A-Share Pre-Market Brief | 2026-10-05 Monday

Market holiday notice: National Day market holiday in effect (markets closed 10/1 through 10/7; trading resumes as usual from Thursday 10/8, SSE announcement¹). This list is the 3rd holiday special issue (no issue on the 10/3-10/4 weekend): time window = morning of 10/2 (Friday) → this morning; use it together with Issues 1 and 2; all conclusions are priced for the 10/8 reopening. Progress bar of holiday pricing: US markets have 5 trading days during the holiday (10/1-10/7); 2 are done, 3 remain (10/5-10/7). The biggest macro event of the holiday has landed: on the evening of 10/2, US September nonfarm payrolls came in at just +29,000, a major miss, and October rate-hike expectations collapsed from about 70% early in the week to about 22% at Friday's close; but "long-end Treasuries rose instead of falling after weak data" (10Y closed at 5.27%, 30Y at 5.62%), and Hong Kong stocks already fell 2.6% on 10/2 — global valuation pressure is unresolved. The 10/8 reopening is a hedge between "cooling rate-hike expectations + RMB strengthening to 6.71 + nearly 70% positive Q3 pre-announcements" and "high long-end rates"; the strength ranking completes a swap this issue: optical communications > storage. Data conventions: A-share quotes are as of the 9/30 close; US stocks as of the 10/2 close (US Eastern), i.e., early morning of 10/3 Beijing time; fundamentals are the 2026 interim reports + announcements on the evening of 9/30. Price limits: main board ±10% | ChiNext/STAR ±20% | BSE ±30%; every stock below is tagged with its board, and limit-up calls follow each board's limit. Corrections this issue (vs Issue 2): ① The meeting held 10/26-29 is the Fifth Plenary Session of the 20th CPC Central Committee (misstated as the Fourth Plenary Session in Issue 2, Lianhe Zaobao 9/21²); ② For the YoY rise in DDR5 spot prices, the "+480%" cited in Issue 2 was an old caliber from late August; at the 10/2 average of USD 58.00 versus spot a year ago (in the 6.8-7.7 USD range), YoY is roughly +650% to +750% (TrendForce³).

---

0. One-line summary for today

Day 5 of the holiday; the biggest macro variable of the holiday has landed: US September nonfarm payrolls +29,000, far below the expected 84,000-90,000; unemployment rose to 4.2%; average hourly earnings +3.0% YoY, the weakest since May 2021 (BLS⁴) — October rate-hike expectations collapsed Friday to about 22% (a December 25bp hike remains the mainstream scenario at 73.5%-80%+, Polymarket 10/4⁵); US stocks closed higher Friday (Nasdaq +1.19%, a record closing high), but the bond market reversed in the second hour, with long-end Treasuries closing higher, not lower (2Y 4.82%, 10Y 5.27%, 30Y 5.62%) — "the bond market has slipped out of the Fed's control" (JRJ⁶). The main-line strength swapped during the holiday: Friday's US session saw two straight days of real gains in optical communications (Coherent +10% Thursday then another +5.59% Friday, Lumentum +3.79%, AAOI +7.71%, CRDO +4.03%, Corning +2.35%, optical-communications ETF LYTE +3.09%), while storage hardware was slammed (Western Digital -10.22%, Seagate -10.21% — Toshiba announced it would double capacity at its Philippines plant, triggering a "supply expansion" trade; SanDisk -3.79%, Micron -2.05% on profit-taking; only SK Hynix ADR rose, +0.84%) — strength ranking for the 10/8 reopening: optical communications > storage. Storage fundamentals are intact: TrendForce raised its Q4 contract-price forecasts across the board (public note: overall DRAM +15-20%, NAND +15-20%; via Goldman Sachs: PC DRAM +13-18%, enterprise SSD +23-28%); Micron's FQ1 guidance of USD 61.5 billion massively beat expectations, and the CFO confirmed over 75% of 2027 output is already locked in — Friday's US decline was in the HDD/NAND supply narrative; for the A-share storage chain (niche DRAM / interfaces / modules), 10/8 sentiment hinges on how Korean names open. Oil prices are being pulled both ways: the G7 announced a release of up to 100 million barrels of reserves over 4 months plus the US dropping its diesel export ban pressured crude (Brent settled Friday at 102.25, -2% on the week), while the Red Sea escalated again over the weekend (106 Saudi airstrikes in 24 hours, a tanker attacked), pushing Brent up +1% at Monday's open to 103.3 USD. Domestic policy was quiet and holiday consumption was mild (commercial-district revenue +5.3%, duty-free sales +8.85% in the first three days, National Day holiday box office at a 13-year low). Driver types: macro (nonfarm payrolls) · real overseas gains (optical communications) · overseas divergence (storage) · prices & geopolitics (oil & gas) · quiet domestic policy. Tone: re-review daily; final version on the evening of 10/7.

