Memory Pulls Back for the First Time in Ten Weeks While the Oil Premium Gives Back Two Thirds — and the Risk It Prices Did Not Go Away
1Week-over-Week Comparison — Segment Changes & Drivers
Comparison window: W44 (2026.10.23–10.30) vs W45 (2026.10.30–11.06). Risk levels are unchanged at 7 red / 4 yellow / 2 green, but the risk forms flipped on two fronts: memory spot pulled back for the first time in ten weeks, and the oil premium gave back two thirds of its geopolitical content — without the underlying risk moving at all. The week in one sentence: a political shock repriced everything except the thing that was actually happening on the water.
| Category / Metric | Last week (W44) | This week (W45) | Direction | What changed |
|---|---|---|---|---|
| 🆕 Macro — US midterms Single driver |
Dec hike odds 80%+ | Dec hike odds ~66% | 🆕 Blue-wave trade | The single largest macro variable of the week, and the first cause of almost every move in this report. The Democrats swept both the House and the Senate on Nov 3. Consequences: crude fell 8% in a single session on Nov 4 to a two-month low; December rate-hike odds fell from above 80% to around 66%; the 10-year Treasury yield fell 10bp on the week to 4.84% (Treasury official, Nov 5); and US AI-chain equities crashed on Nov 5 — Nasdaq -5.5%, S&P 500 -4.3%, Micron -17%, SanDisk -22%, SK hynix ADR -13%, NVIDIA -9%, AMD -17%. Polysilicon futures rose +3.59% in a single session on a friendlier overseas renewables-subsidy expectation. Read every other number in this table as downstream of one political outcome. |
| Memory DDR5 16Gb (2Gx8) Spot session |
USD 61.50 | USD 60.667 (-1.36%) | ↓ First fall in 10 weeks | Nov 5 session USD 60.667 (-1.36% for the week, -1.09% single session) — the first weekly decline in ten weeks, after three consecutive weeks above USD 60. On a 30-day basis +3.00%, year to date +451.7%. Long-cycle: 55.433 (Sep 18) → 61.50 (Oct 30) → 60.667 (Nov 5), +9.47% over the period. The character of this decline matters: it is not a demand signal. It coincides exactly with the Nov 5 selloff in listed AI equities, and original-equipment capacity remains sold out — so this is a valuation-driven profit-taking episode in the listed AI complex passing into the spot tape, not a change in physical demand |
| Memory DDR4 8Gb (1Gx8 3200) Spot session |
USD 46.864 | USD 46.036 (-1.77%) | ↓ 2nd week, wider | Nov 5 session USD 46.036 (-1.77%, -0.46% single session) — a second consecutive weekly decline, and a wider one than last week's -0.52%. A negative week is not a downtrend: 43.637 (late Aug) → 46.143 (Sep 18) → 46.321 (peak, Oct 2) → 45.971 (Oct 16) → 47.107 (Oct 23) → 46.864 (Oct 30) → 46.036 (Nov 5), ten weeks +5.50%. And note the divergence inside the complex: DDR4 8Gb fell 1.77% while the white-label DDR5 tier rose 1.51%, which is the signature of a positioning unwind rather than a demand downgrade |
| Memory DDR5 16Gb eTT (white-label) Spot session |
USD 29.80 | USD 30.25 (+1.51%) | ↑ Only rising tier | USD 30.25 (+1.51%) — the only memory tier that rose this week. For four straight weeks this white-label tier outpaced branded particles; this week it rose while branded DDR5 fell 1.36% and branded DDR4 fell 1.77%. That is a two-sided signal: the spot tape is being sold for positioning reasons in the listed names while unbranded supply, which is priced closer to actual allocation, held firm. Read it as a market-quote signal, not a fundamental one |
| Memory DDR4 16Gb (2Gx8) Spot session |
USD 82.796 | USD 81.679 (-1.35%) | ↓ Range breaks | Nov 5 USD 81.679 (-1.35%, -0.30% single session) — breaking the three-week USD 82.8-83.7 range; 30-day -2.35%, year to date +691.6%. DDR4 16Gb eTT 13.475 (-1.64%). This tier had been the most stable price point in the complex for three weeks, so its break alongside the rest is part of the same market-wide repositioning rather than a tier-specific story |
| Memory module DDR5 RDIMM 32GB Module quote |
USD 2,165.00 (10/19 quote) | USD 2,165.00 (10/26 quote) | → Flat | The Oct 26 module session matched the Oct 19 quote exactly at USD 2,165.00 — a 7-day cross-session comparison, not a weekly change. The substantive point is that module pricing continues to rise more slowly than die pricing, so module makers are still working through previously-priced particle increases and cost pass-through retains a lag. Per the pricing-caliber rule, module and die quotes carry separate session dates and are not cross-derived |
| 🆕 E-fabric & e-yarn Market quote |
7628 at RMB 13.5-14/m | 7628 at RMB 14.6-14.65/m | 🆕 Up again | The most important upstream development of the week, and it happened through the market-quote mechanism rather than a price letter. The Nov 3 monthly quote moved the entire chain: 7628 heavy cloth to RMB 14.6-14.65/m (+1.5/m, about +11.2% month on month); G75 e-yarn to RMB 28,750-28,850/t (+3,000/t, about +11.2%); 2116 thin cloth to RMB 20.2/m (+2.5); 1080 thin cloth to RMB 20.75-20.8/m (+2.5-2.6); E225 yarn to RMB 71,500/t (+10,500-12,500); and D450 fine yarn to RMB 147,500/t (+27,750). No new price letter was issued this week — but under a market-quote mechanism the price rose anyway, and the distinction between "no letter" and "no price change" matters enormously here. 7628 is now more than 275% above its Sep 2025 level of RMB 3.9/m, and has clearly passed both prior historical peaks (about RMB 7 in 2017 and RMB 9 in 2021). Industry inventory is described as close to zero, and producers expect the high cycle to run at least through the end of next year |
| 🆕 Energy — Brent / WTI EIA fixing |
