF weakened from Tuesday's Xinhua fix of 0.8325 even as gold fell and equities were mixed. This is the day's most information-carrying cross. On Tuesday, CHF strength alongside dollar strength was flagged as a genuine haven bid and a divergence worth tracking. Today the franc has given that up while gold fell - meaning the haven bid has migrated out of both CHF and gold simultaneously. Where a haven bid goes when it leaves gold and the franc is the actual question tonight's CPI answers, and this column does not have a good candidate for the destination. That absence is itself worth stating rather than filling with a guess. |
| AUD/USD | 0.6975 | ▼ | Included because the RBA minutes change how yesterday's 0.6983 print should be read. From Tuesday's 0.6983, AUD is slightly lower. The reason is not the dollar: the Reserve Bank of Australia's September minutes show policymakers choosing to hike on increased upside risks to inflation while flagging Middle East uncertainty (via Sina). So the RBA is tightening into an energy shock while the Fed debates one hike - the policy differential is moving against AUD even with the dollar flat, which is why the Australian dollar is not bouncing with the metal. This is the cleanest available argument that gold's weakness on Wednesday is a rates story and not a risk story. |
CHAPTER 04
Fundamentals: the ETF bid reversed, and the sell-side did not move at all
SPDR fell 1.215 tonnes to 1,054.196 tonnes as of October 13 - the first decline in the visible series, printed the morning after a 41 dollar round trip. The target panel is unchanged from this morning because no desk published anything new in the European session; the Citigroup exposure reduction remains the most informative entry in it.
The one thing that genuinely changed today is on the flow side, and it changed sign. SPDR Gold Trust holdings fell 1.215 tonnes to 1,054.196 tonnes as of October 13 - against +2.28 tonnes the previous day and a run of additions before that. Units caveat: the relay prints "104.196 tonnes", which is wrong by a factor of ten; it is read here as 1,054.196 because that reconciles exactly with the 1,055.41 tonnes recorded for October 12 less 1.215, and an unreconciled figure would be worse than a corrected one. The ETF buyer did not merely stop buying - it sold, on the day gold made a 41 dollar round trip and three concurrent buying stories were running.
“Legs one and three are intact today. Leg two turned negative overnight.”
This weakens the floor argument in a way that the morning edition's structure requires us to state. That edition's bull case rested on three legs: central banks, ETF holdings at a record, and speculative length at a six-week low. Legs one and three are intact today. Leg two turned negative overnight. When Tonghuashun writes that central bank buying and ETF inflows solder the floor between 4100 and 4150, one of those two solder streams stopped flowing this morning - and FX678 independently warns that ETF positioning is not secure and a break below 4000 could trigger concentrated redemptions. The floor is now a central-bank floor, which is a slower and lower-volume floor than an ETF floor.
And the seller that replaced the ETF buyer is not the central bank - it is the yield. The flat explanation for today is in one line of the evening note: 10-year stuck near 5.3%, dollar above 102, gold pays nothing, Treasuries pay 5%+ risk-free, so money goes to bonds first. Add the corroboration that base metals are leading the decline - copper -1.44%, nickel -2.30%, tin -2.44% - and this is unambiguously a rates signature rather than a risk signature. That matters for tonight: a rates move reverses on a single print, a risk move does not. The asymmetry is real, it is named by the desk with the most to lose from being wrong, and it is the reason this column is not calling a break of 4150 the start of something larger.
THE DATA
- TD Securities' own desk says the hawkish repricing is exhausted, and it is the desk with the most to lose from being wrong about that. Tonghuashun's futures note carries the institutional view directly: the market has priced at least three further hikes through end-2027, leaving little room to revise higher, so this pullback can be treated as an opportunity to build positions; the key level is 4000. The same note identifies the specific conditions required for gold to actually break out - hike expectations falling materially, or a negative feedback loop across equities, bonds and FX. An asymmetric repricing setup is the honest bull case, and it is stronger than a price chart.
- Structural demand is unchanged and central bank buying is not price-sensitive. The Asian central bank has bought for 23 consecutive months, September's 740,000 ounces being the largest of the cycle, and Guangzhou Jinkong notes Chinese gold imports may be the highest since 2017. FX678 adds that CTA positioning has already started covering gold and platinum shorts - systematic length is adding, not removing. Neither of those cares what happens at 20:30 GMT+8, which is precisely why they are the floor.
