Comparison window: W43 (2026.10.16–10.23) vs W44 (2026.10.23–10.30). Risk levels are unchanged at 7 red / 4 yellow / 2 green, but energy pressure stepped up and MLCC's constraint moved further from price to allocation. The week in one sentence: a realized geopolitical event drove a European physical premium, while a rate-driven precious-metals selloff and a demand-driven tin slide moved in the opposite direction — and the memory complex simply kept grinding higher with volatility narrowing.
| Category / Metric | Last week (W43) | This week (W44) | Direction | What changed |
|---|---|---|---|---|
| 🆕 Geopolitics — Hormuz Realized |
Strike plan reported | 4 tankers struck in one day | 🆕 Realized | Last week's "expected risk" became a realized event. On Oct 30 the IRGC Navy said a fourth tanker hit a mine in an "unauthorized" channel in the Strait and caught fire, carrying over two million barrels and ignoring radio warnings. The method escalated from missiles and drones to a manually attached "magnet limpet mine" — a first since the conflict began, according to the Financial Times citing a maritime security company, and the vessel involved was the "Louise I." The same day an LPG carrier was hit in the Persian Gulf about 160 nautical miles west of the Strait, so risk has now clearly spilled outside the chokepoint, and October's regional attack count passed 30 — the highest monthly total since the conflict began in late February. Assessment: the risk centre of gravity moved from the Red Sea back to Hormuz, and from a single chokepoint to two connected sea areas whose risks are no longer independent. Hormuz carries roughly 20 million barrels a day, about a fifth of seaborne oil, with no pipeline substitute available in the short term |
| 🆕 Energy — Brent / WTI EIA fixing |
Brent 118.82 / WTI 94.99 | Brent 128.12 (+7.83%) | 🆕 6-month high | EIA official Wednesday fixing: Brent USD 128.12/bbl, +7.83% for the week — a six-month high — while WTI fell 4.33% to USD 105.25. The consequence is the Brent-WTI spread widening from USD 8.81 to USD 22.87, roughly three times the top of its normal 3-8 band (cross-check: 128.12 − 105.25 = 22.87). This is the most important structural signal of the week: Brent rising while WTI falls means the premium sits in the European physical end — a supply-risk premium, not a global demand recovery. For plastics, housings and inks the transmission is genuine cost inflation; for tanker operators it is freight inflation. Note the caliber: ICE Brent December at 117.64 (+2.4% daily) and WTI December at 104.12 (+2.08% daily) are a different series and must not be set against the spot fixing above |
| 🆕 MLCC lead-time listings Allocation |
Samsung E-M 52 weeks | Listings at 61 weeks | 🆕 +22 wks vs Sept | The passive-component constraint hardened again, and this time the number comes from the channel rather than the factory. iM Securities on Oct 21 reported that some distributors are quoting lead times of 61 weeks, against 39-49 weeks quoted in September. The cross-comparison for IT-equipment-grade MLCC now reads Samsung E-M 52 weeks, Yageo 28 weeks, Murata 17 weeks — a 35-week spread between the tightest and the loosest major. No new price letter this week, and that is the point: with price already executed, quota and lead time are now the harder constraint. November opens 2027 Q1 price negotiations, and Japanese vendors are widely expected to follow Samsung E-M's increase |
| 🆕 Storage contracts 4Q26 |
3Q26 +18-23% | 4Q26 +13-18% | ↗ Converging, still up | TrendForce's Oct 30 update puts 4Q26 PC DRAM contract prices at +13-18%, converging from 3Q's +18-23%. October module contract averages: 8GB DDR4 USD 170 (month +14.86%), 16GB DDR5 USD 275 (+11.34%), 8GB DDR5 USD 165 (+24.06%). The general bands hold — DRAM +10-15%, HBM-inclusive +15-20%, NAND +15-20% — with enterprise SSD's 4Q increase at +23-28% and potentially exceeding 3Q. The key structural fact: vendors are quoting 15-20% but deals are clustering at 10-15%, and Samsung and SK hynix may cut PC DRAM output by more than 40% next year against an original 20% expectation — while notebook shipments fall 9% QoQ in 4Q and PC-maker DRAM inventory stretches from 6-9 weeks to 7-10 weeks. Demand is weakening and prices are still rising, which confirms supply-side dominance rather than demand-side strength. November contract prices are expected above October |
| 🆕 Silver breaks USD 50 Precious metals |