Brent 128.12 / WTI 105.25 | Brent 107.39 (-16.18%) | ↓ Premium given back | EIA official Wednesday fixing: Brent USD 107.39/bbl, -16.18% for the week — a six-month-to-date low of two months — while WTI fell 4.93% to USD 100.06. The consequence is the Brent-WTI spread collapsing from USD 22.87 to USD 7.33, back inside the normal 3-8 band (cross-check: 107.39 − 100.06 = 7.33). Last week's structural signal — the premium concentrated in the European physical end — has resolved, because roughly two thirds of the premium has been given back. But the driver was political, not economic: US-Iran talks resumed through Qatar on Nov 2, and the midterms on Nov 3 delivered a single-party sweep. Implication for procurement: keep Brent-linked clauses on plastics, housings and inks but do not bank a further reduction this period, and recompute Q4 fuel surcharges against the lower price |
| 🆕 Geopolitics — attacks continue Risk not resolved |
4 tankers struck in one day | 5+ vessels in week one | 🆕 Unchanged | This is the week's central contradiction, and it should be read explicitly. The premium collapsed on the election and the talks, but UKMTO recorded at least five vessels reporting incidents from Sunday Nov 1 onward — a frequency comparable to October's 35, the highest monthly total since the conflict began. On Nov 4-11/6 three tankers were damaged in the Gulf and the Strait, with two striking mines and catching fire. Strait throughput is running below 60% of capacity, while loading volumes at Gulf and Omani ports are 36% above pre-war levels — meaning vessels are being routed around or waiting, not flowing freely. On Nov 5 Iran submitted a nuclear proposal through Qatar, with seven conditions stated in advance; on Nov 5 the president said there was no plan to strike Iran, while on Nov 6 reporting indicated US military plans are ready with thousands of air and naval personnel standing by. Conclusion: the premium has largely been given back but has not been cleared, and it can be re-injected. Keep the three-lane war-risk quote window open. |
| 🆕 Tanker cost structure OPEC official |
No BDTI / VLCC series | 30% of oil cost | 🆕 Gap filled | An official read that closes this report's most important data gap. OPEC's own assessment: on Middle East-to-East Asia routes, VLCC freight and insurance together now account for roughly 30% of the total delivered crude cost, against about 5% at the end of January. Per barrel, freight went from USD 3.18 to USD 37.13 (roughly 12x) and insurance from USD 0.06 to USD 12.29 (over 200x), a combined USD 49.42/bbl. OPEC's own conclusion is that tight shipping capacity explains only part of the increase — the overall deterioration in the maritime risk environment is the main cause. That is the cleanest available quantification of why delivered energy costs can stay elevated while the crude benchmark falls |
| 🆕 Middle East port congestion Record |
Not measured | 8.86 days waiting | 🆕 Record | Linerlytica data (Oct 30): average waiting time at Fujairah has reached 8.86 days and at Khor Fakkan 7.25 days, with Jebel Ali around 3 days; transshipment from Khor Fakkan and Fujairah to Jebel Ali and other hubs, which used to take about 5 days, now takes 12-15 days. Yard utilization is at 84% at Salalah and 76% at Jeddah. Implication: the delay is now showing up in schedule reliability and inventory-in-transit cost, not yet in headline freight rates — which is why the container index has stayed flat while the physical situation deteriorated |
| MLCC lead-time listings Negotiations open |
61 weeks | 61 weeks (held) | → Held | The distributor listing of 61 weeks held (iM Securities, Oct 21) — the constraint is not easing. November opens 2027 Q1 price negotiations, with Japanese vendors widely expected to follow Samsung E-M's increase. And a new round from a second major is now in its landing window: high-cap products at more than 50%, mid-to-low-cap at 10-30%, planned for late October through November, with the pricing method shifting to customer-by-customer quotes. Implication: the per-part-number, per-account notification channel is now the primary price channel — a blanket percentage framework no longer applies |
| 🆕 Aluminum electrolytic lead times New |
12-20 weeks | 40 weeks | 🆕 Lengthening | A new lead-time signal on a component that had been treated as a price story. Channel monitoring in early November puts 400-450V industrial-voltage snap-in and bolt-type aluminum electrolytic lead times (105°C / 125°C grades) at 40 weeks for Nichicon, Rubycon and Panasonic, with institutional forecasts of 48 weeks or more by 2027 Q1 — attributed to aluminum foil corrosion and the capital intensity of the forming line, which means low supply elasticity. Note that channel readings differ by scope this week, ranging from roughly 18-22 weeks on general grades to 40 weeks on the 400-450V industrial segment; treat 40 weeks as the figure that governs year-end locking decisions, and ask each supplier to confirm the lead time applicable to the specific part number and voltage grade rather than accepting a catalogue range. Implication: for power supplies, inverters and energy-storage PCS projects, this is a year-end lock-in decision, and it should be made on lead time rather than on the aluminum price |
| 🆕 Tantalum pricing New round |