- Three independent desks mark the same downside zone, and it is well below spot. Guangzhou Jinkong: support at 4150, then 4100, October 7 low at 4066. Tonghuashun: 4080-4100 as the support band, with 4000 the strategic central bank accumulation zone. FX678: 4120 next, then 4000. Three desks, two continents of coverage, one zone between 4,060 and 4,120 - and gold is 25 dollars above it.
- The dovish tail is real and it is named by an actual bank. Deutsche Bank's Tim Baker: inflation has not come in hotter than expected and the risk "seems tilted toward a downside inflation surprise, which would prompt markets to re-price in a dovish Fed." Williams remains on record that a single further hike could be appropriate with no case for more. A 3-1 hawkish vote on the board is not the same as a 3-1 hawkish outcome, and the asymmetry sits with the dovish tail.
- Moribund base metals confirm this is a dollar story, not a risk story. Copper -1.44%, aluminium -1.31%, nickel -2.30%, zinc -1.08%, tin -2.44% at the 17:xx GMT+8 snapshot (Jin10). If this were a genuine risk-off move, base metals would not be leading it. They are not. Gold's weakness alongside a weak complex but a firm dollar is a rates signature, and rates can reverse in one print.
- SPDR turned negative for the first time in the visible series. Holdings fell 1.215 tonnes to 1,054.196 tonnes as of October 13 (Jin10 via Tonghuashun's morning note) - against +2.28 tonnes the day before. The configuration-fund bid that this column has cited as the reason dips get bought stopped bidding on the very day gold made a 41 dollar round trip. FX678 adds the mechanism from the other side: short-term ETF positioning is not secure, and a break below 4000 could trigger concentrated redemptions. The floor argument now depends on central banks alone.
- October hike odds have gone up three editions running while gold went nowhere. 18.3% -> 20.5% -> 22.7%, with December at-least-one at 87.8%. Gold is down over that stretch. A market that prices more tightening and less gold at the same time is a market that has found its marginal seller, and the candidate is the leveraged side, not the allocation side. FX678's technical read is unambiguous: daily structure is bearish, price below the 100-day near 4250 and the Bollinger midline near 4220, RSI around 43, MACD histogram negative and not yet exhausted.
- Hammack's inflation list is broader than energy, and two voters now share it. Tariffs, AI data-centre electricity and component demand, and the surge in AI-infrastructure bond issuance. She warns that companies planning price increases beyond their input cost growth is an early warning of inflation thinking becoming entrenched - the exact mechanism that makes a single soft CPI insufficient. Barkin and Goolsbee independently name AI as an inflationary force today. Three voting members on the record tying the AI buildout to current prices is a structural argument for a higher terminal rate, not a data-dependent one.
- The institutional channel nobody is trading. Waller in print today: "this administration poses a threat to Fed independence. So far the threat has not truly materialized." If independence risk is priced as a term-premium factor - and Williams has just told the market that term premium and supply are what set the long end - then real yields can rise for reasons that have nothing to do with the CPI print. That is the one bear argument that survives a soft number.
- The geopolitical premium is being priced as noise, again. Houthi missile attacks on Saudi Arabia continue and Hormuz shipping risk stays elevated, with Brent above 106 and WTI probing 95 - and gold is down. Day three of oil up and gold down. The morning edition noted the strike-talk reports were priced as noise; today's tape prices the actual ongoing missile exchange as noise too. When realised hostilities stop moving gold, the haven bid has gone somewhere the tape is not showing - and the franc has just declined as well.
- Goldman Sachs4,650+12.2%Cut from 4,900 - the 2026 year-end fair value, with the 2027 target held at 5,400. The bank estimates central bank buying at 60 tonnes a month across 2026-27 and attributes roughly 23% of the 2027 upside to central bank demand. Two mechanical notes carry forward: gold call option open interest sits at three times average, and dealer hedging could mechanically amplify a move beyond the baseline forecast. The cut is a Fed-path cut, not a gold-structure cut.