---

1. News roundup

Slow-variable tracking: ① 3 US trading days remain in the holiday (10/5-10/7) + the FOMC minutes at the early hours of 10/8; ② a December hike is still the mainstream scenario — whether long-end Treasuries (10Y 5.27% / 30Y 5.62%) can fall back is the lifeline for growth stocks on 10/8; ③ the two-way oil switch: Red Sea / Strait of Hormuz attack frequency vs the pace of G7 reserve releases (heavy diesel in the first 20 days); ④ FTSE A50 futures previous settlement 14,022 points (Eastmoney²⁸), offshore RMB 6.7065 (sharply stronger this year, Investing.com²⁹); ⑤ Hong Kong market movement over 10/5-10/7 (Stock Connect absent — sentiment preview only); ⑥ Mate90 daily first-sales / official scorecards; ⑦ the 10/8 reopening brings a dense batch of 134 Q3 reports + TSMC's September revenue; ⑧ the Fifth Plenary Session of the 20th CPC Central Committee on 10/26-29; ⑨ the revised Integrated Circuit Layout-Design Protection Regulations take effect on 10/15.

---

2. Strongest bullish branches, ranked

Branch dictionary for the whole issue: optical communications / optical modules (two straight real gains during the holiday, up to No. 1) | storage supercycle (first overseas divergence, fundamentals intact) | oil & gas / tanker shipping (two-way tug-of-war) | semiconductor equipment / domestic chain | Huawei chain (Mate90 shortages + JAC confirming talks; no verifiable main-attack target this issue) | auto chain (September sales) | shareholder increases, buybacks & events | consumer & tourism (mild holiday data). Each stock belongs to exactly one branch; Sections 3 and 7 are consistent with this section. Branch strength uses S/A+/A/B+/B/C; a stock's bullish level may float half a notch up or down from its branch strength based on company-level evidence (e.g., ACM Shanghai at B+ is above its branch's B because its backlog is direct, company-announcement-grade evidence).

Branch notes:

