A-Share Pre-Market Brief | 2026-09-09 Wednesday

Scope boundaries of this issue (please gauge the strength of the conclusions accordingly) ① News window = 2026-09-08 15:00 (A-share close) → 2026-09-09 07:10. News outside the window that is still being priced in is listed separately with a "pricing progress" note and is not treated as a new positive. ② A-share quotes are on the 2026-09-08 close basis (Tencent qt.gtimg.cn); the limit-up / failed-limit-up / limit-down pools are East Money's official pools push2ex, with len(pool)==tc verified (73==73). ③ The Dragon-Tiger List is the complete 2026-09-08 dataset (59 listed entries, 39 institutional-seat entries), released after 18:00 and therefore available to the pre-market slot — this is the question yesterday's post-close recap could not answer, and the one this issue is responsible for answering. ④ Limit-up determination is tiered by board price limit: Shanghai/Shenzhen main board ±10%, ChiNext/STAR Market ±20%, Beijing Stock Exchange ±30%. Every stock in this issue is tagged with its board. ⑤ Domestic futures are on the 2026-09-08 daily settlement basis; for night-session products (crude oil / copper / coke / corn / soybean meal) an additional 09-09 early-morning reading is given and labeled. Urea, white sugar, cotton and rapeseed meal have no night session, so their readings are still the 9/8 day session and will only update during today's trading. ⑥ ⚠️ CNBC international agricultural products (wheat / corn / soybeans / sugar) currently show volume=0, i.e. these are 9/8 settlement prices rather than live overnight prices; this issue uses them only in their "settlement price" sense, never to describe overnight moves. ⑦ No northbound flow data (daily net buy disclosure was discontinued from 2024-08-19); this issue gives no northbound figures.

⚠️ Data-sourcing and search log (this section is not sent to clients):

I. Channels that worked

• Quotes via Tencent qt.gtimg.cn; historical candles via Tencent web.ifzq.gtimg.cn/appstock/app/fqkline (East Money stock_zh_a_hist returned Connection aborted for the whole batch — the sixth consecutive trading day of cluster failure).
• Sector/stock fund flows via push2delay.eastmoney.com — following the lesson written back by the previous issue, "when clist returns 502, try the backup domain first," this worked on the first attempt and there was no need to fall back wholesale to Tonghuashun. That memory is now closed-loop.
• Dragon-Tiger List stock_lhb_detail_em + stock_lhb_jgmmtj_em both normal (59/39 entries).
• Announcements via East Money np-anotice-stock list + np-cnotice-stock full text, using eiTime rather than notice_date to fix the publication moment.
• US equities/commodities via the CNBC quote endpoint.

II. Channels that failed

• ⚠️ cninfo.com.cn returned 502; announcement full text was rerouted to East Money np-cnotice-stock, which returned the complete text, so conclusions were unaffected.
• ⚠️ cls.cn/nodeapi/telegraphList returned 404 (endpoint changed); CLS newswire was rerouted to ak.stock_info_global_cls, which returned 20 items.
• ⚠️ ak.index_us_stock_sina(".VIX") raised IndexError; VIX history was not obtained, so §1.3 only makes a conditional statement about the VIX phantom prior close and asserts nothing.
• ⚠️ ak.futures_zh_realtime('wheat')/('methanol') raised KeyError — those product names are not in the new Sina mapping table; wheat was switched to CNBC @W.1.