- Deutsche Bank4,700+13.4%Raised, as an annual average. One of the two houses at the top of the tight cluster rather than an outlier - and the fact that Deutsche sits at 4,700 while Goldman was cut to 4,650 is the cleanest illustration available that this panel is re-pricing around the Fed path rather than around gold's structure.
- Standard Chartered4,650+12.2%Q4 2026 average against roughly 4,350 realised in Q3 to date, looking for a year-end repair. Independently quantifies the decoupling: gold's correlation to nominal 10- and 30-year yields has moved to about -20% and -10% from a historically much tighter negative relationship - a different desk with a different model arriving at UBS's conclusion, which is the strongest cross-validation in this panel.
- UBS4,600+11.0%December 2026, rising to 5,400 by September 2027. Maintains a constructive twelve-month view and expects central bank buying of 750-1,000 tonnes a year - demand that is price-insensitive and accelerates on weakness. The most useful quantification in this panel: gold's correlation to 10- and 30-year nominal yields has drifted to about -20% and -10%, and its correlation to 2- and 5-year real yields weakened from -30%/-38% to -16%/-22% within a single month. Support band 3,990-4,030; technical repair requires 4,280-4,320.
- Morgan Stanley (house)4,600+11.0%From the September 22 commodity outlook, framing 2026 precious metals positively on central bank demand having more than doubled versus the pre-2020 average, with a 5,700 second-half scenario and 2027 seen above 5,000. The gap between the house average of 4,600 and its own 5,700 bull case is the widest in the panel - a desk that cannot say where it is going, only which way the tail points.
- HSBC4,560+10.1%2026 average, 2027 at 4,925. Sees the remainder of 2026 trading in a 3,800-4,700 range and ending the year near 4,750, grounded in diversification demand, central bank buying and ETF inflows as a medium-term floor. Note where that range floor sits: 3,800, not 4,100. This desk is not defending 4,100 - it is defending 3,800.
- J.P. Morgan (desk average)4,500+8.6%Q4 2026 quarterly average, framed as Fed-driven volatility rather than a destination. Q3 evidence supports the caution: adjusted revenue 50.91bn missed the 51.37bn consensus and fixed-income trading revenue fell 4% y/y - the first decline in over two years - while equities revenue rose 49% to a record 4.95bn. Book value nearly doubled to 302.1bn (+95.9%) on the conversion of 21.9bn of preferreds, a financing action rather than earnings, and buybacks were cut to a token 664mn. Dimon's own framing - deficits, sticky inflation, elevated asset prices, rising rates, geopolitics - is the desk's macro view in one sentence. Reported alongside this row: the same bank's longer-dated 6,000 target. Not averaged with it.
- Bank of America4,000-3.5%Q4 2026 average, with 4,813 for 2027 - the shape of a cyclical-trough view. The reasoning is explicit and it is not a Fed call: physical demand is weak across jewellery, bars and coins, and that caps the near term even with investment demand stable. The only desk in the table pricing a demand-side rather than a policy-side problem.
- Morgan Stanley (Gower)4,000 floor-3.8%"I think we should look at 4,000 as a very strong floor." Grounded in central bank physical demand being "very very strong" and Chinese gold imports potentially the highest since 2017. Excluded from the median: this is a statement about where the metal stops falling, not where it goes. Do not read it as bearish - and note it carries the same -3.8% as BofA's average for a completely different reason.
- Street median (8 dated entries)4,625+11.6%Median of the eight dated, numeric targets above - Goldman 4,650, Deutsche 4,700, Standard Chartered 4,650, UBS 4,600, Morgan Stanley house 4,600, HSBC 4,560, JPMorgan 4,500, Bank of America 4,000. Excludes the Gower floor call as a non-target and excludes Citigroup, which changed a position rather than publishing a number. At +11.6% above the 4143.20 close, the panel gap is produced by the high targets rather than by any consensus that the year ends higher.
Nothing in this panel moved during the Asian or European session, and that silence is itself information. The morning edition recorded the panel at a 4,000-4,700 band with a 4,625 median across eight dated numeric targets, and as of 17:55 GMT+8 on October 14 not one desk has published anything new. The reason is structural: the panel turns on the Fed path, the Fed path turns on 20:30 GMT+8, and no bank re-rates its fair value two hours before the print that decides it.