London 54.09 / Shanghai 13,109 | London 49.107 / Shanghai 12,093 | ↓ Support broken | London silver spot fell to USD 49.107/oz, -9.21% for the week, its first break below USD 50 since November 2025 (COMEX at 49.135, -6.20% on the day); Shanghai silver futures fell 7.75% to RMB 12,093/kg from 13,109. Last week's judgement that "USD 50 is the primary support" has been falsified and the framework has to be rebuilt — with real yields not yet turning, bounce height is limited and USD 45-47 is the next reference. The cause is funds and rates, not demand: CFTC managed-money net selling, September CPI at 3.7% above expectations, Fed December-hike odds above 80%, and rising real yields. Silver's industrial weighting, smaller market and thinner liquidity make it fall harder than gold (-9.21% vs gold's -3.61%). Procurement read: silver-bearing parts — conductive silver paste, relay and connector silver contacts, plating layers, silver-bearing solder — have a re-negotiation window now. But the long-term variable is silver consumption, not silver price |
| 🆕 Gold loses two round numbers Precious metals |
Shanghai 870 (night session) | Shanghai 855.18 / London 3,875.19 | ↓ Safe-haven failed | Caliber handled strictly: last week's archive stored Shanghai gold at RMB 870/g on the Oct 30 night session and this week is RMB 855.18/g on the Oct 30 close — night session and close are not the same point in time, so no weekly change is computed and the cell is left blank rather than estimated. The direction comes from the outside market: London spot at USD 3,875.19/oz, -3.61% on the day, losing both the 4,000 and 3,900 levels and hitting the lowest since October 2025. The important insight is that safe-haven behaviour failed: gold fell while a Gulf conflict escalated, because precious metals are currently priced off the rate side, not the geopolitical side. Anyone using "geopolitical tension, therefore buy gold" as a decision rule is working from a broken heuristic |
| Memory DDR5 16Gb (2Gx8) Spot session |
USD 61.20 | USD 61.50 (+0.49%) | ↑ 3rd week above 60 | Oct 30 session USD 61.50 (+0.49% for the week, +0.27% single session) — a third consecutive week above USD 60, with the gain continuing to narrow (+1.27% → +0.49%). On a 30-day basis +10.94%, year to date +459.4%. DDR5 16Gb eTT USD 29.80 (+2.76%). Long-cycle: 55.433 (Sep 18) → 61.20 (Oct 23) → 61.50 (Oct 30), +10.94% over the period |
| Memory DDR4 8Gb (1Gx8 3200) Spot session |
USD 47.107 | USD 46.864 (-0.52%) | ↓ First pullback | Oct 30 session USD 46.864 (-0.52% for the week, 0.00% single session) — the first pullback after the stage high of USD 47.107 set on Oct 23. A negative week is not a downtrend: the full sequence is 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), nine weeks +7.40%, and the current level is still above Oct 16's 45.971. This is a shallow consolidation inside a high-level range, not the start of a slide |
| Memory DDR4 16Gb (2Gx8) Spot session |
USD 82.87 | USD 82.796 (-0.09%) | → Narrow range | Oct 30 USD 82.796 (-0.09%, -0.15% single session) — a third week in a narrow USD 82.8-83.7 band; 30-day -3.44%, year to date +702.3%. DDR4 16Gb eTT at 13.70 was the week's largest memory decliner (-1.62%), and the narrowing gap between white-label and branded dies is itself a signal — the discount that made DDR4 16Gb attractive for substitution is closing |
| Memory module DDR5 RDIMM 32GB Module quote |
USD 2,176.00 (10/05 quote) | USD 2,165.00 (10/19 quote) | ↘ Slight pullback | Caliber correction this week: the 2,176.00 figure was mislabelled last week as the Oct 19 session when it is actually the Oct 5 session. Re-listed on the correct session basis, Oct 5 2,176.00 → Oct 19 2,165.00, -0.51% across a 14-day cross-session gap, which is not a weekly change. The substantive point: module pricing is rising noticeably slower than die pricing, which means module makers are working through previously-priced particle increases and cost pass-through has a lag. Per the pricing-caliber rule, module and die quotes carry separate session dates and are not cross-derived |
| Metals — tin Futures |
RMB 376,030/t (10/23) | RMB 360,880/t | ↓ Week's largest decline | SHFE tin RMB 360,880/t, -4.03% — the week's largest industrial-metal decline — and it was concentrated in two sessions: Oct 28 closed at 387,450, Oct 30 at 360,880, a two-day move of -6.86%. LME tin 48,770 USD/t (-2.46%). Drivers: soft technology-giant earnings, tightening policy expectations under high inflation, and the end of the peak season. But supply has not improved: Chinese tin social inventory is 6,342t (-347 on the week), Yunnan smelter utilization is at 25.56% with maintenance extended, LME tin stocks are only 4,030t, and cancelled warrants have risen to 10.17%. Implication: the fall is demand-expectation driven, not supply loosening — batch your solder buying and keep one third of exposure open |