III. Five cross-checks actively performed on method, all of which changed conclusions

1. ⚠️⚠️ [Most important item in this issue] The attribution of COHR's surge — the search engine stitched in another piece of stale news. Searching "Coherent stock surge September 8 2026" returned a summary stating that "the FCC's proposed ban on Chinese optical module imports is the main catalyst." Verification showed that story is from Bloomberg/Reuters/Caixin coverage dated 2026-08-04/05, more than a month ago. Had it been copied over, the entire attribution of today's optical communication branch would be false, and it would imply exactly the opposite direction (negative for A-shares). This is the same trap as the previous issue's "phosphorus chemicals = a six-month-old administrative order"; memory [[ai-attribution-stitches-old-news]] has now recurred on two consecutive days. Changed to: the FCC ban is handled as an "existing regulatory overhang risk" and is not treated as today's catalyst; COHR's true same-day catalyst is not fully closed-loop, so only the verifiable first-hand GLW/Verizon announcement is accepted.
2. ⚠️⚠️ [Yesterday's "agriculture catalyst not identified" is now closed-loop — and it was the company itself that explained it] 万向德农 (Wanxiang Doneed, 600371)'s 9/8 17:46 unusual-movement announcement names it explicitly: "the market has been discussing climate topics such as a super El Niño extensively." From that, a reverse search found that the WMO confirmed on 2026-09-03 that El Niño had formed, with an 81% probability of developing into a super-strength event in October–December. This is the catalyst the previous issue failed to find on two consecutive days. More importantly, in the same announcement the company itself denied the transmission: "the relevant topic mainly concerns commodity corn… the company's main business is corn seed, which sits upstream in the industry chain and belongs to a different link from the commodity corn market," and it described the seed industry as being in a "destocking cycle with supply exceeding demand." Attribution and falsification come from the same first-hand document.
3. ⚠️ The previous issue's figure "urea −0.22%, zero futures confirmation" was inaccurate. Daily data show UR2701 closed at 1822 on 9/8, +0.33%, and +2.82% over the last 5 trading days; white sugar SR2701 was +0.92% on 9/8, +2.32% over 5 days, and its 9/8 high of 5510 was a window high. The previous issue's 15:34 live snapshot read an intraday price rather than the settlement price. The conclusion must be softened: it is not "zero confirmation," but "confirmation is weak and exists only on the soft-commodity leg."
4. ⚠️ The previous issue's cited "wheat at 757.25 cents, the only agricultural product genuinely rising" has already been given back. @W.1 settled at 747.00 on 9/8, −1.35% versus that reading. The reason is that the reading taken by the previous issue on the afternoon of 9/8 Beijing time was in fact the 9/4 (Friday) settlement price — 9/7 was the US Labor Day holiday, so that reading was already one trading day stale at the time.
5. ⚠️ Tanker shipping is not "completely unpriced." The first draft said "the geopolitical catalyst has not yet been priced by A-shares." After checking the 60-day range: 招商轮船 (China Merchants Energy Shipping, 601872) is +15.78% over the last 20 days, 中远海能 (COSCO Shipping Energy Transportation, 600026) +24.11%, 招商南油 (Nanjing Tanker, 601975) +29.09%, with 60-day percentiles of 66.7 / 73.1 / 84.1 respectively. The theme is already running; the new catalyst is an accelerator, not an igniter — the framing has been revised throughout §2 and §5.

IV. What this issue did not obtain / did not do (listed honestly)

• The true prior close of the VIX was not obtained. CNBC gives prev=15.30, but 15.30 is suspected to be the phantom close of the 9/7 Labor Day holiday. If the true prior close is the 9/4 reading, the VIX's actual rise is far larger than the apparent +2.75%. This issue quotes only the level 15.72 and does not cite a percentage change.
• The same-day first-hand catalyst for COHR's +7.10% is not closed-loop. The stale FCC story has been ruled out; media mention "high-voltage silicon carbide for AI data centers," but no company announcement or press release from that day was found to corroborate it. Treated as "attribution unknown."
• The reason 中天科技 (Zhongtian Technology, 600522) / 亨通光电 (Hengtong Optic-Electric, 600487) / 烽火通信 (FiberHome Telecommunication, 600498) fell 36%–41% over the last 60 days and are halved from their highs was not identified. This is the largest information gap in this issue's recommendation of that branch — a stock down 52% may have been unfairly sold off, or its fundamentals may genuinely have deteriorated, and this issue is not able to distinguish between the two. It is explicitly flagged in §5 and position sizing language is limited accordingly.
• VLCC freight rates (BDTI/TD3C) were not obtained; the tanker branch has only two legs — the event and the share price — and lacks the hardest leg, freight rates.
• The publishing sector catalyst is still not closed-loop with a first-hand source. The "State Administration of Press and Publication's Implementation Rules for the Cultural Digitalization Industry" that was found comes from a Sina Finance AI-generated limit-up analysis page, a low-credibility source that this issue does not accept as fact; the branch is downgraded on that basis.

V. Problems found by risk-auditor in quality control and already fixed (the real error list of this issue's first draft)

Before finalizing, risk-auditor flagged 18 red flags, of which 8 were substantive errors in the first draft, all now fixed:

