DDR5 Holds Above USD 61 and DDR4 8Gb Turns Up — but Brent's 8.56% Fall Does Not Mean the Strait Risk Is Resolved
1Week-over-Week Comparison — Segment Changes & Drivers
Comparison window: W42 (2026.10.09–10.16) vs W43 (2026.10.16–10.23). Overall risk levels are unchanged at 7 red / 4 yellow / 2 green — but two risk forms changed, and one new compliance variable arrived. The week in one sentence: memory moved from generational divergence into a broad rally, the oil premium came down, and the freight-and-insurance premium went up.
| Category / Metric | Last week (W42) | This week (W43) | Direction | What changed |
|---|---|---|---|---|
| Memory DDR5 16Gb (2Gx8) Spot session |
USD 60.433 | USD 61.20 (+1.27%) | ↑ Second week above 60 | Oct 23 close USD 61.20 (+1.27% for the week, +0.55% single session) — a second consecutive week above the USD 60 round number (W42 broke it at 60.433). On a 30-day basis +10.40%, year to date +456.5%. The long-cycle read: 55.433 on Sep 18 → 58.900 (Oct 9) → 60.433 (Oct 16) → 61.20 (Oct 23), a clean stair-step rather than a spike |
| Memory DDR4 8Gb (1Gx8 3200) Spot session |
USD 45.971 | USD 47.107 (+2.47%) | 🆕 Turns up | The most important memory change of the week: DDR4 8Gb rose +2.47% to USD 47.107, ending two consecutive weekly declines (30-day +2.09%, year to date +795.6%). Alongside DDR3 4Gb turning up +1.18% to 13.842, this means memory has moved from generational divergence into a broad rally — the generational spread actually narrowed from USD 14.46 to USD 14.09. The prior "substitution-driven pullback" narrative no longer holds: if the old tier is rising too, the driver is capacity allocation squeezing the 2Gx8 and 1Gx8 tiers as bits are pulled toward server DDR5 and HBM, not weak end-demand |
| Memory DDR5 16Gb eTT (white-label) Spot session |
USD 27.800 | USD 29.00 (+4.32%) | ↑ Strongest tier | USD 29.00 (+4.32%, +1.09% single session) — the strongest tier in the whole memory complex this week, and a third consecutive week of outpacing branded particles. Read together with the DDR4 8Gb turn-up, the signal is that scarcity is being priced wherever qualification friction is lowest, not where brand premium is highest — a different buying problem from the one we faced two weeks ago |
| Memory DDR4 16Gb (2Gx8) Spot session |
USD 83.660 | USD 82.87 (-0.94%) | ↓ Sole decliner | Oct 23 close USD 82.87 (-0.94%) — the only decliner this week, and it breaks the three weeks of marginal flatness that preceded it (on a 30-day basis -3.36%, year to date +703.0%). DDR4 16Gb eTT 13.925 (-1.33%); DDR4 8Gb eTT 5.970 (flat). This tier remains ~35% above DDR5 16Gb, so it is still the structurally expensive side of the market; the 30-day pullback is a normalization inside an intact uptrend, not a reversal |
| Memory module DDR5 RDIMM 32GB Module quote |
USD 2,176.00 (10/05 quote) | USD 2,176.00 (carried) | → No new session | No new module session this week; the Oct 5 quote of USD 2,176.00 is carried forward unchanged (+0.58% versus the Sep 28 quote of 2,163.50). Per the pricing-caliber rule, module and die quotes carry separate session dates and must never be cross-derived — no module-side weekly change is stated, and the die-side rally is not presented as confirmation of it |
| NAND / eMMC Cross-session |
TLC 18.833 · SLC 4.353 | TLC 18.592 (10/19 session) | ↓ Cross-session | 512Gb TLC at USD 18.592 on the Oct 19 session, -1.28% against the Oct 5 reading of 18.833 — a 14-day cross-session change, not a weekly move, and labeled as such. SLC 2Gb stays at the Oct 5 reading of USD 4.353. The honest read is that NAND is still correcting while DRAM re-firmed, so a NAND-based forecast should be marked down while a DRAM-based one should not |