| Metals — copper, aluminum, zinc Futures |
Cu 107,590 / Al 23,275 / Zn 25,230 / Ni 119,650 | Cu 108,050 / Al 23,395 / Zn 25,850 / Ni 117,190 | ↑ Mixed, mostly up | Last week's pattern fully reversed: copper +0.43% to 108,050 ends two consecutive falls; aluminum +0.52% to 23,395 ends three consecutive falls; zinc +2.46% to 25,850 was the strongest Shanghai metal; nickel -2.06% to 117,190 saw last week's rebound fail to extend (LME nickel stocks at a record high, with Indonesian quota easing and stainless/new-energy demand weakening). Shanghai and London diverged again: LME copper +1.33% to 14,286 against Shanghai's +0.43%, and LME tin -2.46% against Shanghai's -4.03% — two separate series that must not be merged into one percentage, because copper's import exposure and tin's domestic pricing structure differ. Support for the base metals is inventory-driven: Chinese aluminum ingot social stocks fell 23,900t to 527,700t and LME aluminum fell 4,200t to 231,900t, both declining. Implication: the cost-improvement window on aluminum electrolytic capacitors and structural parts is now closed — do not wait for further declines |
| Copper concentrate TC Supply floor |
-237.80 USD/dmt | -241 USD/dmt | ↓ Deteriorated | Import copper concentrate spot TC at USD -241/dmt, from -237.80 last week — a near-decade low, and smelters are now paying the mines to take ore. Driven by more overseas mine disruptions plus concentrated domestic smelter maintenance, and China's refined copper output has fallen year on year for three consecutive months (Jul-Sep) for the first time in nearly five years. Implication: until TC improves, copper's downside is constrained and copper-intensive input costs will not fall with the copper price. Note the inventory counterpoint: LME copper stocks rose 850t to 273,800t (+0.31%), a dramatic slowdown from last week's +20.45% build, with the US warehouse share still high — visible global stocks are rising while domestic spot remains tight, so "stocks up" cannot be read one-way as looseness |
| Freight — container Spot |
SCFI 3,681.34 (10/23 print) | SCFI 3,687.35 (+0.16%) | ↗ Third rise, narrowing | SCFI printed Oct 30 at 3,687.35, +0.16% — a third consecutive weekly rise, with the gain narrowing from +0.42% to +0.16%. CCFI at 1,881.82 (+0.30%, from 1,876.27). The critical point is that these readings pre-date the Oct 30 Hormuz attacks, and freight pass-through typically lags by one to two weeks — next issue's print is the first real test. BDI had no new print and is carried at 2,511. For reference, the prior weeks' lane structure showed Europe and Mediterranean rising on energy and restocking while US lanes fell |
| Insurance — war-risk Highest priorityre-quote |
10% of ship value | No updated quote | ↑ Re-quote needed | The single highest-priority re-quote item this week. The war-risk rate on crude carriers, which jumped from below 3% to 10% of ship value per transit last week, has no updated quote this week and is deliberately not estimated. After four tankers were struck on Oct 30, that rate carries a real risk of a second step-up — obtain fresh quotes from your broker for the Red Sea, the Persian Gulf and the Gulf of Oman lanes now. Keep the three-part caliber intact: rate (per ship value per transit), surcharges, and rerouting fuel cost of USD 2.5-4.5 million per voyage. Note that fuel remains 58-60% above pre-war levels even after this week's crude move, so the amplification on a rerouted voyage stays elevated |
| 🆕 Macro — Fed and rates New |
— | Rates held, Dec hike odds >80% | 🆕 New variable | The Fed held its policy rate unchanged at the Oct 27-28 meeting, and market-implied odds of a December hike now exceed 80%. September CPI printed 3.7% year on year (against 3.6% expected) and core CPI 2.6% (against 2.5%). On Oct 30 the 2-year Treasury yield fell 13bp and the 10-year fell 6bp, with the 10Y-2Y spread widening 7bp. Two transmission channels, and they run in opposite directions from the geopolitical story: rising real yields pressure non-yielding assets — which is why gold and silver fell while a Gulf conflict escalated — while raising US dollar funding costs. This is the mechanism behind the precious-metals selloff and it is why "geopolitics implies safe-haven demand" no longer holds |
| 🆕 MLCC shortage spreads Demand pull |