Conflicts flagged, not reconciled - unchanged from the morning edition: JPMorgan's desk quarterly average is 4,500 while Greg Shearer at the same bank holds a 6,000 longer-dated target - same series, two horizons, not averaged here. Morgan Stanley's "4,000 is a very strong floor" is a floor call and is excluded from the median; its house average of 4,600 sits well below its own 5,700 bull case.
And the single most informative entry has still published no number at all. Citigroup reduced its gold exposure outright after reassessing the Fed - the prior spot long rested on a dovish turn and a Treasury buyback expectation, both removed by the post-Jackson-Hawk repricing - and now names US rates as the single primary driver. Because it carries no numeric target it sits outside the table rather than inside it with a fabricated number. It remains the most informative row in this column's coverage, because it is consistent with 22.7% October and 87.8% December pricing rather than with Williams - and it was consistent with those numbers on Tuesday too, which is why nothing about it has changed. What has changed on Wednesday is that three voting hawks are now on the record, and one of them had been missed.
Delta is each target's distance from the 4,143.20 level at 17:55 GMT+8 on October 14, expressed in percent - the convention used in the 10-11 through 10-14 editions - and is not a change versus a prior published target. Because this edition has no New York settlement, the reference level is a live European-session price and the deltas will move again once New York trades. Morgan Stanley's 4,000 entry is a stated floor, not a target, and is excluded from the 4,625 median. JPMorgan's 4,500 desk quarterly average and its 6,000 longer-dated target are the same house reported separately, never merged. Citigroup is absent from the table because it published no number - only an exposure reduction - and inventing one to fill the row would misrepresent it. Ratings are reproduced as characterised in the source notes; this column has not independently verified each bank's current formal rating. Not investment advice.
CHAPTER 05
Levels and watchlist: 4150 was tested and given back twice today, 4200 was never bid, and the only catalyst left is 20:30 GMT+8
A fourth session without a hold above the round number; a day range of 29.35 dollars made almost entirely downward; and the sell-side panel that has not re-rated in the two hours before the print that decides it.
THE DATA
| Type | Level | Basis and implication |
|---|---|---|
| Strong resistance | 4200 / 4170.35 | Four consecutive sessions have failed to hold the round number, and today it was never bid - the European high was 4170.35, some 30 dollars below. A level that has to be defended by sellers stops being interesting the moment buyers stop arriving. A market that has stopped bidding the round number has inherited it as resistance without a test, which is a quieter and far more durable configuration than a fourth rejection. Above: 4220 as the Bollinger midline (FX678), 4200-4220 as the London 20-day average band with 4250 next and the 100-day near 4250-4260 depending on source, then 4280-4320 as UBS's technical-repair band. A sustained close above 4220 is the trigger to revisit the October highs. |
| Mid-range | 4141.00 / 4150 | 4150 is the only number that matters between now and 20:30 GMT+8. Today's low was 4141.00 (Cngold) against Tonghuashun's 4142.70 at 17:32 - the level has been tested intraday and given back, once in the low print and again by the 4143.20 close. Three independent desks mark support converging immediately below: Guangzhou Jinkong at 4150 then 4100 with the October 7 low at 4066; Tonghuashun at 4080-4100; FX678 at 4120 with 4000 beyond. The disagreement between the desks is 4150 versus 4100-4120, and gold is currently 6 dollars above the top of it. Holding 4150 keeps the range; losing it on a soft CPI would be a genuine anomaly worth trading, and losing it on a hot one is merely the expected path. |