• Optical communications (No. 1, upgraded this issue): This is the first time in the holiday that "the same branch posts sector-wide real gains on two consecutive US trading days" — Coherent +10% on 10/1 and another +5.59% on 10/2, Lumentum up two days in a row, AAOI (small-cap optical components) +7.71%, and NVIDIA's two-day rally to an intraday high endorsing total compute demand. Two consecutive days of capital behavior is harder evidence than any narrative; the A-share optical-module chain is a high-certainty mapping for the 10/8 reopening. But note this issue's evidence is entirely price action — Innolight / Eoptolink have no new company-level announcements during the holiday (order-type hard catalysts are absent; quality control has logged this as a data gap). The suppressors are unchanged: long-end Treasuries at high levels, and volatility across the holiday's remaining 3 US trading days.
• Storage (No. 2, downgraded this issue): You must separate the "trading" layer from the "fundamentals" layer. Trading: Friday's US session saw a rare single-day plunge in the HDD/storage-equipment chain (Western Digital / Seagate -10%); the trigger was Toshiba's announcement that it would double capacity at its Philippines plant before FY2027 — the market traded "supply will eventually expand." SanDisk and Micron (up on 10/1 after earnings, then -2.05% Friday on profit-taking) fell along; only SK Hynix ADR closed higher, showing capital split between "DRAM scarcity" and "NAND/HDD supply expansion." Fundamentals: TrendForce raised Q4 contract-price forecasts on 10/2 (overall DRAM +15-20%, NAND +15-20%; eSSD +23-28% via Goldman Sachs); Micron's FQ1 guidance of USD 61.5 billion (up another ~13% QoQ), the CFO's confirmation that over 75% of 2027 output is locked in, and HBM sold out through 2027 — the makers' multi-year commitments show no loosening whatsoever. The A-share storage chain is mainly DRAM niche (GigaDevice), interfaces (Montage), and modules (Longsys / BIWIN) — a different line from Friday's US NAND/HDD supply narrative, though opening sentiment will inevitably be dragged; on the morning of 10/8, watch how Korean stocks open at 9:00 (SK Hynix / Samsung). All contract prices and guidance above are institutional forecasts / company guidance, not accomplished facts.
• Oil & gas / tanker shipping (No. 3): Forces on both sides strengthened simultaneously during the holiday — pressing: the G7 releasing up to 100 million barrels over 4 months (heavy diesel in the first 20 days, directly hitting diesel crack spreads) and the US dropping the diesel ban; pushing: the weekend Red Sea escalation (106 Saudi airstrikes, tanker explosions, Brent opening at 103.3). Note the core driver of tanker shipping is not oil prices but the effective capacity lost to blockades: TD3C equivalent time-charter rates broke USD 1 million/day in mid-September (1.099 million on 9/15, 1.2125 million on 9/17, +425% vs pre-conflict; Q2 average USD 427,000/day, +1008% YoY) — no public prints for the latest rates since October (they eased ahead of the holiday on de-escalation hopes and heated up again over the weekend); actual levels to be verified; if rates hold at the plateau, a high QoQ rise for the two state-owned tanker majors in Q3 is a high-probability event (an inference, not a disclosure), but if there is a ceasefire / resumed navigation, rates will fall just as fast. Keep at B+, a two-way branch.
• Huawei chain (No. 5): Mate90 high-end versions staying out of stock says demand is not weak, but the constraint of "no official sales figures" is unchanged; supply-chain (CIS / panels / PCB) name-drops still lack authoritative teardown sources. JAC Motors' "confirmed talks" is a high-position rumor-grade confirmation (talks ≠ orders) — fine for holiday fermentation, not fine as a main attack. No verifiable main-attack target in this branch this issue.
• Auto chain (No. 6): BYD's September numbers are positive (record single month, exports 180,700), and Q2 core per-vehicle profit hit a recent-year high after adjusting out about RMB 8 billion in forex losses (Tencent News 8/31³⁰ — broker/media analyst caliber, not announcement figures); but the cumulative January-September figure is -3.94%, the price war continues — keep at B.

---

3. Overall stock bullishness ranking

Ranked in descending order by the Section 4 scoring model (whole-market coverage; main board, ChiNext, and STAR treated alike; ordered by evidence strength). Holiday special: all conclusions target the 10/8 reopening. Changes vs Issue 2: Zhongji Innolight 中际旭创 +3 to No. 1 (optical communications' two straight real gains), Eoptolink 新易盛 +4 to No. 3 (GB300 1.6T second-source share + gross margin above Innolight + Q2 QoQ +70.8%), TFC 天孚通信 +2, COSCO Shipping Energy 中远海能 +1 (Red Sea escalation + freight-rate plateau), Giantec 聚辰股份 -2 (most of the profit surge comes from fair-value changes and other non-recurring items — a temperature gap vs the DDR5 narrative); GigaDevice / Montage / Longsys / BIWIN unchanged.

---

4. Stock scoring model

Total score = sum of seven base components (capped at 95) + risk deduction (0 to -15). Components and weights: source authority 0-15 / benefit directness 0-20 / earnings elasticity 0-15 / industry position & fundamentals 0-15 / expectation gap 0-10 / theme durability 0-10 / A-share trading traits 0-10 / risk deduction 0 to -15.

(Stocks below No. 11 get simplified ratings, same caliber; sub-scores omitted — see Section 3 for totals.) Scoring notes: Zhongji Innolight risk deduction -7 (operating cash flow -44%, cash-to-profit ratio 0.13 — the divergence from profit must be counted); Montage -8 (the Korean prosecutors' antitrust raid hanging overhead + valuation double-kill risk); COSCO Shipping Energy durability just 6 (rates depend on the war and negotiation progress — more event-driven than storage's "makers' multi-year commitments"), risk -9 (G7 releases + ceasefire both point down, and there are no public rate prints for October); Giantec -9 (of the RMB 530 million net profit attributable to parent, about RMB 480 million is one-off items such as fair-value changes; revenue only +4.7%, gross margin down YoY). All fundamental data were independently cross-checked via the akshare financial-statement API (2026 interim reports; cross-check date 2026-10-05) — see Section 5.