1. ⚠️⚠️ In the §7A scoring table, 5 of 10 rows did not add up, with 中粮糖业 (COFCO Sugar) and 奥浦迈 (OPM Biosciences) each off by 20 pts — and the ranking is built on that score, meaning the ranking was fake. Every row has been recomputed and the §6 master ranking re-sorted (招商轮船 73 / 中远海能 69 / 长飞 67…), plus two framing notes added (the seven components sum to a maximum of 95, not 100; negative factors go uniformly into the "risk deduction" column rather than as negative values in positive components).
2. ⚠️⚠️ The correction on 中京电子 (Zhongjing Electronics) was applied only to §4, while §7 and §9.3 still carried the refuted "1 buy 3 sell" — a half-finished correction is more dangerous than no correction, because readers lower their guard on the rest of the document once they see "this issue has already corrected itself." Now synchronized document-wide. Process lesson: after every correction, run a full-text search on the keywords that were changed.
3. ⚠️⚠️ The criterion explicitly voided in §8.1 ("whether 中天/亨通 follow the rally") was resurrected by the document itself in §8.2 — a client executing on §8.2 would take the opposite action to §8.1. Changed to use the directional alignment of 长飞's A/H shares as the criterion.
4. ⚠️⚠️ Removing 中天/亨通 used a 2025FY basis to negate the 2026H1 upcycle, and switched rulers: the two were negated using "revenue share" while 烽火 was affirmed using "segment gross margin" — yet 烽火's 23.9% share is less than double 中天's 14.0%. More seriously, the internal log said "the reason for the decline was not identified," while the client-facing text said "it most likely comes from the copper/grid businesses" — this is exactly "I couldn't find it" being quietly rewritten as "it didn't happen." All such statements have been downgraded to pending verification, and the reason for removal changed to the defensible formulation "the original bull case (low position) does not constitute a bullish logic."
5. ⚠️ The two arguments for "the Corning contract is a competitive negative" cancel each other out: if the two markets are already bilaterally walled off, then the loss from "removing one customer" is zero. Also, the original exclusivity clause text was not obtained. Changed to "direction of marginal impact unclear."
6. ⚠️ Two entries in the §1.1 "in-window news" table were out of bounds (the GAC customs data was released on 9/8 and the 亚盛集团 announcement on 9/7), even though this issue specifically created §1.2 to hold out-of-window items and then did not use it. This directly undermines the terminal judgment that "the only hard in-window information is geopolitics" — the window boundary was loose in our own hands. Labels have been corrected.
7. ⚠️ Compliance: the explicit position recommendation "half position" appeared, in direct conflict with the disclaimer and the CLAUDE.md hard rules. It has been deleted, and the risk disclosure expanded from 2 lines to 6 items (including reading timestamp, incomplete-coverage statement, suitability threshold, holdings statement), plus a new §10 known-limitations table with 12 items.
8. ⚠️ Multiple uses of "the only" do not hold, of which "金正大 is the only one of the four with PB>2" is directly falsified by this issue's own data (云天化 PB 2.29); separately, "万向德农 +5.04%" had been written into "the three that hit limit-up." All instances corrected.

Two further items where this issue accepted the QC direction but reached a different conclusion:

• The "60% of 奥浦迈's +230% net profit is paper gains" figure of 60% does not reconcile under either calculation: on a net-profit-attributable-to-parent basis it is 49.2% (6,098.58/12,403.26), and on an incremental basis 70.5%. That figure is cited 5 times in the text and is the core reason for removing the recommendation slot. Uniformly changed to 49.2% with the basis stated.
• The August CPI cannot adjudicate an El Niño catalyst that was only released on 9/3 — this QC point is very well taken: August CPI measures prices from 1–31 August and physically cannot include climate expectations formed after 9/3, yet the first draft wrote it as triggering a hard action of "no participation in the entire agriculture branch." The whole passage has been rewritten.

VI. QC points not adopted / only partly adopted by this issue

• QC argued that "all eight news links in §1.1 point to the cls.cn homepage, so clients cannot verify them." CENTCOM's official release and the original NBC article have been added, but the two IRGC statements still exist only as Chinese-language paraphrase, and each is now labeled "no English first-hand source obtained" inside the table.
• QC suspected that "the same set of institutional names appearing on the sell side for both 奥浦迈 and 上海电影 may be a row-misalignment in data retrieval." The 上海电影 set was pulled by this agent itself via stock_lhb_stock_detail_em, not relayed by a sub-agent, and has been rechecked as correct; Shanghai-Hong Kong Stock Connect, Guotai Haitong HQ and Goldman Sachs (China) are all high-frequency active seats, so appearing on the sell side of two stocks on the same day is normal. But QC's point that "RMB 193 million vs RMB 28.77 million was written up as an equal-weight contradiction" is indeed a wording error, and has been corrected to a magnitude mismatch.

---

0. One-sentence summary of the day

The strongest catalyst is geopolitical, not industrial: on 9/8 local time, US forces destroyed 5 Iranian crude oil tankers; Iran promptly warned that it would strike tankers berthed at ports in Kuwait and Bahrain and demanded that crews evacuate, while also hitting the US Azraq air base in Jordan with ballistic missiles; layered on top are OPEC's sharp August production drop and the simultaneous disruption of two Saudi export routes. This entire chain occurred after the A-share close on 9/8, and is the hardest, most verifiable incremental information inside the window. WTI is quoted at USD 94.26 (+1.32%, live quote at 19:02 ET), with none of yesterday's gain given back — the verification point the previous issue set for the energy line, "falsified if more than half the gain is given back," is judged not falsified before the open.