| Strong support | 4060 / 4100 | This level has to do the work the ETF used to do. SPDR's 1.215-tonne decline is the first net outflow in the visible series, and FX678 warns concentrated redemptions below 4000. Three desks converge on 4060-4120 - Guangzhou Jinkong's 4066 October 7 low and 4100, Tonghuashun's 4080-4100 and 4000, FX678's 4120. Below that: UBS's stated support band 3,990-4,030, Morgan Stanley's Gower calling 4,000 "a very strong floor", HSBC's remaining-year range floor at 3,800. Four independent houses bracket the same zone from four directions, and the PBoC's 740,000-ounce September purchase was made in the 4,100s. Between 3,990 and 4,030 is where this column would stop being bearish and start calling it an accumulation zone - and note it would only be reached on a hot print plus a hawkish repricing, because a soft print sends gold the other way entirely. |
| Strong resistance | 4280 / 4320 | UBS's technical-repair threshold of 4,280-4,320: sustained closes above confirm the pullback as consolidation within the uptrend rather than a reversal. It sits above the 100-day near 4250-4260 and below the 50-day at 4334.80 - the geometrically correct place for a trend-repair signal. The panel median of 4,625 is 481 dollars above spot and has not moved in the two hours before the print that decides it. Gold needs 4150 to hold, 4200 to turn, 4280 to be repaired and 4335 to be in a trend again. Four hurdles, in that order, and none of them is being approached tonight. |
- 120:30 GMT+8 - September CPI, and the asymmetry is not where the morning assumedConsensus: headline +0.6% m/m (from +0.4%), 3.6-3.7% y/y (from 3.4%); core +0.2% m/m (from +0.3%), 2.5% y/y (from 2.4%). Unchanged from the morning edition and still three independent readings that cluster - the WSJ survey via Sina, Etnet's calendar, and Taurex. What has changed is the starting point. October hike odds have moved 20.5% -> 22.7% and December at-least-one to 87.8%, so the bar for a dovish surprise is higher than it was at 08:00. A core print at exactly 0.2% that merely confirms consensus is now a hawkish-consistent outcome, not a dovish one.
The honest bull argument is exhaustion, and it is on the record. Tonghuashun's desk: at least three further hikes are priced through end-2027 and there is little room to revise higher, so this pullback is an opportunity to build positions. Deutsche Bank's Tim Baker: the risk seems tilted toward a downside inflation surprise that would re-price a dovish Fed. Both arguments point the same way - and both are arguments about the second derivative rather than the print.
The bear argument is that one soft print is not enough. Hammack's early-warning language - companies planning price increases beyond input cost growth as the mark of inflation thinking becoming entrenched - describes a mechanism that a single monthly core number cannot reverse. Two voting members independently name AI data-centre electricity and component demand and AI-infrastructure bond issuance as price pressures. Expect a relief bounce on a soft number and a test of 4180-4200, per the evening note, but do not expect the repricing to be permanent until the second confirmation.
Also before the print: Lagarde 16:30 and Bowman 17:40 GMT+8 - Bowman falls two hours ahead of the data and is the last voice before it. Then Barkin 01:30 and the Beige Book at 02:00 GMT+8. A speech risk sits between the last speaker and the number. - 2The morning edition's Hammack error, and why a two-source correction is not a bookkeeping noteThe 08:00 edition listed Hammack as "carried forward unchanged, no new information this session" and counted the board as 2-1 hawkish. Her NABE keynote was relayed overnight by Xinhua Finance, Cailian Press and Guangzhou Jinkong Futures with an unconditional tightening call: "the economy needs to tighten further", smaller adjustments made relatively quickly rather than waiting and risking a larger correction later, and a restrictive stance is appropriate. She is a 2026 voting member. The count is 3-1, not 2-1.
Why this is stated in the body rather than fixed quietly. A reader comparing the 08:00 and 18:00 editions will see a changed vote count with no explanation and have to assume one of two things - that the original was careless, or that it was massaged. Both would be wrong. The morning read was assembled while Hammack's speech was circulating only as a carried-forward note; the overnight relays carry the actual quotes. Wednesday's tape - gold -18 dollars, October odds +2.2pp - is the Hammack speech being priced. That is why the correction matters: it is the mechanism of the move, not an errata.