---

5. Top 10 stocks in detail

51 Zhongji Innolight 中际旭创300308Optical comms · No. 1 overall · priority deep-dive (A+/74) · ChiNext ±20%

• Related news: The second day of the optical-communications sector's two straight real gains on 10/2 (US Eastern) — Coherent +5.59%, Lumentum +3.79%, AAOI +7.71%, CRDO +4.03%, Corning +2.35%, optical-communications ETF (LYTE) +3.09% (Eastmoney⁷); NVIDIA +1.34%, an intraday record of USD 237.88, endorsing total compute demand. No new company-level announcements during the holiday.
• Bullish logic: A direct-order-type positive. Cloud-vendor capex → 800G ramp + 1.6T starting; about 94% of the company's revenue comes from overseas cloud vendors. The two straight sector-wide real gains are the most direct sentiment and pricing mapping for the 10/8 reopening — what they affect is order visibility and valuation, not pure sentiment.
• Branch stage: Late fermentation nearing climax. A-share optical modules showed huge volume with resilience on 9/30; the two overseas gains leave room for a gap-up at the reopening. The formation is complete — heavyweights (Innolight / Eoptolink) + components (TFC) — with no limit-up ladder yet (no limit-up on 9/30): a "trending main upward move," not a consecutive-limit-up theme.
• Fundamental verification: H1 revenue RMB 41.78 billion, +182.5%; net profit attributable to parent RMB 13.65 billion, +241.7%; Q2 single-quarter net profit RMB 7.92 billion, QoQ +38.1%; gross margin 46.3% (akshare cross-check). The flip side you must see at the same time: H1 operating cash flow of just RMB 1.80 billion, -44% YoY, cash-to-profit ratio ~0.13 — receivables/inventory pressure under rapid revenue growth is common to this cycle and is counted in the risk deduction.
• Industry position: Global No. 1 in optical modules by share (LightCounting 2025 sales caliber); Eoptolink second.
• Technical sentiment: ChiNext ±20%; 9/30 volume of RMB 12.85 billion, price resilient — chips changed hands ahead of the holiday; the gap-up magnitude at the reopening depends on whether optical communications can keep rising over the holiday's remaining 3 US trading days.
• Final call: The name with the hardest real-gain evidence this holiday — priority deep-dive; verification points = the gap-up magnitude and volume absorption at the 10/8 open; if the gap-up exceeds 8%, watch for intraday divergence.

52 GigaDevice 兆易创新603986Storage · No. 2 overall · priority deep-dive (A/72) · main board ±10%

• Related news: TrendForce raised Q4 contract-price forecasts on 10/2 (overall DRAM +15-20%, NAND +15-20%, ITHome¹³, forecast caliber); DDR5 16Gb spot average USD 58.00, YoY roughly +650% to +750% vs a year ago (TrendForce³); Friday's US storage hardware was slammed but SK Hynix ADR rose alone +0.84% — capital split between "DRAM scarcity vs NAND/HDD supply expansion," and GigaDevice sits on the kept side.
• Bullish logic: Niche DRAM + NOR follow-through benefits directly, affecting quarter-by-quarter gross margin (Q2 already 66.6%) and full-year earnings upgrades; not a sentiment mapping.
• Branch stage: The first overseas divergence within an ongoing main upward move. Makers' multi-year commitments are unchanged; the first real US decline appears on the trading side. 10/8 opening sentiment will be set by how Korean names (SK Hynix / Samsung) open.
• Fundamental verification: H1 revenue RMB 11.57 billion, +178.7%; net profit attributable to parent RMB 6.86 billion, +1091.5%; Q2 gross margin 66.6%; operating cash flow RMB 6.05 billion (best in the chain); debt ratio just 11.5% (akshare cross-check). NOR flash is global No. 2 (share ~18.5-20%, Omdia / Frost & Sullivan caliber; Issue 2's "top three" phrasing was conservative).
• Industry position: NOR global No. 2; China's leading memory designer; a scarce mainland niche-DRAM target.
• Technical sentiment: Main board ±10%; ~47% pullback from the July high, PE (TTM) ~31x — better position and odds than the chain's high-valuation names; 9/30 volume down -3.72% completed earnings-risk hedging.
• Final call: The storage-chain name with "the smallest overseas-sentiment shock and the hardest fundamentals" — priority deep-dive; risk = Friday's US storage sentiment dragging the first 30 minutes of the 10/8 open.