The second line is optical fiber, but its appeal is not tonight's news — it is a set of earnings disclosed three weeks ago: the US optical communication complex rallied broadly overnight — Lumentum +11.04%, Corning +7.56%, Coherent +7.10% — and only Corning has a closed-loop first-hand announcement (on 9/8 it signed a 2027–2032 supply agreement with Verizon for more than 80 million miles of optical fiber, amount undisclosed). But after verification it must be cooled down: this contract is a bilateral procurement deal between Corning and Verizon, in which A-share companies play no role whatsoever; and the China–US optical fiber markets already carry bilateral trade barriers (⚠️ the specific tariff rates and effective dates were not obtained from first-hand government documents in this issue, and are labeled pending verification), so the direction of its marginal impact on Chinese manufacturers is unclear — it is neither an order pass-through nor safely asserted to be a competitive negative.

The evidence that actually holds up is in the earnings reports, and it is dual-sourced: 长飞光纤 (YOFC, 601869)'s 2026H1 ex-non-recurring net profit attributable to parent was +1,680%, with gross margin rising from 28.30% to 53.36% (26Q2 single quarter 60.52%); 烽火通信 (FiberHome, 600498)'s optical fiber and cable segment gross margin was +16.78pp over the same period on revenue +48.0%. Two independent sets of financials point the same way, proving that optical fiber price increases are industry-wide. ⚠️ But it must also be said plainly: 长飞's results were made public on 21 August, and the stock is already +261.9% year to date — tonight's news is lagging confirmation for it, not new information.

⚠️ This issue's single most important correction happened on this line: the first draft recommended 中天科技 (600522) and 亨通光电 (600487) as "the three optical fiber names, with the lowest positions"; after checking segment reports, optical communication accounts for only 14.0% and 8.4% of their revenue, and their main businesses are submarine cable, power grid and copper conductors — they are not on this transmission chain, and their recommendation slots have been withdrawn.

Where the money went: institutional seats turned around yesterday, and turned cleanly. The previous issue's hardest evidence was that "the top 7 net institutional buys were all inside the compute-hardware complex"; today's answer is: not anymore. The top three institutional net buys on 9/8 were 上海电影 (Shanghai Film, 601595) (+RMB 193 million, 2 buy 0 sell), 集泰股份 (Jointas Chemical, 002909) (+RMB 71 million) and 奥浦迈 (OPM Biosciences, 688293) (+RMB 54 million, 3 buy 0 sell); in PCB only 中京电子 (002579) remains, and within its institutional seats it is 2 net buys vs 1 net sell (see §4); optical modules had zero institutional seats on the list.

Driver ranking: geopolitical event > overseas industry > climate theme > policy (already refused pricing).

Pre-market state judgment: today is neither a continuation day nor a rotation day; it is an "add-a-leg" day. Yesterday's high-to-low rotation (heavyweight tech sold, agriculture/resources bought) was neither falsified nor reinforced overnight: FTSE China A50 night session +0.10%, essentially flat; the three US indices closed lower (Dow −1.18%, S&P −0.58%, Nasdaq −0.32%) but with extreme internal divergence — optical communication and semiconductor equipment rallied hard while Nvidia −2.01% and Micron −1.61% fell. August CPI is released today at 09:30, simultaneous with the open — ⚠️ but it measures August prices, while the El Niño alert was issued on 9/3, so it has no adjudicating power over this round of the agriculture narrative; it can only be read as a pre-price-increase baseline.

---

1. News overview

1.1 In-window news (2026-09-08 15:00 → 2026-09-09 07:10)

1.1B ⚠️ Market backdrop (2026-09-08 close, the reference frame for branch strength)

⚠️ What this table means for today: 9/8 was a "index down, stocks up" high-to-low rotation day — the ChiNext Index was −1.15% and STAR 50 −1.52%, while 3,257 stocks rose and 0 hit limit-down. But the failed-limit-up rate has already doubled from 17.0% to 33.6%, so the difficulty of relay trading is at a high level, and that is the main constraint on chasing limit-ups today. Also note: in the limit-up sector distribution there were only 2 telecom equipment names and 2 optics & optoelectronics names — there was no limit-up wave in optical communication on 9/8, so this issue's second branch is built on a direction that "made no money yesterday."

1.1C ⚠️ One macro backdrop this issue's first draft badly underestimated: the Fed is being priced for a rate hike

⚠️ This is not background noise; it has three direct consequences that this issue's first draft did not handle at all:

1. It may be the actual reason the Dow's −1.18% materially underperformed the Nasdaq's −0.32%, rather than the "structural divergence" this issue originally wrote.
2. It is a headwind for every high-PB growth stock — pointing directly at this issue's third recommendation slot 长飞光纤 (601869) (PB at a historical high) and at the withdrawn 奥浦迈.
3. Directionally it supports relative outperformance by resource/energy value stocks, aligned with this issue's first branch.