| 🆕 MLCC lead times Form switch |
Samsung E-M 30 weeks | Samsung E-M 52 weeks | 🆕 +73%, +22 weeks | The single most violent change on the passive side this week, and it is not a price. Korean press on Oct 20, corroborated by iM Securities, has Samsung Electro-Mechanics MLCC lead times stretching to 52 weeks in October, from 30 weeks in September — +73%, or +22 weeks. 52 weeks is effectively "over a year," at which point ordinary price negotiation stops working and the only mechanism left is allocation plus long-term booking. The whole industry is moving together: Yageo 30-52 weeks, Viking 52 weeks+, NIC 30-40 weeks. Channel price for 1206-476 high-cap stayed at RMB 850-890 per 2,000 pieces with no new price letter this week |
| 🆕 Taiwan export controls Compliance |
No list change | AI servers listed | 🆕 New | Taiwan has added AI servers and advanced semiconductors to its export-control list, effective this month, with transfers into mainland China now requiring a licence and violations carrying criminal liability. Implication: Taiwan-origin AI-related parts need a compliance-traceability and licence-path review, and transshipment or third-country shipping arrangements need to be re-examined. This is a new compliance variable rather than a price event, but it is the kind that stops shipments outright |
| 🆕 Energy — Brent Spot |
USD 129.95/bbl | USD 118.82/bbl (-8.56%) | ↓ Premium given back | Weekly EIA official Wednesday fixing USD 118.82/bbl, -8.56%. WTI rose +15.81% to USD 110.01 from an abnormally discounted 94.99, so the Brent-WTI spread collapsed from USD 31.22 to USD 8.81 and is back inside the normal 3-8 band — this is spread convergence as much as a crude decline, and the geopolitical premium has been given back rather than extended. Two causes: the New York Times reported on Oct 21 that the US and Israel plan to defer strikes on Iran until after the midterms, plus expectations of the Oct 15 IEA reserve release. But read the freight rows below before drawing any relief conclusion |
| 🆕 Geopolitics — Red Sea / Saudi Escalation |
Hormuz strike plan reported | Houthi strikes on Saudi Arabia | 🆕 Risk rotated | On Oct 22 Houthi forces claimed four attacks on Saudi targets within 48 hours — the Riyadh defense ministry, two airports and oil facilities — the Bab al-Mandeb is under pressure, and the US State Department ordered non-essential staff and dependants out of Saudi Arabia. Assessment: the risk centre of gravity rotated from the Strait of Hormuz back to the Red Sea and the Gulf. The consequence lands on insurance and freight rather than on the crude benchmark: super-tanker war-risk insurance jumped from below 3% to 10% of ship value, and Gulf-to-Asia super-tanker freight printed a record USD 37.37/bbl. Oil premium down, logistics-and-insurance cost up — these are two separate bills and only one of them came down. Saudi Arabia continues negotiating crude loading routes outside Hormuz |
| 🆕 Super-tanker war-risk & freight Insurance |
Below 3% of ship value | 10% of ship value | 🆕 Step change | The number to act on this week: crude-carrier war-risk insurance moved from below 3% to 10% of ship value per transit, and Gulf-to-Asia super-tanker freight set a record at USD 37.37/bbl. This sits on top of the standing container-market three-part cost — war-risk rate 0.35%-0.5% of ship value per transit on Red Sea / Gulf-of- Aden routing (roughly 0.1% at the start of the year), plus surcharges, plus rerouting cost of USD 2.5-4.5 million per voyage — with fuel still 58-60% above pre-war levels even after Brent's 8.56% fall, so the amplification on a rerouted voyage stays elevated |
| Freight — container lanes Spot |