AI server high-cap | Now reaches optical modules | 🆕 Broadening | TrendForce on Oct 29: NVIDIA's Vera Rubin ramps monthly from 4Q26 and AMD has already procured dedicated high-cap MLCC for MI450, with ODMs expecting 2027 Q1-Q2 to be the tightest high-cap MLCC window. Critically, the shortage has now spread to communication-grade MLCC used in optical modules — which means the squeeze is no longer confined to the AI server BOM and now reaches a second, less-discussed chain. Consumer and mid-cap spot prices fell about 5% in early October before stabilizing, and the high-end increase has narrowed, but AI-grade parts remain tight |
| 🆕 Connector capacity Volume ramp |
224G in volume supply | Q3 results confirm ramp | ↗ Volume progressing | Two domestic leaders reported Q3 on Oct 28. One posted Q3 revenue RMB 6.882bn (+47.87%) and net profit RMB 816m (+172.13%), with liquid-cooling orders for leading intelligent-computing platforms growing so fast that capacity has become insufficient and orders have built a backlog, Q4 capacity continuing to expand, and a subsidiary's optical-communications orders booked into 2027; its 112G connector components are scaling and 224G customer validation is progressing. The other posted Q3 revenue +43% with R&D up nearly 70%, 112G in stable mass production, 224G completing engineering validation and entering small-batch supply, NPO opto-electronic interconnect validated for small-scale Q4 output, and 448G in pre-research; capacity of RMB 612m across two phases and 30 lines, with the first lines starting up in November; the 224G parts are qualified for GB200 / B300-class platforms. Read together: domestic 224G is moving from validation into small-batch and then volume — but client certification still runs 12-18 months |
| USD / RMB FX |
6.7180 (10/23 parity) | 6.7138 (10/30 parity) | ↗ RMB firmer | Oct 30 central parity 6.7138, up a further +42bp versus the Oct 23 endpoint of 6.7180 — the strongest since February 2023; the path was steady (10/26 6.7162 → 10/30 6.7138). But size the benefit correctly: a 0.06% weekly move against storage and MLCC increases of double digits does not hedge anything. It is a small buffer, not an offset, and it should not slow your locking pace |
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; energy pressure stepped up while the MLCC constraint moved further from price to allocation.
| Segment | 1H 2026 | Current (Oct 30) | 2027 Outlook | |
|---|---|---|---|---|
| Memory DRAM / DDR5 | High | Red alert | Red alert | 3rd week above $60; 4Q26 PC DRAM +13-18%; PC cut plans >40% |
| NAND / eMMC | High | Elevated | High | 4Q26 NAND +15-20%; enterprise SSD 4Q may exceed 3Q |
| MLCC Ceramic Capacitors | Red alert | Red alert | Red alert | Listings 61 weeks; shortage now reaches optical-module MLCC |
| Tantalum Capacitors | High | Red alert | Red alert | Concentrate pass-through still incomplete; 40-52 wk lead times |
| Aluminum-Electrolytic / Film | Yellow alert | Red alert | High | Aluminum +0.52% ends 3 down weeks; cost-improvement window closed |
| MCU Microcontrollers | High | Red alert | Red alert | Industrial MCU lead times 72-80 wks, booked into 2028 |
| Analog / Power Management | High | Red alert | Red alert | onsemi executing; Taiwan +10-15% unannounced 3rd week; Renesas 76 days out |
| Power Devices | High | Red alert | Red alert | AI and automotive take capacity; legacy MOSFET capacity net exit |
| PCB / CCL | Yellow alert | Red alert | Red alert | Cu/Al/Zn up, TC -$241/dmt; 2027 capacity orders land 2028 at earliest |
| Copper / Base Metals | High | High | High | Tin -4.03% but supply tight; LME copper stocks +850t only |
| Freight / War-risk | Yellow alert | High | High | 10% ship-value war-risk, no fresh quote; 4 tankers struck Oct 30 |
| Connectors / RF | Yellow alert | Elevated | Elevated | 224G into small-batch then volume; lines start Nov; cert 12-18 months |
| Precious Metals | Watch | Watch | Watch | Silver <$50, gold <3,900 — rate-driven, safe haven failed |
| Origin / Compliance | Watch | High | Elevated | Taiwan export-control list; RMB 6.7138 — too small a hedge |
DDR5 16Gb held above USD 60 for a third week at USD 61.50 (+0.49%), with gains narrowing (+1.27% → +0.49%). DDR4 8Gb pulled back -0.52% from its stage high — a shallow consolidation inside a nine-week +7.40% uptrend, not a reversal. The more consequential development is contractual: 4Q26 PC DRAM at +13-18%, with vendors quoting 15-20% but deals clustering at 10-15%, and Samsung and SK hynix now planning more than 40% PC DRAM output cuts next year — against notebook shipments falling 9% QoQ. Demand weakening while prices rise is the clearest possible proof that this is supply-side dominance.