4th round in execution | Further +20-30% | 🆕 New round | A further round landed on Nov 1, with one major increasing T520 / T521 / T530 series a further 20-30% (media-reported caliber; the round attribution differs from the previously reported "fourth round at +25-40%", and should be verified against authorised distributor quotes), while a second major followed on polymer tantalum at +15-30% and on manganese dioxide grades at +10-20%. The longest lead time reading moved to 84 weeks, with the 40-52 week caliber maintained. For context, tantalum ingot spot was USD 6,920/kg in June 2026, up 158% from end-2025. Implication: tantalum is now the tightest lead-time item in the passive complex, and the "old price window closed on Nov 1" note means the pre-increase quote is no longer available |
| Metals — industrial complex Futures |
Cu 108,050 / Al 23,395 / Zn 25,850 / Ni 117,190 | Sn 342,760 / Ni 122,380 / Cu 104,700 / Al 23,080 / Zn 25,370 | ↓ Broad decline, one gainer | A liquidity shock hit the industrial complex: the Nov 6 morning close showed Shanghai tin RMB 342,760/t (-5.02%), copper RMB 104,700/t (-3.10%), zinc RMB 25,370/t (-1.85%) and aluminum RMB 23,080/t (-1.35%), with nickel the single gainer at RMB 122,380/t (+4.43%) on Indonesian enforcement disruption and short covering. LME tin fell 5.48% to USD 46,100/t and LME copper 3.58% to USD 13,774/t — internal weakness and external weakness together this time. Tin is now down for a second consecutive week, and the driver is identified as the AI-chain correction plus a liquidity shock amplifying the move, not a change in supply. Nickel matters more than its weight suggests: it is the MLCC internal-electrode input, so electrode cost pressure turned back upward this week |
| Precious metals Rebound |
London Ag 49.107 / Au 3,875.19 | London Ag 51.78 / Au 4,088.12 | ↑ Support recaptured | Precious metals rebounded as the rate outlook eased: London silver recovered to USD 51.78/oz (+5.45%), taking it back above the USD 50 level that it broke last week, with Shanghai silver at RMB 12,409/kg (+2.61%) and London gold at USD 4,088.12/oz (+5.49%); Shanghai gold RMB 870.02/g (+1.74%). The driver is the election outcome reducing December-hike odds from above 80% to around 66% and a weaker dollar. Implication for procurement: the silver-bearing negotiation window that opened last week has already narrowed again — conductive silver paste, relay and connector silver contacts, plating layers and silver-bearing solder should be re-quoted this week rather than left to drift. And the structural point from last week still holds: silver price is not the long-term variable, silver consumption is |
| Dry bulk & freight Spot |
BDI 2,511 (10/23) | BDI 2,694 (11/4) | ↗ Rebound | BDI rebounded to 2,694 on Nov 4 — up 7.29% against the Oct 23 reading of 2,511, which is a 12-day comparison and not a weekly move. Capesize led with +7.22%, and the Panamax index ended a 15-session slide. On container lanes, freight-agent readings for Nov 6 put North Europe around USD 3,595 and the Mediterranean around USD 4,375 per 40ft, pushed up by northern European congestion, the French port strike and carrier increases, while US lanes fell 5-6% on the week (US West roughly USD 7,500-8,300, US East USD 9,631-9,900 per FEU) as the peak season ended and November increases did not land. A major carrier raised full-year EBITDA guidance for a third time to USD 14.0-14.5bn from USD 10.5-12.5bn, noting Q3 spot rates averaged 46% above Q2 and pointing to port congestion and vessel turnaround deterioration as a major cause of the high rate level |
| 🆕 Auto electronics demand Basis switch |
+13.6% (Sept NEV sales) | -25% (Oct passenger retail) | ↓ Basis switched | An important change of measurement basis rather than of direction. October passenger-vehicle retail was 1.693 million units, down 25% year on year and up 5% month on month, with new-energy retail at 1.148 million units, down 10%, and penetration at 67.8%. The year-on-year automotive electronics reading therefore switches from the September +13.6% (based on new-energy vehicle sales) to -25% (overall passenger retail), and vehicle-grade demand expectations must be marked down accordingly. The structural cause of the capacity squeeze — AI taking the allocation from automotive-grade parts — is unchanged, so the guidance is: negotiate automotive-grade parts on annual framework prices and do not place expediting orders |
| Power devices Window |
Unannounced, 3rd week | Unannounced, 4th week | → Held | The Taiwan pricing window has now been open for four consecutive weeks, and the internal reading is that Q4 adjustment odds are rising while the timing keeps slipping. One major's +5-15% (effective Oct 10) remains in execution and under verification. Implication: if the announcement slips past the midterms, you negotiate from a weaker position with a 30-50+ week lead time behind you. Also relevant this week: the Section 232 refined-copper decision is now 39 days overdue and still suspended rather than cancelled |
| USD / RMB FX |
6.7138 (10/30 parity) | 6.7168 (11/6 parity) | ↘ Slight depreciation | Nov 6 central parity 6.7168, a 30bp depreciation against the Oct 30 endpoint of 6.7138 — the first depreciation after three consecutive weeks of appreciation, and a weekly move of only 0.04%. Implication: the local-currency improvement on dollar-denominated imports pauses, and the natural hedge against memory, MLCC and analog price increases weakens. Keep FX-linkage clauses on import quotes. Note also the rate path: the 10-year Treasury yield fell 10bp on the week to 4.84%, with the 2-year at 4.46% and the 30-year at 5.25% |
2Five Forces Driving the Market This Week
3Risk Map — 14 Segments × Three Time Windows
Color block = supply / price tightness. Bold outline = moved this week. Red = rising / high risk, green = falling / loose supply-demand (China market convention). Overall rating unchanged at 7 red / 4 yellow / 2 green; the risk forms flipped on memory spot, energy and PCB inputs.