⚠️ But what this issue failed to do must be stated: the rate-hike expectation leg was not incorporated into the §7A scoring model; it appears only in 长飞光纤's risk deduction. This is a known structural defect of this issue.

1.2 ⚠️ Two items out of window but still being priced (listed separately, not treated as new positives)

---

2. Strongest positive branches in descending order

---

3. ⚠️ First, the prerequisite question the previous issue demanded: is today a continuation day or a rotation day?

Item 1 of §6.4 of the previous issue required: "you must first answer whether tomorrow is a continuation day or a rotation day, then choose the ranking variable," on the grounds that the same "relative strength + institutional flows" ranking systematically inverts on a rotation day. This section answers that first.

Answer: neither — today is an "add-a-leg" day. Three pieces of evidence:

1. The overnight session delivered no directional verdict. FTSE China A50 futures night session +0.10% (14,578 points), essentially flat; US indices closed lower but with extreme internal divergence (optical communication +7% to +11% vs Nvidia −2.01%), so the information content at the index level is close to zero while the information content at the structural level is large.
2. Yesterday's rotation itself was neither extended nor overturned. Yesterday's strong agriculture and chemicals got no new confirmation from the commodity side overnight (urea/white sugar have no night session, and international agricultural products show volume=0); yesterday's weak tech diverged overseas (equipment up, compute chips down). Neither side received a new verdict.
3. The main increment is a leg unrelated to either of the above: geopolitical tanker shipping. It belongs neither to yesterday's strong side (agriculture) nor to yesterday's weak side (tech), so it is not bound by the continuation/rotation dichotomy at all. ⚠️ Strictly speaking there are two increments — the Corning announcement (9/8, first-hand, in window) is also new information — but it points to optical fiber, which is adjacent to yesterday's weak side, and after verification it turns out to be only lagging confirmation of industry β for A-shares, so it is not listed as a separate leg.

Direct implication for the ranking variable: since today is not a single-state day, a single ranking variable should not be used. This issue therefore uses different primary ranking variables for different branches, and states explicitly in §7 which ruler each branch uses:

• Tanker shipping uses event timing + position (the event is new, the position is no longer low);
• Optical fiber uses position + catalyst hardness (position extremely low, catalyst medium-hard);
• Agriculture uses commodity confirmation (the entire disagreement sits on this dimension);
• Pharma uses institutional seat direction (one of the few directions where institutional seats were one-way buyers on 9/8).

---

4. ⚠️ The question the previous issue specifically asked the Dragon-Tiger List to answer: are institutions still buying PCB / optical modules?

Item 2 of §6.4 of the previous issue: "The Dragon-Tiger List is the pre-market slot's unique informational edge, and tomorrow morning it must be used to answer: are institutional seats still buying PCB/optical modules, or have they turned around and bought agriculture and chemicals?"

Answer: neither — institutions turned around and bought media, chemical materials and pharma, and institutional participation in PCB and optical modules has fallen to near zero.

There are 39 institutional-seat buy/sell records for 2026-09-08. ⚠️ In the raw table, 集泰股份 and 我爱我家 each appear twice and 欧福蛋业 (Oufu Egg Industry) three times (the same stock generating multiple rows because of multiple listing reasons); statistics were computed after de-duplicating by code, so as not to repeat the earlier error of "adding the buy and sell lists together and overstating the institutional share."

Four conclusions that must be read out:

1. The optical module branch has zero institutional seats. 中际旭创 (Innolight, 300308, ChiNext ±20%), 新易盛 (Eoptolink, 300502, ChiNext ±20%), 光迅科技 (Accelink, 002281, Shenzhen main board ±10%) and 天孚通信 (T&S Communications, 300394, ChiNext ±20%) all failed to appear on the 9/8 Dragon-Tiger List. The previous issue's pattern of "the top 7 net buys were all inside the compute-hardware complex, most with 3–4 institutional seats" has completely disappeared.
2. In PCB only 中京电子 remains, and institutions are internally split. Seat-by-seat verification: the three "institution-dedicated" seats were respectively a net buy of +RMB 60.14 million, a net sell of −RMB 29.52 million, and a net buy of +RMB 6.74 million, totaling +RMB 37.36 million (reconciling with the aggregated figure of RMB 37.37 million), i.e. 2 net buyers and 1 net seller. The net amount is only 0.94% of that day's RMB 3.974 billion turnover — this is a divided book, not accumulation. The institutional story in PCB has been told.

⚠️⚠️ This issue must publicly correct one of its own errors and explain a data trap that will be stepped on repeatedly. The first draft relied on the "buy-side institution count / sell-side institution count" fields of stock_lhb_jgmmtj_em and wrote 金正大 as "1 buy 2 sell" and 中京电子 as "1 buy 3 sell," concluding from that they were divided books. After pulling stock_lhb_stock_detail_em seat by seat, both turned out to be wrong:

• All three of 金正大's institutional seats were in fact net buyers (+RMB 38.66 million / +RMB 1.56 million / +RMB 7.66 million = +RMB 47.87 million). The latter two appeared on the sell list only because their sell amounts were large enough, but their net amounts were still positive.
• 中京电子 was 2 net buy 1 net sell, not 1 buy 3 sell.