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. The matrix below tracks where AI orders are being absorbed and what is being displaced as a result — a standing weekly tracker, because this is the mechanism behind the whole price complex rather than a one-week event.
Last week the metals tape fell broadly while fabric rose. This week the tape split: copper +0.43%, aluminum +0.52% and zinc +2.46% all rose while tin fell 4.03%. More importantly, copper concentrate TC deteriorated further to USD -241/dmt and Chinese refined copper output has now fallen year on year for three straight months. The implication is uncomfortable for buyers hoping for cost relief: the input-cost floor is rising even where the headline metal price is not.
| Node | This week (Oct 30 close) |
|---|---|
| Copper & copper foil | SHFE copper RMB 108,050/t (+0.43%), ending two consecutive weekly falls; LME copper 14,286 USD/t (+1.33%). The pattern inverted again — external strength, internal weakness, after last week moved the other way. Drivers: 2-year and 10-year Treasury yields falling 13bp and 6bp eased valuation pressure, concentrate TC deteriorating to USD -241/dmt reinforced the supply constraint, and domestic refined copper social stocks drew down against a rising LME total. Procurement: last week's "discount the input cost by the weekly copper move" logic does not apply this week — the processing-fee concession window on copper foil, enamelled wire and lead frames has narrowed, and a small copper price rebound is not a reason to accept a processing-fee increase, because the TC deterioration squeezes smelter margins, not converter margins |
| Aluminum | SHFE aluminum RMB 23,395/t (+0.52%), ending three consecutive weekly declines (W41 -1.34% → W42 -0.70% → W43 -0.52% → W44 +0.52% on a different base). Support is inventory-led: Chinese aluminum ingot social stocks fell 23,900t to 527,700t and LME aluminum fell 4,200t to 231,900t — both declining, with new Xinjiang capacity coming on gently. Procurement: the cost-improvement window on aluminum electrolytic capacitors and structural parts is now closed — last week's "the decline is narrowing to flat" has been delivered as a turn up, so do not wait for further relief |
| Tin / solder & plating | SHFE tin RMB 360,880/t (-4.03%) — the week's largest industrial-metal decline, and concentrated in two sessions: 387,450 on Oct 28 to 360,880 on Oct 30, a two-day -6.86%. LME tin 48,770 (-2.46%). Drivers: soft technology-giant earnings, tightening policy expectations under high inflation, and the end of peak season. But supply has not improved: Chinese tin social inventory 6,342t (-347 on the week), Yunnan smelter utilization 25.56% with maintenance extended, LME stocks only 4,030t, cancelled warrants up to 10.17%. Base case is a weak range of RMB 355,000-380,000/t. Procurement: do not chase the fall — batch solder buying into three tranches covering four weeks of use and keep one third of exposure open, because a tightening of Myanmar or Indonesian export policy on this inventory level would produce a sharp move. A negative week is not a downtrend, and this one is demand-expectation driven rather than supply loosening |
| Zinc / nickel | Zinc +2.46% to RMB 25,850/t was the strongest Shanghai metal, on continued deterioration in ore supply, falling zinc concentrate treatment charges, smelter cuts at home and abroad, and weaker sulfuric acid prices squeezing smelter margins. Nickel -2.06% to RMB 117,190/t as last week's rebound failed to extend, with LME stocks at a record high on eased Indonesian quotas, a further 2.91% decline in global mine supply and softer stainless and new-energy demand. Zinc pressure returns to galvanized steel and die-cast alloys; nickel relief eases MLCC electrode-powder cost — but nickel powder is priced on conversion cost, so pass-through still takes one to two quarters and that is not a reason to demand an immediate powder price cut |
| E-yarn G75 / e-fabric 7628 | No new hike letter this week; 7628 holds at RMB 13.5-14/m and G75 at RMB 25,800-26,000/t, now in execution verification. Chinese e-yarn nameplate capacity is 1.368Mt/yr of which only about 20kt is low-Dk furnace process, and 2026 is a concentrated cold-repair year on 2013-2015 ignitions, with repairs running three to six months. AI-server glass-fabric consumption runs 18-24 metres per unit against roughly five times that for a conventional server. Tightness persists into 2027, and with copper and aluminum now rising the PCB input basket has stopped providing relief entirely |
| AI high-speed CCL (M8/M9) | PPE resin — ~70% of global supply is Saudi-produced with only about 65% of capacity restored, and a maintenance outage is scheduled inside November, the largest single materials risk for Q4. This week adds a second problem: the Strait attacks raise the logistics and insurance cost of Saudi resin exports, so the risk is now "capacity short" plus "cannot ship it out" at the same time — plan six to eight weeks of cover. Electronic-grade PPE is up +261% year to date |