And the substantive part of her speech had not been read at all this morning. Her inflation list runs beyond energy to tariffs, AI data-centre electricity and component demand, and the surge in AI-infrastructure bond issuance. She flags companies planning price increases beyond input cost growth as an early warning of inflation thinking becoming entrenched. Barkin and Goolsbee name AI as inflationary today as well. Three voting members on the record tying the AI buildout to current prices is a structural higher-terminal-rate argument, and it is the argument that survives a soft CPI. The morning edition treated AI as a 2027 productivity offset because Williams framed it that way; the hawkish bloc frames it as a 2026 cost, and the market has been listening to the second group. - 3Day three of oil up and gold down - and the haven bid has left both CHF and gold on the same dayWTI probed 95 dollars intraday, up about 0.93%, with Brent recovering to 106.75 (Tonghuashun, 17:32 GMT+8). Gold fell 0.43%. That is three consecutive sessions of oil rising and gold falling. The transmission is unremarkable and fully specified: oil raises headline inflation, which raises the expected terminal rate, which raises the real rate, which raises the opportunity cost of a zero-coupon asset. Tonghuashun's technical note adds the second half - a firm dollar means non-US buyers also pay a currency translation cost on top. Dual squeeze, not a single one.
Supply side is tightening on every measure available today. Saudi Aramco's Nasser: global commercial onshore oil inventory is below 6 billion barrels and may need two years to normalise. OPEC+ has confirmed the 2.2 million bpd voluntary cut runs to end-2026, and global commercial crude inventories have fallen for five consecutive weeks to well below the five-year average. Houthi missile exchanges with Saudi Arabia continue and Hormuz shipping risk stays elevated. WTI's daily Bollinger bands have shifted from closing yesterday to opening upward today.
And the more interesting item: the franc declined too. USD/CHF weakened from Tuesday's 0.8325 to around 0.8335 while gold fell. This column flagged CHF strength as an outlier divergence on Monday and Tuesday. Today both legs gave it up simultaneously - the haven bid has not vanished, it has moved. Where a haven bid goes when it leaves gold and the Swiss franc on the same afternoon is the question this column cannot answer from the tape available, and saying so is more useful than guessing. - 4The IEA report landed at 16:00 GMT+8 and OPEC cut demand again - the cost-push configuration is intactOPEC secretary general Haitham Ghais cut the 2026 global oil demand growth forecast to 400,000 bpd from 1.4 million at the start of the year, while raising 2027 for a sixth consecutive month to 2.55 million bpd, arguing deferred demand shows up in 2027. A separate relay of the same cut has it at 180,000 bpd down from 380,000 - recorded as a claim and not reconciled here.
The trade reads through CPI, and it reads through it as a cost push. A demand downgrade combined with a supply shortfall, no refinery relief and diesel at +71% y/y against gasoline's +41% (Wood Mackenzie, October 13) is a configuration that raises headline without raising real demand - precisely the environment in which a hiking cycle hurts gold, and precisely why Collins named diesel by name. The headline jump to 3.6-3.7% is arithmetic. Core is the only part of tonight's print that carries information.
Australia's RBA hiked in September on rising upside inflation risks while flagging Middle East uncertainty, per the minutes published today. That matters more than it looks: a G10 central bank tightening into an energy shock while the Fed debates a single hike means the policy differential is moving against AUD regardless of what the dollar does - and AUD/USD's softness today despite a flat dollar is the proof. The energy shock is producing global tightening patterns, not just a gold-specific squeeze. - 5The dollar held above 102 and nobody has explained why - the missing piece in today's tapeAfter closing below 102 on Monday and Tuesday, DXY held above the handle through the full European session. The technical note in circulation puts it near 102.50 intraday with a daily pivot at 102.156, a 101.56-102.82 band, the upper Bollinger band at 102.95 capping a move below the year-high at 102.61, the 100-day at 100.40 and the lower band at 100.05. RSI near 64.5 - not oversold, buying still dominant - with the MACD above zero and a narrowing histogram. UOB is reported as seeing the dollar's upward momentum weakening and favouring consolidation or a mild pullback.
The candidate explanation is Williams, and it is counterintuitive. The dovish vice chair is the one whose remarks pulled the index down to 102.16 on Tuesday. Yet the index rose today. The most coherent reading is that the market read Hammack as a partial offset - not a counterweight to Williams, but a voting member who does not accept his 2027 endpoint. A vice chair's ceiling plus a voter's floor is a narrower expected path than either statement implies alone, and the dollar prices the range rather than the ceiling.