SCFI 3,664.11 | SCFI 3,679.58 (+0.42%) | ↘ East-West divergence | SCFI printed Oct 17 at 3,679.58 (+0.42%), a second consecutive weekly rise. The composition matters more than the index: Europe 2,335/TEU (+2.7%) and Mediterranean 3,250 (+8.3%) rose on energy and restocking, while US West 7,525/FEU (-0.8%), US East 10,405 (-1.1%), South America 5,529 (-6.7%) and Australia/New Zealand 2,553 (-7.6%) all fell. East-West divergence: Europe-line export costs must be re-estimated upward, and US-line costs are easing. CCFI had no new print and stays at the Oct 10 reading of 1,905.40 |
| Dry bulk — BDI Spot |
2,545 | 2,511 (-1.34%) | ↓ Sharp narrowing | BDI printed Oct 23 at 2,511, -1.34% on same-caliber endpoints (Oct 16 → Oct 23) — a second weekly fall, but a dramatically narrower one than last week's -15.00%. Driver: shipping-cadence normalization after the holiday plus softer iron-ore dispatch. Reverse anchor: BDI snaps back quickly once iron-ore shipping rhythms recover. Dry bulk is not AI-driven and does not mirror container pricing |
| Metals — base Futures |
Cu 108,970 / Al 23,345 / Zn 25,610 / Ni 115,800 / Sn 378,380 | Ni 119,650 / Cu 107,590 / Al 23,275 / Zn 25,230 / Sn 376,030 | ↗ Diverging, not falling | The metals tape stopped being a broad decline and became a divergence. Nickel rose +3.32% to RMB 119,650/t — the only Shanghai base metal to gain, ending two consecutive weekly falls, which puts MLCC internal-electrode nickel-powder cost pressure back on the rise (though nickel powder is priced on conversion cost, so pass-through still takes one to two quarters). Copper fell -1.27% to 107,590 with LME copper -1.74% to 14,098 — both exchanges now weakening together, reversing last week's internal-weak/external-strong split. Aluminum -0.30% to 23,275 (a third consecutive weekly decline, with the size narrowing -1.34% → -0.70% → -0.30%). Zinc -1.48% to 25,230. Tin -0.62% to 376,030 (sharply narrower than last week's -3.50%), still below 380,000. Lead flat at 16,220 (no new print); LME nickel 15,995 USD/t. New: LME copper inventories turned to a build of +46,350t to 272,950t, with US warehouses holding 169,475t, or 62% — tight ore against loose stocks, which caps downside but breaks immediately if the tariff lands |
| Copper concentrate TC No new print |
-237.80 USD/dmt | -237.80 USD/dmt (unchanged) | → No new reading | No new concentrate TC reading this week; the -237.80 USD/dmt historic low is maintained (last week's print was also -237.80 — the number did not change, the commentary did). Combined with the LME inventory build, the picture is tight ore, loose stocks: the copper price decline is driven by tariff-expectation disappointment and a broadly weaker dollar-priced commodity complex, not by supply relief, so downside room is limited. And HVLP4 copper-foil processing fees held at RMB 35,000/t with high-end HVLP foil still short by 15-25% — proof again that high-end foil is priced by capacity, not by copper |
| E-fabric / e-yarn (PCB) Execution check |
7628 at RMB 13.0-13.3/m | 7628 at RMB 13.5-14/m | ↑ Up again | 7628 e-fabric market readings moved up to RMB 13.5-14/m — the third week of execution verification, and up again after two weeks of flatness, with G75 e-yarn holding at RMB 25,800-26,000/t. Kingboard Laminates issued its eighth round of increases for the year on Oct 13 (+15% on 2116 and thinner cloth plus PP), now in execution. Corroboration on a monthly caliber, with the specific release date unconfirmed: a Chinese brokerage's October tracking has 7628 +RMB 1.5/m, G75 +RMB 3,000/t and 2116/1080 thin cloth +RMB 2.5/m. Implication: PCB input cost pressure did not ease with the metals, and a second CCL round in Q4 remains live |
| Foundry — TSMC (carried) No new readout |