| Semiconductor materials | Japan photoresist +15% is effective (Oct 1): high-end ArF +16-22%, HBM immersion ArF up to +24%; wafers rising across all sizes. TSMC's Q4 revenue guidance of +61% and 2027 capex of USD 60-64bn confirm a second pass-through into foundry cost is coming, not that it has finished |
Because the metal that matters most is priced off concentrate, not off the exchange. TC at USD -241/dmt means smelters are now paying mines to take ore — a negative TC is structurally at odds with a falling copper price, which is exactly why refined copper output has declined year on year for three straight months, the first such run in nearly five years. Meanwhile LME copper stocks rose only 850t to 273,800t (+0.31%), a dramatic slowdown from last week's +20.45% surge, and the US warehouse share is still high: visible global stocks are rising while domestic spot stays tight. Reading "stocks up" one-way as "looser" is wrong this week.
Tin illustrates the mirror image. It fell 4.03%, and it looks like relief — but Chinese social inventory is down 347t, Yunnan smelters are running at 25.56% with maintenance extended, LME stocks are only 4,030t and cancelled warrants have risen to 10.17%. The price fell on demand expectations and macro sentiment, not on supply. That is a different mechanism from copper's supply squeeze, and it carries a different reversal risk.
Aluminum and zinc make the same point from the other side: both are supported by falling inventories, so the window for cost relief on electrolytic capacitors, plated parts and structural components has closed rather than opened. Taken together: the PCB input basket is no longer a source of negotiation leverage — the only remaining concession worth chasing is a Q4 adjustment clause on tin-based solder, and it should be sized as a one-off, not a trend.
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.
| Application | Demand | Visibility | YoY | What changed this week |
|---|---|---|---|---|
| AI servers / compute | Extreme | Into 2027 | +31% TrendForce 2026 global AI server shipments (>2m units) |
Two domestic connector makers reported Q3 on Oct 28 — one with net profit +172.13% on revenue +47.87% and liquid-cooling orders for leading intelligent-computing platforms so over-subscribed that capacity has become insufficient and orders have built a backlog, expanding into Q4, with a subsidiary's optical-communications orders booked into 2027; the other with revenue +43%, R&D up nearly 70%, 224G moving from engineering validation into small-batch supply, NPO opto-electronic interconnect validated for small-scale Q4 output, and the first of 30 new lines starting in November. TrendForce (Oct 29): Vera Rubin ramps monthly from 4Q26, AMD has procured dedicated high-cap MLCC for MI450, and ODMs expect 2027 Q1-Q2 to be the tightest high-cap MLCC window |
| Consumer electronics (phone / PC) | Contracting | 6-12 months | -22% IDC 3Q26 global phones, released Oct 14; largest single-quarter decline on record |
New negative readings this week. A TrendForce survey has 4Q26 major notebook ODM order volumes forecast to fall 10-15% year on year on average, with handset brands cutting low-to-mid-tier models; notebook shipments are down 14% and 9% QoQ in 3Q and 4Q. Gartner: memory price rises lift memory from 16% to 23% of PC material cost, with PC average selling prices up 17%. One major PC vendor posted 3Q personal-systems revenue of USD 11.8bn (+18%) but units -16% and operating margin down from 5.4% to 4.6%. Read: the PC supply chain is being repriced by memory, not by end demand |
| Automotive electronics | Firm (structural) | 3-6 months | +13.6% September new-energy vehicle sales |
Cost side turned against buyers: copper +0.43% and aluminum +0.52% push vehicle copper-foil and aluminum cost back up, and aluminum electrolytic capacitor and structural-part cost no longer improves. Automotive MLCC quota pressure is unrelieved — all grades at 52 weeks — so on automotive parts the lead-time risk now clearly exceeds the price risk |
| Industrial control | Firm | 2-4 months | +18.9% September industrial robot output |
No new readings this week. European energy costs remain high — German and French industrial power above EUR 120/MWh — and now compounded by Brent at +7.83%, raising cost pressure on European industrial customers. Industrial MCU lead times of 72-80 weeks, booked into 2028, are unchanged |
| Photovoltaics | Weak | 1-3 months | -36% to -49% CPIA 2026 full-year forecast, 160-200GW — forecast caliber |