| Segment | 1H 2026 | Current (Nov 6) | 2027 Outlook | |
|---|---|---|---|---|
| Memory DRAM / DDR5 | High | Red alert | Red alert | -1.36%, first fall in 10 wks — valuation, not demand |
| NAND / eMMC | High | Elevated | High | 4Q26 NAND +15-20%; capacity sold out to 2029 |
| MLCC Ceramic Capacitors | Red alert | Red alert | Red alert | 61-week listings hold; 2027 Q1 talks open; per-customer quoting |
| Tantalum Capacitors | High | Red alert | Red alert | New +20-30% round Nov 1; longest lead time 84 weeks |
| Aluminum-Electrolytic / Film | Yellow alert | Red alert | High | 400-450V grades to 40 weeks; 48+ wks forecast for 2027 Q1 |
| MCU Microcontrollers | High | Red alert | Red alert | Industrial MCU 72-80 wks, booked into 2028 |
| Analog / Power Management | High | Red alert | Red alert | TI 2027 annual negotiation open, notified per part number |
| Power Devices | High | Red alert | Red alert | Taiwan +10-15% unannounced, 4th week — window still open |
| PCB / CCL | Yellow alert | Red alert | Red alert | 7628 +11.2% again; Nov CCL +15%/+25% in execution |
| Copper / Base Metals | High | High | High | Liquidity shock: tin -5.02%, Cu -3.10%; nickel +4.43% |
| Freight / War-risk | Yellow alert | High | High | Freight+insurance 30% of oil cost; congestion at record; 301 fee Nov 9 |
| Connectors / RF | Yellow alert | Elevated | Elevated | No new disclosure; 224G certification still 12-18 months |
| Precious Metals | Watch | Watch | Watch | Silver recaptures $50 on softer hike odds — window narrowing again |
| Origin / Compliance | Watch | High | Elevated | 301 port fee expires Nov 9; Taiwan list still in force |
4Memory — The First Weekly Decline in Ten Weeks, and Why It Is Not a Demand Signal
DDR5 16Gb fell -1.36% to USD 60.667 — its first weekly decline in ten weeks — and DDR4 8Gb fell -1.77% to USD 46.036, a second consecutive fall with a wider margin. But the white-label DDR5 tier rose 1.51% while branded DDR5 fell, the Nov 5 session coincided with a 5.5% Nasdaq drop and a 17% single-day fall in one memory maker, and original-equipment capacity remains sold out to 2029. That combination identifies this as a positioning unwind in the listed AI complex passing into the spot tape, not a change in physical demand.
Baseline index
YTD +185%
YTD +275%
YTD +451.7%
YTD +775.7%
4BAI Capacity Allocation — Why Everything Else Is Getting Squeezed
Memory prices are not rising because end demand is strong — notebook shipments are falling and PC-maker inventory is stretching. They are rising because HBM, advanced packaging and AI-grade components are absorbing capacity, and the marginal tier loses. This matrix tracks where AI orders are being absorbed and what is displaced as a result. It is a standing weekly tracker because it is the mechanism behind the whole price complex, not a one-week event.
HBM makers
Domestic memory
Niche memory suppliers
Leading-edge foundry
Analog and power management
MLCC makers
Power device makers
High-speed connectors
Substrate, high-layer PCB and CCL
5PCB, CCL & Metals — Fabric Jumped on a Market Quote, Metals Fell on a Liquidity Shock, and Neither Touched the Floor
Two things moved at once and in opposite directions. Metals fell broadly — tin -5.02%, copper -3.10%, zinc -1.85%, aluminum -1.35% — on an AI-chain correction and a liquidity shock. Glass fabric and e-yarn jumped again through the monthly market quote, with no price letter issued. And the input-cost floor did not move: copper concentrate TC remains at a near-decade low and high-end foil processing fees are still 6-7 times ordinary foil, so a falling copper price has not become a falling input cost.