53 Eoptolink 新易盛300502Optical comms · No. 3 overall · priority deep-dive (A/68) · ChiNext ±20%

• Related news: Same as Zhongji Innolight (two straight real gains in optical communications); at the company level it won a second-source share of NVIDIA's GB300 1.6T, with small-batch validation in Q2 and capacity ramping in H2 (Sina 7/19³¹), plus 2026-2030 long-term agreements with Meta / Microsoft / Amazon (Caixin³²).
• Bullish logic: Direct-order type. 1.6T is the 2027 elasticity source; the LTAs lock in the demand base; under two straight real gains, its elasticity is the same source as Innolight's.
• Branch stage: Same as Innolight (trending main upward move).
• Fundamental verification: H1 revenue RMB 20.91 billion, +100.3%; net profit attributable to parent RMB 7.53 billion, +91.0%; gross margin 48.4%, 2.1pct above Zhongji Innolight; Q2 net profit RMB 4.75 billion, QoQ +70.8% (akshare cross-check).
• Industry position: One of the optical-module duopoly, ~20% global share in 800G.
• Technical sentiment: ChiNext ±20%; pulled back with the sector before the holiday, endorsed by two overseas gains; more elastic than Innolight, and more volatile.
• Final call: The best "quality-to-valuation match" within optical communications — priority deep-dive; risk = concentration in five North American customers (~95-96% overseas share), with no announcement evidence for the exact single-customer share — do not cite media estimates.

54 Montage Technology 澜起科技688008Storage / interfaces · No. 4 overall · priority deep-dive (A/67) · STAR ±20%

• Related news: Q4 server DRAM contract-price forecast +10-15% (TrendForce¹³, forecast caliber); Micron's HBM sold out through 2027, with over 75% of 2027 output already locked in (Wallstreetcn relaying the earnings call³³) — total server-memory expansion is the "volume and structure" logic for interface chips.
• Bullish logic: Rising DDR5 penetration and larger per-module capacity → RCD / SPD / MRCD interface chips gain in both unit count and value; it gains from volume, not price hikes — high earnings quality.
• Branch stage: The most fundamentally solid "picks-and-shovels" name, not directly hit by Friday's US HDD/NAND supply narrative.
• Fundamental verification: H1 revenue RMB 3.34 billion, +26.7%; net profit attributable to parent RMB 2.00 billion, +72.3%; gross margin 65.3%; debt ratio 7.5%; operating cash flow RMB 1.33 billion (akshare cross-check). The memory-interface triopoly (Montage / Renesas / Rambus) holds a combined share above 93%, with Montage first globally at 36.8% (Frost & Sullivan).
• A risk fact that must be added: On 2026-07-16, Korean prosecutors launched an antitrust raid on three memory-interface chip companies; Montage's A/H shares plunged over two days (Leida Finance et al.³⁴; no new developments seen as of publication — to be verified) — the regulatory overhang is a permanent valuation discount until the investigation lands.
• Technical sentiment: STAR ±20% (500,000 CNY / 2-year access threshold); PE (TTM) ~80x with high growth priced in; pulled back with STAR on 9/30.
• Final call: Priority deep-dive, but treat it position-wise as a "high-score white horse," not an "elasticity theme"; verification points = any new developments in the Korean raid and the final magnitude of Q4 server DRAM contract prices.

55 Longsys 江波龙301308Storage · No. 5 overall · watch closely (A/66) · ChiNext ±20%

• Related news: Q4 contract-price forecasts raised — enterprise SSD +23-28% via Goldman Sachs (JRJ¹⁴, forecast caliber); Friday's US declines in Western Digital / Seagate (-10% area) were triggered by Toshiba's supply expansion — a sentiment negative on the NAND/HDD line, not the same source as Longsys's "low-cost inventory + price pass-through" profit logic, but opening sentiment will be infected.
• Bullish logic: Profit = (spot/contract spread) × inventory revaluation; the H1 gross margin of 58.9% (61.3% in Q2) comes exactly from this; enterprise SSD (revenue RMB 2.14 billion, +208.8%) opens a second curve.
• Branch stage: A high-beta name in the first divergence of an ongoing main move.
• Fundamental verification: H1 revenue RMB 24.09 billion, +136.3%; net profit attributable to parent RMB 10.58 billion, +71529% YoY (only RMB 10 million in the prior-year period); the flip-side numbers are just as glaring: operating cash flow -RMB 3.15 billion, inventory RMB 25.78 billion = 60.1% of total assets, debt ratio 56.1% (all as of 6/30; Q3 changes await the Q3 report) (akshare cross-check). No third-party ranking source was found for "China's No. 1 enterprise SSD"; the accurate phrasing is "one of the few independent module makers in China with scaled shipments of the eSSD + RDIMM combination."
• Industry position: China's leading independent module maker (dual brands).
• Technical sentiment: ChiNext ±20%; already pulled back ~13% in late September; PE (TTM) ~11.6x is a classic cyclical low-PE trap.
• Final call: Watch closely; verification points = whether it can quickly recover after an opening dragged by US storage sentiment on 10/8, and the Q4 contract-price pass-through guidance for gross margin.