New this week: London silver below USD 50 (49.107, -9.21% for the week), so conductive silver paste cost eases materially and silver-bearing parts have a re-negotiation window. The rule must be restated: 2026 photovoltaic silver consumption is down 20% on silver-coated copper and copper-plating substitution — consumption, not the silver price, is the long-term variable. Polysilicon at 35,330 (-2.62% on the day); green alert retained |
| Energy storage | Recovering | 1-3 months | +42% CNESA September newly installed power — tracked separately from solar |
No new concentrated procurement data this week. Cost side continues to improve: lithium carbonate futures 90,540 (-0.66% on the day), battery-grade spot average 93,608 (-2,535 day on day), weekly average 98,000 (-2.49% week on week), so cell cost keeps falling — but copper and aluminum turning up is a reverse pressure on busbars and structural parts |
| 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. Domestic carriers' 4Q26 base-station tendering started Oct 1. New watch item: communication-grade MLCC used in optical modules has started to show shortages (TrendForce Oct 29) — so passive-component lead times in the base-station and optical-module chain need to be monitored separately from the AI-server chain |
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 — war-risk re-quote4 tankers struck Oct 30; rate at 10% of ship value with no fresh quote | This is the highest-priority action item in the report, and it is a phone call, not a negotiation. The crude-carrier war-risk rate jumped from below 3% to 10% of ship value per transit last week and has no updated quote this week — we deliberately do not estimate it. After four tankers were struck on Oct 30, a second step-up is plausible. Get fresh quotes from your broker this week for the Red Sea, the Persian Gulf and the Gulf of Oman lanes separately, and separate the crude-carrier rate from the container rate in your model — they are different instruments with different triggers. Keep the three-part caliber intact: rate per ship value per transit, surcharges, and rerouting fuel cost of USD 2.5-4.5 million per voyage with fuel still 58-60% above pre-war. | Re-price the route, not the commodity: a 22.87 Brent-WTI spread means the premium is in the European physical end, so contracting origin flexibility is worth more than hedging the crude benchmark. Hold air-freight contingency; keep a Suez-return option for Europe lanes where the carrier still offers it; negotiate a war-risk rate cap with overrun sharing rather than accepting an open-ended rate. |
| Taiwan power / MCU+10-15% planned, still unannounced — third week, last comfortable window | Third consecutive week we flag this, and it is now the last comfortable comparison window. Panjit, TSC, Epistar and NIC plan +10-15% on spot products and, as of Oct 30, still have not issued a notice — and Taiwanese media language has already shifted from "rising from October" to "Q4 adjustment of non-contract prices", which is a softer commitment. Complete the new-order negotiation with Taiwan makers this week and secure written magnitude and effective date, so their increase does not stack passively on onsemi's already-live +5-15%. If it slips past the November midterms, you will be negotiating from a weaker position with a 30-50+ week lead time behind you. | Cross-brand pin-to-pin on MOSFET / IGBT / SiC; package arbitrage; qualify second sources on long-lead SiC now. Renesas's Jan 1, 2027 repricing belongs in 2027 budgeting, not in this cycle's scramble. Note that legacy power capacity is net exiting because AI and automotive take the schedule — so the second-source decision is more urgent than the price is. |
| MLCC — 61-week listingsDistributor lead times to 61 weeks; shortage reaches optical modules | The constraint has moved from the factory queue to the channel queue, which changes who you negotiate with. With factory lead times at 17-52 weeks depending on the vendor and distributor listings at 61 weeks, the effective wait is longer than any factory number suggests. Get written allocation and lead-time commitments from your distributor, not just a price, and place allocation requests now rather than at the point of need. November opens 2027 Q1 price negotiations and Japanese vendors are widely expected to follow Samsung E-M's increase — so if you need 2027 volume, the negotiation window to influence it is closing in weeks, not months. Watch that the shortage has now reached communication-grade MLCC for optical modules, which means your second chain is starting to move. | Generic 0402/0603 and consumer grades decouple from AI high-cap — consumer and mid-cap spot prices fell about 5% in early October before stabilizing, so there is genuine slack below the AI tier. Rationalize part numbers to improve allocation leverage; qualify a mainland alternative in the industrial tier, priced at current levels rather than reused assumptions. |