| Node | This week (Nov 6 close) |
|---|---|
| E-yarn & e-fabric | The chain moved again through the monthly quote mechanism, not a price letter. Nov 3: 7628 heavy cloth RMB 14.6-14.65/m (+1.5/m, about +11.2% month on month); G75 e-yarn RMB 28,750-28,850/t (+3,000/t, about +11.2%); 2116 thin cloth RMB 20.2 (+2.5); 1080 thin cloth RMB 20.75-20.8 (+2.5-2.6); E225 yarn RMB 71,500 (+10,500-12,500); D450 fine yarn RMB 147,500 (+27,750). Year to date: 7628 +275%+, G75 +121%, thin cloth +190%+. 7628 is more than 275% above its Sep 2025 level of RMB 3.9/m and has clearly passed both prior peaks (about RMB 7 in 2017 and RMB 9 in 2021). G75 inventory is described as near zero, and D450's larger increase shows AI-chain demand elasticity sitting in fine yarn. Thin cloth continues to outrun heavy cloth as product mix upgrades, and most suppliers are prioritising long-term agreement customers. This is the reason the glass chain is treated as structurally short rather than as a market that happens to be firm |
| Copper foil | SHFE copper RMB 104,700/t (-3.10%) with LME copper 13,774 (-3.58%) — internal and external weakness together this week, reversing last week's external-strength split. The temptation is to read the copper decline into foil pricing, and it should be resisted: high-end HVLP foil processing fees sit at RMB 170,000-200,000/t against RMB 26,500/t for ordinary 18-micron HTE and RMB 24,000/t for 6-micron lithium battery foil — a 6-7x generational gap, and the gap did not narrow. HVLP foil demand is forecast up more than 260% in 2026 with high-end supply short by close to 50%; the HVLP4 gap is about 1,500 tonnes in 2026 and 2,500 tonnes in 2027. Foil is priced off conversion fee and capacity, not off copper, and there is no new fee reading this week |
| Laminate (FR-4) | Two increases are now in execution at once. One major's eighth increase of the year (issued Oct 13: 2116-and-below thin cloth board material and prepreg, +15%) is under verification, and another major's November increase (M6-and-below boards +15%, prepreg +25%, reported Oct 28) is under verification. Orders are scheduled to January 2027 and the supply gap is unchanged. Note the caliber discipline: the absence of a new letter this week means the price letter channel is quiet, not that the price is unchanged — the fabric underneath rose anyway. Looking ahead, one Japanese major has been reported to be raising some conventional-grade board material by +200% with lead times extending from Jan 1, 2027 — and the qualifier is essential: that figure applies only to certain conventional glass-epoxy grades and must not be applied across high-speed grades, whose separate announcement covers a far smaller increase of around +20%, because the conventional capacity is what AI demand has displaced. Fold the conventional-grade portion into 2027 budgeting, and do not extrapolate it to high-speed laminate. |
| AI high-speed CCL (M8/M9) | One institution's forecast: AI-server CCL demand moves from 42 million sheets in 2026 to 131 million in 2028, a 77% CAGR, with average pricing rising from USD 165 to 282 to 362, roughly a 48% CAGR. Another puts the 2026 high-speed CCL market at about USD 8bn. The gap is around 70%, lead times have stretched from 2 weeks to 8-12 weeks, and leading orders are scheduled into 2027. On the platform itself: PCB value per card has doubled from about USD 400 on the previous generation to over USD 800, with 40-78 layers and CCL moving from M6 to M8 — which is the mechanism by which a laminate price move becomes a board cost move |
| Tin / solder & plating | SHFE tin RMB 342,760/t (-5.02%), a second consecutive weekly decline, with LME tin 46,100 (-5.48%) — the identified driver is the AI-chain correction plus a liquidity shock amplifying the move, not a change in supply. Procurement: do not extrapolate a supply-driven decline from a sentiment-driven one. Note also that the AI chain is the same chain whose high-end CCL demand is forecast to compound at 77% — the pullback is in the listed equity, not in the order book |
| Nickel / MLCC electrodes | SHFE nickel RMB 122,380/t (+4.43%) — the strongest Shanghai industrial metal of the week and the only gainer, on Indonesian enforcement disruption and short covering. MLCC internal-electrode nickel-powder cost pressure therefore turned back upward in the same week that MLCC lead times held at 61 weeks — the cost floor and the queue moved together. Nickel powder is priced on conversion cost, so pass-through still takes one to two quarters |
| PPE resin (hard supply constraint) | ~70% of global PPE resin is produced in Saudi Arabia, where only about 65% of capacity has been restored, and a maintenance outage is scheduled inside November — the largest single materials risk for Q4, now with logistics and insurance layered on top because the Strait attacks raise the cost of moving Saudi product. Electronic-grade PPE is up +261% year to date |
| Glass furnace (rigidity) | Once lit, a glass furnace must run 8-10 years; 2026 is a concentrated cold-repair year, with 2013-2015 firings reaching the end of their 8-10 year life and repairs running 3-6 months, while very few new lines are lit. Laminate expansion runs 18-24 months with about 12 months of customer certification — so supply has essentially no elasticity. One major has already pre-booked roughly half of the next two years' new loom capacity from an equipment supplier whose delivery queue runs to 2028 |
Three separate mechanisms, and this week each of them was visible at once. First, the metal price fell for macro reasons, not supply reasons — a liquidity shock and an AI-chain correction, which is why tin fell harder than copper despite tighter tin inventories. Second, the binding constraint upstream is glass and foil capacity, not metal: the November quote moved 7628 up 11.2% and D450 fine yarn up RMB 27,750/t in the same week copper fell 3.10%, and high-end foil processing fees are 6-7 times ordinary foil and did not move at all. Third, the floor is set by concentrate — with copper concentrate TC at a near-decade low, a falling copper price reflects a financial and macro move rather than a supply improvement.