Root cause: that field counts "which list a seat appears on," not "the net direction"; and the same seat can appear on both the buy and sell lists and be counted twice. This is the same trap as "the Dragon-Tiger buy and sell lists have overlapping rows, so adding them directly overstates the institutional share," in a different guise. Lesson: any qualitative judgment about institutional direction must drop down to seat-by-seat net amounts and cannot use the aggregate table's count fields. 奥浦迈 in §6.1 and 上海电影 and 亚盛集团 in §4 of this issue have all been re-verified seat by seat to this standard (results: 上海电影 2 buy 0 sell confirmed, 亚盛集团 0 buy 2 sell confirmed).

3. ⚠️ 亚盛集团, on its 4th consecutive limit-up, was net sold by institutions, 0 buy 2 sell. A stock on a 4-day limit-up run with a seal-to-turnover ratio of 6.244 and a one-word limit board had one-way selling from institutional seats. This corroborates its liquidity condition of RMB 89 million turnover and a volume ratio of 0.09: the seal belongs to retail and hot money, and institutions are distributing.
4. The only clean one-way buying was 奥浦迈 (3 buy 0 sell) and 上海电影 (2 buy 0 sell). But their credibility must be judged separately: 上海电影's controlling shareholder denied all material matters in writing that same evening (item 13 of §1.1), so the institutional buying has no fundamental landing point; 奥浦迈, after seat-by-seat verification, has a cleanly established institutional direction but an overstated intensity (see the note below and §6.1).

⚠️ 奥浦迈's "share of turnover" must be given as two numbers, not just the higher one. The 3 institution-dedicated seats bought a combined RMB 53.5029 million and there were no institution-dedicated seats on the sell list — that part is entirely true. But the whole-list net buy was only RMB 24.4814 million, or 4.60% of that day's total turnover, rather than the 10.05% on the institutional basis: the Shanghai Connect dedicated seat appeared on both the buy and sell sides (bought RMB 48.2654 million / sold RMB 40.9170 million, net inflow only RMB 7.3485 million), and the sell list also contained Guotai Haitong HQ at −RMB 17.57 million, Goldman Sachs (China) Shanghai at −RMB 12.38 million, CICC Shanghai at −RMB 11.34 million, and CITIC Shandong Qingdao at −RMB 10.96 million. Conclusion: a clean direction ≠ strong absorption. This is the same class of basis trap as "the Dragon-Tiger buy and sell lists overlap, so adding them directly overstates the institutional share."

---

5. Branch detail and key evidence

5.1 Branch 1: tanker shipping / oil & gas — the only S-grade event chain inside the window, but not the starting point

Event chain (all occurring after the 9/8 A-share close):

• US Central Command posted on social media: US forces "destroyed" 5 Iranian crude oil tankers on 9/8 local time; over the preceding two days the IRGC had twice attacked a US warship with ballistic missiles.
• IRGC warning: in view of the US strikes, crews of tankers berthed at ports in Kuwait and Bahrain should evacuate immediately, and both anchorages and ports will be targets.
• IRGC statement: it has fired ballistic missiles at the US Azraq air base in Jordan.
• Survey data: OPEC crude output fell sharply in August, ending two consecutive months of recovery; two Saudi oil export routes are simultaneously under threat.

Price confirmation (independent of share prices):

⚠️ The verification point the previous issue set for the energy line can already be judged pre-market: the original wording was "whether WTI/Brent can hold tonight — if international oil prices give back more than half of today's gain, this branch is falsified at tomorrow's open."

Verdict: not falsified. Two independent calculations point to the same conclusion:

• WTI: quoted at USD 94.38, above its 9/8 settlement of 93.03, i.e. the change is positive and the give-back is zero (this can be judged without knowing the 9/8 gain base).
• Shanghai crude SC2610: 9/8 day session +3.05% (close 720.6), night session 714.7 = −0.82%, so the give-back is about 27% of that day's gain, far short of half.

⚠️ Basis note: this issue did not obtain WTI's gain base for 9/8 (CNBC's previous_day_closing is the 9/8 settlement of 93.03, while 9/7 was US Labor Day), so on the WTI side the criterion was switched to "is it above the 9/8 settlement," which does not depend on a base; the original criterion that requires a base is fulfilled on the SC side.