Q3 net T$763.1bn (+69%) | Unchanged this week | → No change | No new foundry readout this week — the Oct 15 call remains the standing reference: Q3 net profit T$763.1bn (+69% YoY), Q4 revenue guidance USD 52.3-54.3bn (+61%), 2026 full-year growth about 44%, 2026 capex USD 60-64bn, 2027 CoWoS capacity to grow more than 45%, utilization about 96.5% through mid-October with everything from 16/12nm down to 2nm at 100%. Carried forward unchanged, not a new event |
| Power devices Window narrowing |
Execution week 1 | Execution, window narrowing | ↗ Last window | onsemi's +5-15% remains in execution (effective Oct 10) — the task is still verifying that invoices match the lettered magnitude, pursuing current-price treatment on the shipping tail, and opening 4Q-1Q27 framework talks. Taiwan makers (Panjit / TSC) at +10-15% for October are still unannounced as of Oct 23 — the second consecutive week we have flagged this, and the window is visibly narrowing. Renesas's second repricing lands Jan 1, 2027, 76 days out |
| USD / RMB FX |
6.7253 (10/16 parity) | 6.7180 (10/23 parity) | ↗ RMB firmer | Oct 23 central parity 6.7180, a further +73bp versus the Oct 16 endpoint of 6.7253 — the strongest since February 2023; offshore 6.6727. Local-currency cost of USD-priced memory / MLCC / analog keeps improving. But do not read this as a reason to relax locking pace — agency quotes are USD-denominated and the currency gain is largely offset by OEM price rises running in the same direction |
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 — but two risk forms changed: MLCC from price to lead time, and energy from crude premium to insurance and freight.
| Segment | 1H 2026 | Current (Oct 23) | 2027 Outlook | |
|---|---|---|---|---|
| Memory DRAM / DDR5 | High | Red alert | Red alert | Broad rally: DDR4 8Gb +2.47% turned up; spread narrows to $14.09 |
| NAND / eMMC | High | Elevated | High | Oct 19 session $18.592, -1.28% cross-session — still correcting |
| MLCC Ceramic Capacitors | Red alert | Red alert | Red alert | FORM: price → lead time. Samsung E-M 30 → 52 weeks |
| 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 -0.30%, third weekly fall but narrowing; anode +RMB 495/t floor |
| MCU Microcontrollers | High | Red alert | Red alert | Lead times booked into 2028; Taiwan window narrowing |
| Analog / Power Management | High | Red alert | Red alert | onsemi execution; nickel +3.32% lifts electrode cost; Renesas 76 days out |
| Power Devices | High | Red alert | Red alert | onsemi +5-15% live; Taiwan +10-15% unannounced, window narrowing |
| PCB / CCL | Yellow alert | Red alert | Red alert | 7628 up again to 13.5-14; Kingboard 8th round live; foil fees hold |
| Copper / Base Metals | High | High | High | Now diverging: nickel +3.32%, copper -1.27%; LME stocks +46,350t |
| Freight / War-risk | Yellow alert | High | Elevated | Tanker war-risk <3% → 10% of ship value; SCFI lane divergence |
| Connectors / RF | Yellow alert | Elevated | Elevated | 224G volume delivery confirmed; domestic capacity beginning to scale |
| IMU / Sensors | Watch | Yellow alert | Elevated | No change at any tier — steady state, structural value only |
| Origin / Compliance | Watch | High | Elevated | Taiwan adds AI servers to export-control list; RMB 6.7180 |
4Memory — The Narrative Corrects: This Is a Broad Rally, Not Divergence
Last week we said memory was diverging by generation. This week that changed: DDR4 8Gb rose +2.47% and ended two consecutive declines, DDR3 4Gb turned up +1.18%, and the white-label eTT tier rose +4.32% for a third straight week of outpacing branded dies. The generational spread narrowed from USD 14.46 to USD 14.09 — scarcity is now priced across generations, not just in DDR5.