56 BIWIN Storage 佰维存储688525Storage · No. 6 overall · watch closely (A/65) · STAR ±20%

• Related news: Same as Longsys; the company swung to an H1 profit of RMB 7.17 billion, within the 7/15 pre-announcement range (7.0-7.5 billion), with figures cross-checked by akshare as true — against the backdrop of the storage supercycle + low-cost-inventory dividend (full-year 2025 net profit attributable to parent was only RMB 850 million).
• Bullish logic: Edge-AI storage (phones / PCs / wearables) capacity upgrades + price pass-through; wafer-level packaging and testing bring cost and customization moats.
• Branch stage: Same as Longsys; the stock fell 15% on the pre-announcement day (7/16) — the market is already trading "the dividend is unsustainable."
• Fundamental verification: H1 revenue RMB 15.57 billion, +298.1%; gross margin 54.8%; operating cash flow -RMB 6.26 billion, inventory RMB 18.17 billion = 54.5% of assets, debt ratio 61.6% — the triple-peak signal is more concentrated than Longsys's (all as of 6/30) (akshare cross-check). "Global No. 1 in edge-storage modules" was found only in prospectus marketing; the verifiable Frost & Sullivan caliber is "the world's only independent storage-solutions provider with wafer-level packaging and testing capability" and "global No. 2 in embedded storage."
• Industry position: A leading independent module maker in edge storage.
• Technical sentiment: STAR ±20%; the interim dividend of RMB 4.24 per 10 shares is rare for a cyclical name, showing management's confidence in cash flow.
• Final call: More elastic than Longsys and riskier than Longsys — watch closely; once storage sentiment ebbs, the triple-peak structure amplifies the drawdown.

57 TFC Optical 天孚通信300394Optical comms · No. 7 overall · watch closely (B+/61) · ChiNext ±20%

• Related news: Two straight real gains in optical communications (components move first: AAOI's +7.71% Friday is the small-cap component weathervane); CPO / optical engines and MPO connectors are positioned for the 1.6T generation (Caixin Weekly³⁵).
• Bullish logic: A platform-type optical-components supplier benefiting from the rising component value in the 800G → 1.6T transition; not a direct one-to-one comp with any single company.
• Industry stage / fundamentals: H1 revenue RMB 2.83 billion, +15.1%; net profit attributable to parent RMB 1.20 billion, +33.9%; gross margin 60.9%; operating cash flow RMB 1.00 billion; debt ratio 17.6% (akshare cross-check) — growth clearly below Innolight / Eoptolink, yet the most expensive valuation: PE (TTM) ~122.5x (9/30 close of RMB 260.77).
• Technical sentiment: ChiNext ±20%; pulled back before the holiday, endorsed by two overseas gains.
• Final call: Good quality, poor odds — watch closely; suited to buying pullbacks, not to chasing the gap-up on overseas real gains.

58 ACM Shanghai 盛美上海688082Semiconductor equipment · No. 8 overall · watch closely (B+/59) · STAR ±20%

• Related news: Backlog RMB 17.073 billion, +88.2% YoY (announced the evening of 9/30, as of 9/29; the company notes it is unaudited, announcement PDF³⁶).
• Bullish logic: Orders → revenue conversion over 12-18 months; domestic No. 1 in cleaning-equipment share (60.5% among domestic makers); electroplating as the second curve.
• Fundamental verification: H1 revenue RMB 3.72 billion, +13.9%; net profit attributable to parent RMB 990 million, +42.1%, but ex-non-recurring profit RMB 570 million, -15.31% YoY — the headline net-profit surge is mainly government subsidies and other non-recurring items; H1 operating cash flow of only RMB 90 million; PE (TTM) 84.5x (akshare + Tencent Quotes cross-check). A "surging 50%" headline misleads — you must read it on the ex-non-recurring caliber.
• Technical sentiment: STAR ±20%; announcement-grade orders catalyst — the most recognizable equipment name.
• Final call: The orders are real, the earnings quality is hollow — watch closely; verification points = the pace of order-to-revenue conversion and when ex-non-recurring growth turns positive.