⚠️ One piece of background that changes the branch's characterization and that this issue's first draft was entirely unaware of: this is not an isolated event but an ongoing US–Iran conflict. The CENTCOM release states that the destruction was a response to "two IRGC ballistic missile attacks on US warships within the past two days"; US media are already covering it continuously under a "2026 US–Iran war" frame (NPR had related coverage as early as 9/5), and headlines about "oil approaching USD 100" have appeared. This means:

• ✅ The branch's durability is stronger than that of a "single event";
• ⚠️ but it also means this line has been running for more than one day, consistent with the price fact that the three tanker names are up 15.78%–29.09% over the last 20 days — this issue's claim of "not yet priced" applies only to the batch of new events on the night of 9/8, not to the whole theme;
• ⚠️ and it cuts both ways for tanker companies: as freight rates rise, war-risk insurance premiums and vessel asset safety deteriorate in step, a layer this issue has not quantified.

⚠️ But it must equally be made clear that this is not the starting point:

⚠️ One piece of flow evidence that runs opposite to the share prices: on 9/8 the shipping & ports sector had a main-force net outflow of RMB 284 million (with the sector up +0.66%), and all three tanker stocks had volume ratios ≤1.08. In other words, before the attacks happened, money was flowing out of tanker shipping. There are two ways to read this, and this issue does not pick a side:

• Bullish reading: the event occurred after the close, so yesterday's outflow has nothing to do with today's event, and today is a brand-new pricing starting point;
• Bearish reading: the sector was already distributing on shrinking volume, and the geopolitical event merely provides one opportunity to spike and unload.

Criterion (distinguishable intraday today, written into §8): watch whether the event premium converts into turnover, not the size of the gap-up.

5.2 Branch 2: optical fiber / optical communication — lowest position and freshest catalyst, but "optical fiber" and "optical modules" must be separated

Overnight US market (all at the 9/8 close, with volume and 10-day average volume for comparison):

⚠️ The one thing most worth reading out of this table: what rallied is the "physical layer of light" (fiber, optical components, lasers), and what did not rally is the "chip layer of light" (Marvell's DSP/custom ASIC, +0.83% on shrinking volume). The previous issue listed MRVL/COHR/GLW side by side as adjudicators; in practice the three gave completely different answers, and they cannot be merged into a single conclusion that "optical communication rallied."

First-hand catalyst (Corning, 2026-09-08): Verizon and Corning announced a multi-year, multi-billion-dollar supply agreement to supply more than 80 million miles of high-density optical fiber and connectivity solutions from 2027–2032, for broadband expansion and the nationwide long-haul backbone required by AI hyperscale customers; the agreement is intended to support Corning in further expanding its US domestic manufacturing footprint. ⚠️ Financial terms were not disclosed; "multi-billion dollars" is the company's own qualitative wording, and this issue gives no specific amount.

⚠️ Grading under hard rule 2: this is an "industry trend positive," not a "direct order positive." The reasons must be stated: ① this is a US carrier procuring from a US supplier; ② one stated purpose of the agreement is to expand US domestic capacity; ③ A-share companies have no named role whatsoever in it. Its significance for A-shares is a single point — it validates that the optical fiber demand created by the AI long-haul backbone is a multi-year 2027–2032 cycle, not a one-off tender.

⚠️ The regulatory overhang must be counted separately: the FCC proposes banning imports of Chinese-made new-model optical modules (2026-08-04/05 draft, intended to take effect within the year; Chinese suppliers account for roughly 60% of global datacom optical module revenue). That draft targets optical transceivers (optical modules), not optical fiber and cable. Therefore:

• Optical modules (中际旭创, 新易盛, 天孚通信, 光迅科技): the higher the overseas revenue share, the greater the regulatory risk;
• Optical fiber and cable (中天, 亨通, 长飞, 烽火): that risk is essentially inapplicable.

A-share position comparison (the strongest dimension of this branch):

⚠️⚠️ But fundamental verification overturns the way this table can be used — 中天科技 and 亨通光电 are not optical fiber names at all.

After pulling the four companies' 2026 interim segment data, this issue found that the very phrase "the three optical fiber names" is wrong:

⚠️ Basis note: 长飞 and 烽火 are 2026H1 segment data; 亨通 and 中天 have not disclosed 2026H1 segment data, so the table uses the 2025FY basis, which is not the same period as the first two and is used only to illustrate business-mix shares, not for direct comparison of gross margin levels.

This single fact simultaneously explains three things that otherwise contradict each other:

1. Why 中天 and 亨通 closed lower against the trend on 9/8 (−0.64%, −1.39%) — their main drivers are submarine cable and power grid, not optical fiber;
2. Why they are down 36% over the last 60 days — ⚠️ this issue still has not identified the reason and makes no assertion. A plausible but unverified hypothesis is that the decline comes from the copper/grid/submarine cable businesses that make up 70%–80% of revenue rather than from optical fiber. This is only a hypothesis; this issue is not able to verify it, and readers should not treat it as explained.
3. Why their overall gross margins barely moved (16.48% / 16.77%) — this is business-mix dilution, not a falsification of the industry.