On the demand side the bifurcation that keeps every PCB sub-tier out of the green zone persists and is now quantified better than before: the 2026 global PCB market growth forecast has been revised up to +18.8%, about USD 101.95bn, with boards above 18 layers up 79.0% and China above 18 layers up 93.5%, and AI-server high-layer and HDI are the increment. In China, 1H26 disclosed PCB expansion investment exceeded RMB 70bn, essentially all of it AI-directed, with zero traditional consumer-line expansion. One capacity figure worth holding on to: equipment lead times run 10-24 months, so capacity ordered in 2027 forms effective supply in 2028 at the earliest — the 2027 gap does not close on this capital cycle, and the glass furnace cold-repair calendar is the reason.
Two institutional views bracket the cycle: one argues the price increase has begun spreading from high-end into general grades and runs at least into 2027; the other sees it conservatively to mid-2027 and optimistically to 2028. Whichever you take, the operative rule for buyers is that no sub-tier of the PCB chain — including the lowest — carries a green rating, because the constraint is upstream capacity and not end demand.
6Segment Snapshot
Memory DRAM/NAND RED
MLCC Ceramic Capacitors RED
Tantalum Capacitors RED
Aluminum Electrolytic RED
MCU / Analog / Power RED
PCB / CCL RED
Foundry / Semi Materials RED
Connectors / RF ELEVATED
IMU / Sensors NEW
Base Metals
Precious Metals
Freight & Insurance
Downstream Demand
Year-on-year growth is refreshed from disclosed monthly and quarterly data each week and labeled with its basis. Actual and forecast calibers are never mixed without saying so, and a change of measurement basis is always stated.
| Application | Demand | Visibility | YoY | What changed this week |
|---|---|---|---|---|
| AI servers / compute | Extreme | Into 2027 | +31% TrendForce 2026 global AI server shipments (>2m units) |
Capital markets and industry fundamentals have separated, and that is this week's story. On Nov 5 the listed AI chain fell sharply (Micron -17%, NVIDIA -9%) while original-equipment capacity remains sold out to 2029 — one maker grew Q3 revenue 327.9% at an 80% operating margin with long-term agreements now covering 14 customers and more than half of 2027-2028 capacity, and DRAM and HBM effectively sold out to 2029; another grew memory revenue 427% with 2027 HBM fully sold. A major Chinese domestic DRAM maker now reports 2027 capacity already reserved by customers. Assessment: order books and capacity reservations have not changed; a valuation re-rating amplified sentiment in the spot tape. That is precisely why this week's memory decline should not be read as a demand signal. |
| Consumer electronics (phone / PC) | Contracting | 6-12 months | -22% IDC 3Q26 global phones, released Oct 14 |
No new readings this week. Maintained: 4Q26 major notebook ODM order volumes forecast down 10-15% year on year, PC-maker DRAM inventory stretched to 7-10 weeks, and handset brands continuing to cut low-to-mid-tier models. Consumer-line material should be bought on demand, not on a forward view — it is the one tier with genuine slack |
| Automotive electronics | Falling (structural divergence) | 3-6 months | -25% October passenger-vehicle retail, released Nov 5 — measurement basis switched |
The year-on-year reading switches from September's +13.6% (measured on new-energy vehicle sales) to October's -25% (measured on overall passenger retail of 1.693 million units, -25% year on year and +5% month on month, with new-energy retail at 1.148 million, -10%, and penetration at 67.8%). Vehicle-grade demand expectations must therefore be marked down. But the structural cause of the capacity squeeze — AI taking the allocation from automotive-grade parts — is unchanged, and automotive memory certification runs 2-3 years, so vehicle-grade parts cannot be re-sourced quickly even as demand softens. Guidance: negotiate automotive-grade parts on annual framework prices and do not place expediting orders. |
| Industrial control | Firm | 2-4 months | +18.9% September industrial robot output |
No new readings. European energy costs remain high — German and French industrial power above EUR 120/MWh — but this week's Brent decline of 16.18% marginally eases input-cost pressure for European industrial customers. Industrial MCU lead times of 72-80 weeks, booked into 2028, are unchanged |
| Photovoltaics | Weak → expectation recovery | 1-3 months | -36% to -49% CPIA 2026 full-year forecast, 160-200GW — forecast caliber |
New this week: the single-party sweep in the US midterms raised expectations for overseas renewables subsidies, and polysilicon futures rose 3.59% in a single session to close at 35,360 (+0.09% on the week). The fundamentals have not changed — inventory stands at 370,000 tonnes — so this is an expectation-driven rebound, not a turn. Meanwhile silver rebounded to USD 51.78 (+5.45%), so conductive silver paste cost pressure returns, reversing last week's relief |
| Energy storage | Recovering | 1-3 months | +42% CNESA September newly installed power — tracked separately from solar |
No new concentrated procurement data. Cost side: lithium carbonate futures at 91,720 (+1.30% on the week, a wide range on short covering), while battery-grade spot fell 1,150 on Nov 5 — the loose supply picture is unchanged and the green alert is retained. Note the opposite directional pressure: aluminum electrolytic lead times at 40 weeks is a far bigger issue for storage PCS than the lithium price is |
| Telecom base stations | Steady (capacity-driven) | 3-6 months | +11.7% 5.254m 5G base stations end-Sep vs 4.705m a year earlier |
No new tender announcements; the domestic carriers' 4Q26 tendering cycle that began Oct 1 is maintained. The structural watch remains: communication-grade MLCC used in optical modules has started to show shortages, so passive-component lead times in the base-station and optical-module chain need separate monitoring from the AI-server chain — and this week's 61-week listing means the two chains now compete for the same constrained capacity |
7This Week's Procurement Playbook — What to Do Now
Standing section, rewritten each week for the changes. Pull-in strategy runs parallel to cost-down / substitution paths.