59 COSCO Shipping Energy 中远海能600026Oil & gas / tankers · No. 9 overall · watch closely (B+/57) · main board ±10%

• Related news: Weekend Red Sea escalation (106 Saudi airstrikes in 24 hours, a tanker exploding south of Al-Mokha, CLS.cn Telegraph¹¹); Brent opened Monday +1% at 103.3; on 9/27 the Strait of Hormuz saw just 1 transit vs a baseline of 85 ships/day (Reuters¹²).
• Bullish logic: The core driver is the effective capacity lost to blockades → freight-rate risk premium: TD3C equivalent time-charter rates broke USD 1 million/day in mid-September (1.099 million on 9/15, 1.2125 million on 9/17, +425% vs pre-conflict; Q2 average USD 427,000/day, +1008% YoY); if rates hold at the plateau, high QoQ growth in Q3 is a high-probability event (an inference — the company has not disclosed it); note that it earns freight rates, not oil prices, so the correlation with the G7 releases pressuring oil is weak. No public rate prints since October (rates eased ahead of the holiday on de-escalation hopes and heated up again over the weekend) — to be verified.
• Fundamental verification: H1 revenue RMB 15.15 billion, +30.0%; net profit attributable to parent RMB 4.54 billion, +143.2% (media figure +141% — the report recomputation prevails); foreign-trade tanker gross profit RMB 4.67 billion, +262.4% (akshare + interim-report cross-check); the world's largest oil-tanker fleet (long-standing broker caliber).
• Technical sentiment: Main board ±10%; tanker rates eased ahead of the holiday on de-escalation hopes and heated up again with weekend events — an event-driven rhythm.
• Final call: A watch-closely two-way branch; verification points = attack frequency and negotiation progress over 10/5-10/7 — any ceasefire rumor is the take-profit exit signal.

510 Giantec Semiconductor 聚辰股份688123Storage · No. 10 overall · watch closely (B/53) · STAR ±20%

• Related news: Q4 DRAM contract-price forecasts raised (forecast caliber); DDR4 discontinued by makers, DDR5 becoming the mainstream for servers / high-end PCs (OFweek³⁷), with SPD / EEPROM benefiting from the penetration.
• Bullish logic: DDR5 SPD global share above 40% (company HK prospectus caliber) — every DDR5 stick ships with an SPD, a pure "volume" logic.
• Fundamental verification: H1 revenue RMB 600 million, just +4.7%; net profit attributable to parent RMB 530 million, +156.1%, but akshare income-statement cross-check shows fair-value-change gains of about RMB 475 million and investment income of about RMB 4 million — one-off items contributed the vast majority of profit; on an ex-non-recurring basis, core profit is only on the order of RMB 100 million; gross margin 51.5%, down YoY (60.3% in the prior-year period). The big profit surge shows a clear temperature gap with the "DDR5 explosion" narrative — Issue 2's score did not reflect this enough and it is downgraded this issue.
• Industry position: A DDR5 SPD leader (prospectus caliber), EEPROM global No. 2 (17.1%, Web-Feet Research).
• Technical sentiment: STAR ±20%; small-cap elasticity; pulled back with the sector on 9/30.
• Final call: Watch closely; the logic holds but earnings have not delivered — add weight only after post-interim-report data verification.

---

6. Pass list

---

7. Intra-branch ranking

Optical communications / optical modules (Branch No. 1 this issue)

Follow-on relationship: on overseas real-gain days, Innolight / Eoptolink set the direction, TFC follows with components, Zhishang / Everbright supply the elasticity, and Naruida is a pure follower. Hardest = Zhongji Innolight; most recognizable = Zhongji Innolight (Eoptolink is the most elastic under capacity constraints); Pass = Naruida; Zhishang = watch only.

Storage supercycle (Branch No. 2 this issue)

10/8 branch call order: first watch how Korean stocks (SK Hynix / Samsung) open at 9:00, then the A-share call auction; GigaDevice (design / niche) is least infected by Friday's US NAND/HDD sentiment, while the module duopoly is the most elastic but also carries the biggest triple-peak drawdowns. Pass = Ingenic; Demingli = watch only.