Deutsche Bank's Tim Baker frames it in one sentence: US inflation is indeed too high and the Fed does need to hike, but inflation has not been hotter than expected and the risk seems tilted toward a downside surprise that would re-price in a dovish Fed. That is a hawkish conclusion carrying a dovish tail - and it is the configuration in which a dollar holds its level going into a data print. Trading a dollar that is standing still ahead of CPI carries a different risk profile from trading one that is falling into it, and gold has just lost the offset without gaining anything in return. - 6Tomorrow: retail sales, and the institutional-risk channel nobody is pricingThursday October 15: US September retail sales. The morning edition flagged a structural calendar conflict on the PPI date - Tonghuashun's weekly calendar places it October 16, ATFX places it October 15 - and that conflict is not resolved in this edition either. Verify the PPI date against the release calendar before positioning for Thursday or Friday.
There is still no Fed chair on the calendar. Chair Warsh has been largely absent from public policy commentary, and the 11/05 closed-door hearing on Cook's removal remains the institutional event with the largest tail. The signals are coming from second- and third-ranking officials and they pointed in opposite directions 24 hours ago - which is exactly why Hammack's overnight speech moved October odds by 2.2 points and Williams's dovish interview did not move them at all. When the vice chair's dovishness is priced at zero and a regional president's hawkishness is priced at 2.2 points, the market is telling you where it locates the decision.
And the item to put on the register: Waller in print today said "this administration poses a threat to Fed independence. So far the threat has not truly materialized." A sitting governor naming political pressure as a term-premium channel is materially different from routine independence rhetoric, because Williams has just told the market that term premium and supply are what set the long end. If independence risk becomes a priced term-premium factor, real yields can rise for reasons the CPI print does not control - which makes it the one bear argument that survives a soft number. Slow channel, not tonight's channel. On the register, not in the position.
Sources · 28›
- 1Cngold live spot gold quote (open/high/low/prev close, refreshed 17:53 GMT+8)
- 2Tonghuashun evening note, 17:32 GMT+8 (JinTou Caijing evening channel)
- 3Tonghuashun futures morning note, 10/14 (FedWatch, SPDR holdings, 南华 viewpoint)
- 4Jin10 (FedWatch relay, SPDR holdings, base metals)
- 5FX678 gold report, 10/14 10:25 GMT+8
- 6Guangzhou Jinkong Futures precious metals daily report, 10/13 data published 10/14
- 7Xinhua Finance (Hammack NABE quotes, Waller on AI and Fed independence)
- 8Cailian Press via Tencent (Hammack NABE speech)
- 9Bloomberg via Sina Global Markets (BOJ inclination, DXY, FX closes)
- 10Bank of China foreign exchange daily (10/13 closes)
- 11Xinhua / Cailian Press New York closing fixings, 10/13
- 12Sina Finance early-knowledge Treasury curve table (10/13 closes)
- 13International Financial News ifnews multi-asset table
- 14CME FedWatch via Tonghuashun and Jin10 (10/14 morning snapshot)
- 15TD Securities US rates strategy (Molly Brooks, CPI risk asymmetry)
- 16Deutsche Bank Americas FX research (Tim Baker, inflation surprise risk)
- 17NABE annual conference keynote, Beth Hammack (via Xinhua Finance, Cailian Press, JRJ)
- 18UBS precious metals research (targets, correlations, support bands)
- 19Goldman Sachs, Deutsche Bank, Standard Chartered, HSBC, Bank of America, JPMorgan and Morgan Stanley research notes
- 20OPEC World Petroleum Congress statements (Haitham Ghais demand revision)
- 21Saudi Aramco CEO Nasser on onshore inventory
- 22Wood Mackenzie AAA diesel data (via Guangzhou Jinkong)
- 23Reserve Bank of Australia September minutes (via Sina Global Markets)
- 24UOB and Chinese-desk technical commentary on DXY and gold (via Tonghuashun evening channel)
- 25IEA Monthly Oil Market Report (16:00 GMT+8)
- 26SPDR Gold Trust daily holdings, report as of October 13
- 27Federal Reserve calendar and FOMC calendar (Oct 27-28 meeting)
- 28Economic calendar cross-checks: Etnet, ATFX, Tonghuashun