Baseline index
YTD +185%
YTD +203%
YTD +456.5%
YTD +795.6%
5PCB & CCL — The Chain Went Up While the Metals Tape Went Down
This is the clearest evidence yet that PCB input cost and metal price are two different things. Copper fell 1.27%, aluminum 0.30%, zinc 1.48% and tin 0.62% — and in the same week 7628 e-fabric rose again to RMB 13.5-14/m and Kingboard's eighth increase of the year went live. Fabric, resin and foil are capacity-constrained; the metals tape is a financial and seasonal story.
| Node | This week (Oct 23 close) |
|---|---|
| E-yarn G75 / e-fabric 7628 | 7628 readings moved up to RMB 13.5-14/m from RMB 13.0-13.3/m — the third week of execution verification, and up again after two flat weeks. G75 e-yarn holds at RMB 25,800-26,000/t. Kingboard Laminates issued its eighth increase of the year on Oct 13 (+15% on 2116 and thinner cloth plus PP), now in execution. Corroboration on a monthly caliber, specific release date unconfirmed: a Chinese brokerage's October tracking has 7628 +RMB 1.5/m, G75 +RMB 3,000/t, 2116/1080 thin cloth +RMB 2.5/m. Tightness in e-fabric, HVLP copper foil and high-speed laminate persists into 2027, and a second CCL round in Q4 remains live. |
| Copper & copper foil | SHFE copper RMB 107,590/t (-1.27%), a second weekly fall with LME copper -1.74% to 14,098 — both exchanges weakening together this time. But no new concentrate TC reading; the -237.80 USD/dmt historic low is maintained, and LME copper inventories turned to a build of +46,350t to 272,950t, with US warehouses holding 169,475t or 62%. That is tight ore against loose stocks: the price fall is tariff-expectation disappointment plus a broadly weaker dollar-priced commodity complex, not supply relief, so downside is capped. HVLP4 processing fees held at RMB 35,000/t with high-end HVLP foil still short by 15-25% — capacity, not copper, sets this price |
| Tin / solder & plating | SHFE tin RMB 376,030/t (-0.62%) — sharply narrower than last week's -3.50% and still below 380,000; LME tin 50,000 (-0.25%). The tin decline is a high-level pullback, not a downtrend: cumulatively -4.10% from the Oct 9 level of 392,110, and the base case is a RMB 370,000-380,000/t range, so do not chase it. Solder and plating cost relief continues but at a slower pace — keep the locking rhythm normal. Tantalum capacitor cost is driven by concentrate and African ore freight, a different source, so this is not a reason to expect tantalum relief |
| Nickel / MLCC electrodes | SHFE nickel RMB 119,650/t (+3.32%) — the only Shanghai base metal to gain this week, ending two consecutive falls. MLCC internal-electrode nickel-powder cost pressure returns, though nickel powder is priced on conversion cost and pass-through still takes one to two quarters. Stainless structural parts follow the same direction. Note the irony: the one metal that rose is the one that feeds the one component already in a 52-week lead-time regime |
| FR-4 CCL | Order books remain full with rationing in place, and price sustainability extends through 2027. Given the fabric verification result, a second CCL price round in Q4 is now the base case rather than a tail risk. Downstream demand stays structurally bifurcated — see the capacity evidence opposite |
| 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 single largest materials-side risk for Q4. Saudi Arabia is also negotiating crude loading routes outside Hormuz, so resin logistics now carry a two-variable exposure. 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 they are constrained by different things. Metal prices are set on a global exchange against macro expectations, currency moves and tariff speculation — which is why copper fell on tariff-catalyst disappointment (the Sep 28 Section 232 decision is now 25 days overdue and suspended) plus a broadly weaker dollar-priced commodity complex. Fabric, resin and high-end foil are set by how much capacity exists and whether anyone will build more. That is why 7628 rose again while copper fell, and why the foil processing fee did not move at all.