Oil & gas / tanker shipping (Branch No. 3 this issue): Hardest = COSCO Shipping Energy (600026 — direct rate elasticity); most recognizable = COSCO Shipping Energy; defensive alternative = China Merchants Energy Shipping (601872 — LTA + dividend profile, low elasticity); linked watch = CNOOC (600938 — earns oil prices but directly suppressed by G7 releases). A ceasefire rumor = the profit-taking signal for the whole branch.

---

8. Opening verification signals for the first reopening day (10/8)

Auction signals: ① Gap-up magnitude for the optical-module chain (Innolight / Eoptolink / TFC) — against two straight real gains, a 3-6% gap-up is the reasonable range; be wary of profit-taking above an 8% gap-up; ② if the storage chain (GigaDevice / Longsys / BIWIN) opens lower on inertia from Friday's US sentiment but recovers within 30 minutes, that is the best "mispricing repair" confirmation; ③ one-word boards and sealed order books (limit-ups sealed from the open — the so-called "20cm" one-word boards — are common in overseas-real-gain branches; mind the ChiNext / STAR ±20% caliber, and don't mistake +12% for limit-up).

Sector signals: ① optical communications: 3 or more limit-ups before 10:00 with the heavyweight (Innolight) holding volume without blowing up = main line confirmed; ② storage: watch the 9:00 Korean open and its linkage with A-shares; ③ tankers: watch whether weekend events keep fermenting (re-check every trading day over 10/5-10/7); ④ index level: FTSE A50 futures (previous settlement 14,022) + Hong Kong's 10/5-10/7 moves (Stock Connect absent — sentiment preview only).

Stock signals: For the top 5 names, absorption (gap-up without filling the gap), large net inflows, and heavyweight-plus-ladder rising together are strong; a single gap-up (one lone name) is weak.

Risk signals (any one triggers a position cut): ① optical communications turns lower on the remaining US trading days (the two real gains falsified); ② long-end Treasuries make a new high (30Y above 5.70%) triggering a synchronized global growth de-rating; ③ US–Iran ceasefire / Strait of Hormuz reopening rumors (the tanker logic reverses); ④ Hong Kong keeps plunging over 10/5-10/7 (a reopening discount priced by overseas capital); ⑤ earnings blowups among the dense Q3 reporting batch (134 names from 10/8).

Must-do homework for the rest of the holiday (every evening, 10/5-10/7): re-check the US optical-communications / storage sectors → update this list's strength ranking; watch US–Iran negotiations and Red Sea attack frequency; watch Mate90 official scorecards; final version on the evening of 10/7.

---

9. Final recommendation

① The 5 most worth watching this issue (for the 10/8 reopening; selected by branch diversification: the model's top 4 + No. 9 COSCO Shipping Energy manually added to cover Branch 3; the model's No. 5-8 (Longsys / BIWIN / TFC / ACM) move down accordingly and are all watch closely)

② The 3 strongest branches this issue

③ Directions not to chase this issue: film & cinema (day-one National Day holiday box office at a 13-year low — a data negative); the p-nitrotoluene anti-dumping concept (no direct A-share capacity confirmation); Huawei-chain rumor stocks (JAC Motors high after 3 limit-ups in 6 days + "talks ≠ orders"; Will Semiconductor with no BOM evidence and H1 mobile CIS revenue -31%); lithium batteries / solid state (no holiday catalysts, and a cooling announcement on the evening of 9/30); any diesel-petrochemical chain directly hurt by the "G7 releases."

④ Final one-line judgment: The holiday enters its second half; the most dangerous macro hurdle (nonfarm payrolls) has made a soft landing (October rate-hike expectations collapsed to 22%), and the main lines completed the strength swap to "optical communications > storage" — for the 10/8 reopening, structure around "optical communications as the main attack, storage bought on dips as Korean names set the tone, tankers event-driven," but long-end Treasuries at 5.3-5.6% and Hong Kong's 2.6% plunge on 10/2 remind us: profit-taking at the reopening's gap-up will not be small — position size matters more than direction.

---

⚠️ Risk disclaimer: This list is only a pre-market information roundup and watchlist, and does not constitute investment advice. A-share volatility risk is extremely high; automatically generated content may carry information-timeliness gaps or industry-chain mapping errors, and must not be used directly as a trading basis.