| Memory contracts4Q26 PC DRAM +13-18%; PC DRAM cut plans >40% | Stop reading spot for the cost signal and start reading contracts. Spot gains are narrowing (+1.27% then +0.49%) while 4Q26 contracts run +13-18%, vendors quote 15-20%, deals cluster at 10-15%, and November contracts are expected above October. Conclude the 4Q26 and 1Q27 volume and price-ceiling negotiations inside the guidance band now, before the signed numbers land. The structural reason to act is supply-side, not demand-side: Samsung and SK hynix may cut PC DRAM output by more than 40% next year against an original 20% expectation, while notebook shipments fall 9% QoQ — supply is being cut into weakening demand, which is exactly why prices still rise. Do not wait for a soft spot tape to buy; it is not coming. | Continue DDR4→DDR5 migration where the design allows, but size the saving correctly — the white-label-to-branded gap is narrowing, and DDR4 16Gb eTT fell 1.62% this week, so the discount that made that tier attractive is closing. Where memory is a large share of the bill of materials, the pressure now shows up as a BOM-structure problem, not a price problem: one major PC vendor has seen memory rise from 16% to 23% of material cost. |
| PCB / CCL & metalsCu/Al/Zn up, Sn -4.03%, TC -$241/dmt | Your leverage on this chain has shrunk, and pretending otherwise will cost money. Copper, aluminum and zinc all rose, inventories are falling, and the cost floor moved up for a reason that is not visible in the headline price: concentrate TC at USD -241/dmt with three straight months of falling Chinese refined output. Do not discount copper-input costs by the copper price, and do not accept a processing-fee increase on the back of a small copper rebound — the TC deterioration squeezes smelter margins, not converter margins, so converter processing fees are the last place to concede. The one concession still worth chasing is a Q4 adjustment clause on tin-based solder, and it should be sized as a one-off, not a trend: tin fell on demand expectations and macro sentiment while inventories kept drawing down. Batch solder buying into three tranches covering four weeks and keep one third of exposure open. | Rational material downgrade; layer reduction; panel optimization; qualify a 2nd CCL supplier; for non-AI products test whether M6 can substitute M8. Keep glass fabric, resin, foil and base metal as four separate line items — this week they moved in four different directions, and a single blended assumption would have been wrong on three of them. |
| Silver-bearing partsLondon silver -9.21%, below $50 for the first time since Nov 2025 | This is the one place a genuine cost concession appeared this week — and it will not stay open. London silver spot at USD 49.107/oz (-9.21%) has broken the USD 50 support that was treated as primary, and Shanghai silver is -7.75% at RMB 12,093/kg. Re-open negotiations on silver-bearing parts now — conductive silver paste, relay and connector silver contacts, plating layers and silver-bearing solder — and ask for re-quotes on the silver component rather than a blanket discount, so the concession is sized correctly and does not leak into non-silver lines. But be precise about why it happened: this is a rates-and-positioning selloff, not a demand improvement, and real yields have not turned, so treat it as a window rather than a new level. | Technical substitution is the real lever here: 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. Note also that gold has now lost two round numbers with safe-haven behaviour failing outright — so any precious-metal-based cost assumption in your contracts should be re-examined, not just re-priced. |
| Consumer and PC exposureNotebook ODM orders -10-15% YoY; memory 23% of PC BOM | Do not let the consumer demand data sit on the dashboard — it changes what you negotiate and what you build. With 4Q26 notebook ODM order volumes forecast down 10-15% year on year and PC-maker DRAM inventory stretching to 7-10 weeks, consumer-tier memory is the one place where destocking is plausible — but the general bands and the December IEA decision point argue against assuming the memory complex rolls over. Work the gap between the tiers: consumer and mid-cap MLCC spot already fell about 5% in early October before stabilizing, while AI-grade stayed tight. For PC-adjacent products, the pressure now arrives as a BOM-structure change (memory from 16% to 23% of material cost, PC average selling prices up 17%) rather than as a component price change, which means engineering and product decisions are affected before purchasing is. | Right-size the consumer BOM ahead of the seasonal peak rather than into it. Where a consumer product does not need AI-grade, spec down — the slack is below the AI tier, not in it. Do not extend the AI-grade part number into a consumer design just because it is already qualified. |
Attention allocation: act immediately in the top-right, monitor continuously in the bottom-left.