The inventory picture reinforces it. LME copper stocks built +46,350t to 272,950t while concentrate TC sat at a historic low — that is metal in a warehouse and ore that is not there. Copper is the one place where a genuine easing could still show up, and even there, the processing fee tells you it will not reach the foil price. For the rest of the chain, treat the metals tape as a negotiating argument on the base-material line, not as a cost forecast.
On demand, the bifurcation that keeps every PCB sub-tier out of the green zone persists: Chinese PCB makers disclosed more than RMB 70bn of expansion investment in 1H 2026, essentially all of it aimed at AI servers and high-speed optical modules, with traditional consumer-electronics expansion at zero. Taiwan's Q1 2026 PCB output reached TWD 245.6bn (+19.6% YoY, a record for a first quarter). High-end capacity takes 12-18 months to build, so the 2027 gap does not close on this capital cycle. Where substitution is genuinely moving the needle is connectors — domestic 224G is now shipping in volume on qualified platforms and a major domestic maker has committed RMB 612 million to a second expansion phase — but certification still runs 12-18 months, so it is negotiating leverage and a 2027 capacity option.
6Segment Snapshot
Memory DRAM/NAND RED
MLCC Ceramic Capacitors RED
Tantalum Capacitors RED
MCU / Analog / Power RED
PCB / CCL RED
Foundry / Semi Materials RED
Connectors / RF ELEVATED
IMU / Sensors NEW
Metals / Freight / FX
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 |
|---|---|---|
| MLCC — 52-week lead timesSamsung E-M 30 → 52 weeks; price is no longer the lever | This is the single most consequential change of the week, and it is not a price. At 52 weeks, ordinary price negotiation is structurally irrelevant — the only mechanisms left are allocation and long-term booking. Move your objective from negotiating price to securing quota: submit allocation requests in writing and early, and start 2027 framework-volume conversations now, because the benchmark is being reset across the whole industry rather than by one vendor. Note that nickel, at +3.32%, is also the one metal that rose — the cost floor under MLCC moved up in the same week that the lead time went to 52 weeks. | Generic 0402/0603 and consumer X5R remain decoupled from AI high-cap — never infer AI high-cap pricing from generic parts. Rationalize part numbers for allocation leverage, and qualify a mainland alternative in the industrial tier where CR5 concentration hurts most. With the domestic price gap closing, re-price alternatives at current levels rather than reusing prior assumptions. |
| Taiwan export-control listAI servers and advanced semiconductors added; mainland transfers need a licence | This is a compliance task with a hard deadline, not a sourcing task. Run a Taiwan-origin review over every AI-server-related part on your BOM and establish, for each, whether it falls under the new control list, whether a licence path exists, and what the traceability documentation requirement is. Re-examine any transshipment or third-country shipping arrangement that relies on an intermediate node. Violations carry criminal liability, which changes the internal approval calculus entirely. Do this before the late-October contract settlements, because a part that cannot be licensed cannot be contracted for. | Qualify non-Taiwan-origin equivalents for the affected line items now — a second source that is already qualified is worth more than a cheaper part that needs a licence review. Build the documentation trail as you go; it becomes reusable evidence for every future audit. |
| Taiwan power / MCU+10-15% planned, still unannounced — the last open window, now narrowing | Second consecutive week we flag this, and the window is visibly closing. Panjit and TSC plan +10-15% on October spot products and, as of Oct 23, still have not issued a notice — the pricing window closes the moment they announce. Complete new-order negotiation with the Taiwan makers this week and obtain written magnitudes and effective dates, so their increase does not stack passively on onsemi's already-live +5-15%. Treat any further delay as a decision you will have to defend later. | Cross-brand pin-to-pin on MOSFET / IGBT / SiC; package arbitrage; qualify second sources on long-lead SiC now rather than after the announcement. Renesas's Jan 1, 2027 repricing belongs in 2027 budgeting, not in this cycle's scramble. |