| Segment — this week's move | Procurement / pull-in strategy (this month) | Cost-down & alternatives |
|---|---|---|
| Hormuz — premium given back, risk notBrent -16.18%, spread to 7.33; 5+ vessels hit in week one | Do not treat the oil decline as a risk reduction — treat it as a pricing opportunity against an unchanged risk. The premium gave back two thirds, but UKMTO recorded at least five vessels reporting incidents in the first week of November, comparable to October's 35, Strait throughput is running below 60% of capacity, and Gulf loading volumes are 36% above pre-war. Keep the three-lane war-risk quote window open and ask your broker to quote the Red Sea, the Persian Gulf and the Gulf of Oman separately, with the crude-carrier rate and the container rate distinguished. OPEC's own assessment — freight and insurance now at 30% of delivered crude cost against 5% at end-January, USD 49.42 per barrel combined — is the clearest evidence that delivered cost can stay elevated while the benchmark falls. Recompute Q4 fuel surcharges against the lower oil price, but keep the buffer plan intact. | Re-price the route rather than the commodity. Where the carrier still offers it, a Suez-return option on Europe lanes still saves 10-14 days and roughly 30% of fuel — but note the French port strike on Nov 5-7 and Middle East congestion now pushing transshipment from 5 days to 12-15 days. Negotiate a war-risk rate cap with overrun sharing rather than an open-ended rate, and hold air-freight contingency for lanes where the risk assessment has moved from probability to duration. |
| E-fabric & e-yarn7628 +11.2%, D450 +RMB 27,750/t — market quote, no price letter | This is the single most actionable item in the report, and it is a process failure waiting to happen. Fabric and e-yarn do not issue price letters — they are quoted from monthly market assessments, so a week with no letter can still contain a price jump. This week it did: 7628 to RMB 14.6-14.65/m and G75 to RMB 28,750-28,850/t on the Nov 3 quote, with D450 fine yarn up RMB 27,750/t. Action: pull your supplier's current quote for every fabric grade on your BOM this week and compare it against the November assessment, rather than against last week's price. Where a grade has moved, request the new-price order confirmation and check for phased execution. And do not let "no price letter this week" become "no change this week" in your own tracking — the two statements mean completely different things. | Where a design permits, thin-cloth and fine-yarn demand is where the tightness is, so substituting toward heavy cloth is not always possible — but the reverse is worth checking. Rationalize fabric grades where a board can move to a less constrained cloth, and qualify a second cloth supplier now: with industry inventory described as near zero and the high cycle expected to run into the end of next year, supplier concentration is the real exposure. |
| Taiwan power / MCU+10-15% planned, unannounced for a fourth week | Fourth consecutive week, and the window is now the main remaining lever. The internal reading is that Q4 adjustment odds are rising while timing keeps slipping — which is a description of a decision that has been made but not communicated. Complete the new-order negotiation with the Taiwan makers this week and secure written magnitude and effective date, so their increase does not stack passively on the +5-15% already live from the US maker. If the announcement slips past the midterms, you will be negotiating from a weaker position with a 30-50+ week lead time behind you and no alternative supply inside the quarter. Also note the Section 232 refined-copper decision is now 39 days overdue — copper-intensive lines face two open questions at once. | Cross-brand pin-to-pin on MOSFET / IGBT / SiC; package arbitrage; qualify second sources on long-lead SiC now rather than after the announcement. One major's January 1, 2027 repricing belongs in 2027 budgeting rather than in this cycle's scramble — but note it is attributed to structural costs in materials, energy, advanced packaging and logistics, all four of which are currently moving against you. |
| MLCC — per-customer pricing61-week listings held; a second major's round landing; 2027 Q1 talks open | The pricing channel changed, and that changes your process. A second major's new round — high-cap more than 50%, mid-to-low-cap 10-30% — is in its landing window and the mechanism has moved to customer-by-customer quotes. At the same time 2027 Q1 negotiations opened in November with Japanese vendors widely expected to follow. Consequence: there is no percentage and no single effective date to negotiate against, so a blanket framework applied across a portfolio will be wrong on part of it. Get per-part-number quotes in writing, with the quote validity period stated, and put your 2027 high-cap volume request in front of suppliers now rather than in December. With 61-week listings, the queue you are in matters more than the price you are quoted. | Generic 0402/0603 and consumer grades remain decoupled from AI high-cap — consumer and mid-cap spot fell about 5% in early October before stabilizing, so there is genuine slack below the AI tier. Rationalize part numbers for allocation leverage, and qualify a mainland alternative in the industrial tier — but price it at current levels, because the domestic price gap has been closing as the majors reprice. |