| Memory DRAM / DDR5Broad rally; DDR4 8Gb +2.47% turns up; spread narrows to $14.09 | The framing changes: this is now a broad rally, not a divergence story, and that removes the comfort of "the old tier is cheap." With DDR4 8Gb up 2.47% and eTT up 4.32%, every generation is rising together, which means the substitution arbitrage is narrowing across the board. Convert reservations to confirmed POs before the 4Q26 contract settlements close in late October — and note that TrendForce's Oct 21 read has NAND at +15-20% and enterprise SSD at +23-28%, so the contract side is not softening. For NAND-based content specifically, the cross-session print at USD 18.592 says mark your forecast down — that is the one place the tape is still correcting. | Continue the DDR4→DDR5 migration where the design allows, but re-price it — the gap has narrowed to USD 14.09, so the saving is smaller than two weeks ago. Freeze the memory sub-BOM early; qualify the pin-compatible successor for industrial parts now. Use the 30-day and year-to-date figures, not a single week's print, to size any long-term commitment. |
| Geopolitics — Red Sea / SaudiHouthi strikes 4x in 48h; tanker war-risk <3% → 10% of ship value | The oil price fell and the risk did not — do not let the Brent decline reset your buffer assumptions. Super-tanker war-risk went from below 3% to 10% of ship value and Gulf-to-Asia tanker freight set a record at USD 37.37/bbl. List war-risk and reroute fuel as separate BOM line items with the full three-part cost — rate (0.35%-0.5% of ship value per transit on Red Sea / Gulf-of-Aden routing, versus roughly 0.1% at the start of the year, plus the new 10% super-tanker rate), surcharges, and rerouting cost of USD 2.5-4.5 million per voyage with fuel still 58-60% above pre-war. Re-confirm per-shipment insurability on Europe, Middle East and Gulf lanes and negotiate a rate cap with overrun sharing, because a step change of this size is rarely grandfathered. | Hold air-freight contingency; pre-book alternative routings; for Europe lanes keep a Suez-return option (saves 10-14 days and roughly 30% of fuel) where the carrier still offers it. Re-price Europe-line exports upward — Europe and Mediterranean container rates rose 2.7% and 8.3% while US lanes fell 0.8% and 1.1%, so shift volume to US lanes where the service allows. |
| PCB / CCL & metals7628 up again to 13.5-14; Cu -1.27% but TC -$237.80 | Two negotiations that must not be confused, and one that is now the base case. On the firm side: 7628 moved up again to RMB 13.5-14/m in the third week of execution verification, and Kingboard's eighth 2026 round is live — a second CCL price round in Q4 is now the base case, not a tail risk, so do not relax lock-in on high-speed laminate. On the softer side: tin's decline is narrowing (-3.50% then -0.62%) and the base case is a RMB 370,000-380,000/t range, so keep solder and plating adjustment clauses on the table but do not build a strategy on further tin weakness. On copper: do not discount the copper-input line by the -1.27% — the concentrate TC is unchanged at a historic low and the foil processing fee did not move at all. | 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 cost line items — this week they moved in four different directions, and a single blended assumption would have been wrong on three of them. |
| Connectors 224GDomestic 224G volume delivery confirmed; RMB 612m phase-two expansion | Domestic 224G has crossed from validation into volume supply on qualified platforms, and a major domestic maker has now committed RMB 612 million to a second expansion phase across 17 lines. This is the moment to start qualification paperwork rather than feasibility talks, and to sequence second sources against a 12-18 month certification clock. But do not plan on relief this cycle — capacity release takes two to three quarters, and the effective high-speed interconnect gap is still above 75%. | Domestic 224G / 112G substitution; cross-brand pin-compatible alternates; consolidate high-speed connector part numbers to improve leverage; keep one incumbent-qualified line as a fallback for safety-critical platforms. |
8Risk Matrix — Likelihood × Impact
Attention allocation: act immediately in the top-right, monitor continuously in the bottom-left.