| Lead-time lock-insAl electrolytic 40 weeks, tantalum 84 weeks, MLCC 61 weeks | Three components crossed thresholds this week and all three should be locked on lead time rather than price. Aluminum electrolytic 400-450V industrial grades moved to 40 weeks, with institutional forecasts of 48+ weeks by 2027 Q1, on low supply elasticity from foil corrosion and the capital intensity of the forming line — for power supply, inverter and energy-storage PCS projects this is a year-end lock-in decision. Tantalum's longest lead time reading moved to 84 weeks after a further increase on Nov 1, and the old-price window closed that day, so pre-increase quotes are gone. MLCC listings held at 61 weeks. Action: convert these from "watch items" into dated purchase commitments this quarter, because each is a queue you cannot shorten later — and each is tightening while the commodity price of its main input is falling, which is exactly the combination that causes a lead time to be missed. | For aluminum electrolytic, the cost relief from a 1.35% aluminum decline is small against a 20-week lead-time extension — the leverage is in the schedule, not the unit price. Consider capacity reservation and dual-sourcing on the specific industrial voltage grades. For tantalum, the standing position that some makers will not accept new long-term agreements and only execute existing orders makes contractual coverage the only available protection. |
| Memory — read contracts, not spotDDR5 -1.36%, first fall in 10 weeks, while capacity stays sold out to 2029 | This week's decline is a sentiment event and it is the wrong thing to react to. The Nov 5 session coincided with a 5.5% Nasdaq fall and a 17% single-day decline in one memory maker, while original-equipment capacity remains sold out to 2029 and long-term agreements now cover more than half of 2027-2028 capacity. Do not treat the spot pullback as a buying window, and do not treat it as a signal to delay either — 4Q26 PC DRAM contracts remain +13-18%, November contracts are expected above October, and a major domestic DRAM maker reports 2027 capacity already reserved with contract liabilities about 15 times end-2025. Action: convert reservations into confirmed POs, and where the white-label tier is the source of a part, lock it there — it rose 1.51% while branded names fell, which is the market telling you where the physical constraint actually sits. | Continue DDR4→DDR5 migration where the design allows, but size the saving against current numbers — the white-label-to-branded gap narrowed again this week. For consumer-facing products, the one place with genuine slack is below the AI tier: 4Q notebook ODM orders are down 10-15% and consumer-grade MLCC spot has already fallen. Do not extend an AI-grade part number into a consumer design just because it is already qualified. |
| Silver-bearing partsSilver rebounded to $51.78, +5.45% — window narrowing again | Last week's window was short and it is closing faster than expected. London silver recovered USD 51.78/oz (+5.45%), taking it back above the USD 50 level it broke only last week, as the election outcome cut December-hike odds from above 80% to around 66% and weakened the dollar; Shanghai silver is RMB 12,409/kg (+2.61%). Re-quote silver-bearing parts now — conductive silver paste, relay and connector silver contacts, plating layers and silver-bearing solder — and ask specifically for the silver component to be re-priced rather than for a blanket discount, so the concession is sized correctly and does not leak into non-silver lines. And be precise about why the window opened: this is a rate-and-positioning move, not a demand improvement, and it reversed in five sessions. | Technical substitution is the durable lever: silver-coated copper and copper plating substitution is already cutting 2026 photovoltaic silver consumption by 20%, and the same substitution is available in contacts and plating. For any silver-bearing part with a design horizon, specification change beats price negotiation over a two-year view. |
| Automotive demandYoY basis switched to -25% on October passenger retail | A change of measurement basis has moved the automotive reading from +13.6% to -25%, and it needs a process response rather than a price response. October passenger-vehicle retail was 1.693 million units, down 25% year on year. Vehicle-grade demand expectations must be marked down, and automotive parts should be negotiated on annual framework prices with no expediting orders — because while the demand side softened, the supply side did not: AI continues to take the allocation from automotive-grade parts, and automotive memory certification runs 2-3 years, so a displaced vehicle part cannot be re-sourced inside a quarter. The combination of softer demand and unchanged structural scarcity is precisely the case where expediting buys the worst price. | Use the softer demand to consolidate part numbers and rationalize the automotive bill of materials while the qualification clock is still running elsewhere. Where a vehicle-grade part is dual-sourced already, this is the window to rebalance toward the source with the shorter qualification, not to chase the spot price. |
8Risk Matrix — Likelihood × Impact
Attention allocation: act immediately in the top-right, monitor continuously in the bottom-left.