⚠️ On this basis this issue withdraws the recommendation slots for 中天科技 and 亨通光电. The first draft treated "lowest position + lowest PB" as a bullish reason, whereas the reason the position is low is precisely that their real main businesses are weakening, and optical fiber at 8%–14% of revenue cannot rescue the income statement. This is a textbook case of "right branch, wrong stock."

✅ Conversely, the optical fiber price increase itself is independently double-verified and real:

• 长飞光纤 2026H1: revenue RMB 9.809 billion +53.64%, ex-non-recurring net profit attributable to parent RMB 2.449 billion +1,680.48%, gross margin 28.30% → 53.36% (+25.06pp); 26Q2 single-quarter gross margin 60.52%, rising monotonically for six quarters. The company attributes it in its own words to "both product sales volume and average price increasing" (volume and price rising together, not a price increase alone).
• 烽火通信 2026H1 optical fiber and cable segment: gross margin 25.48% → 42.26% (+16.78pp), revenue +48.0%. This is a second evidence source independent of 长飞, proving the price increase is industry-wide and not one company's accounting artifact.

⚠️⚠️ And the Corning–Verizon catalyst must be downgraded one further notch after verification — on the North American leg it is marginally negative:

1. This is a bilateral contract locking Verizon's next 6 years of optical fiber supply to Corning. For Chinese manufacturers it does not add contestable North American demand; on the contrary it removes one large customer for 6 years.
2. Bilateral trade barriers exist between the two markets. According to media reports, China levies anti-dumping duties on US-made G.654.C single-mode optical fiber (Corning 37.9%, OFS 33.3%, Draka Americas and others 78.2%). ⚠️ This item is labeled "pending verification," and it contains an internal date contradiction: the cited "levied from September 2025" and "valid until 2028-04-21" are incompatible (anti-dumping measures normally run 5 years), and the latter looks more like the expiry date of some earlier measure. This issue did not obtain the first-hand MOFCOM announcement, so this fact is used qualitatively only and supports no quantitative conclusion. ⚠️ On the reverse direction, US tariffs on Chinese optical cable — two searches in this issue failed to obtain a first-hand US government document, so it is labeled "pending verification" and this issue does not cite Chinese-media paraphrase.
3. ⚠️ For 长飞, this news is "lagging confirmation" rather than new information. 长飞's 26Q2 gross margin of 60.52% and ex-non-recurring profit of RMB 1.988 billion were already publicly disclosed in the interim report on 2026-08-21, and the stock is already +261.9% year to date. Corning's contract merely confirms after the fact something the market has been trading for three weeks.

⚠️ One further note: claims circulating on Chinese social media that "长飞 is the only Chinese manufacturer capable of large-scale delivery to North American cloud providers" and that it "uses a Mexican plant to circumvent US restrictions" have no corresponding statement anywhere in the company's own disclosures, and this issue does not accept them. 长飞's geographic disclosure has only two columns, "Chinese mainland / other," and the company does not disclose a North American revenue share at all, so the transmission path cannot be verified.

The correct formulation for this branch is therefore: the optical fiber upcycle is real (double-verified), but the Corning contract is not its cause and does not constitute an order pass-through to A-shares; the only tradable point today is that "global optical fiber β has been confirmed once more," and among the names this issue verified, the only one with sufficient purity is 长飞光纤 (85.3%).

Calendar catalyst: the 27th China International Optoelectronic Exposition (CIOE) runs today (9/9) through 9/11 at the Shenzhen World Exhibition & Convention Center, held alongside the IICIE integrated circuit innovation expo and elexcon Shenzhen International Electronics Show, covering 320,000 square meters with over 5,000 exhibitors. ⚠️ This is a calendar event that news scanning cannot see — it is not new news and will not appear in any newswire, yet it is a genuine event catalyst for today.

5.2.1 ⚠️ 长飞光纤's valuation: East Money's "total market cap RMB 345.1 billion" is wrong

长飞 is dual-listed as A+H, with 421,476,183 H shares making up 50.91% of the 827,905,108 total shares. East Money's "total market cap" is computed as A-share price × total shares, which systematically overstates it.

Blended market cap = 406,428,925 A shares × RMB 416.85 (RMB 169.4 billion) + 421,476,183 H shares × HK$171.20 × 0.86482 (RMB 62.4 billion); the FX rate is the 2026-09-08 PBOC central parity of HK$100 = RMB 86.482.

⚠️ Two risks surface from this, which this issue believes the market is underpricing:

1. A/H premium +181.5% — the H share converts to RMB 148.06 versus RMB 416.85 for the A share, so for the same earnings, A-share investors pay 2.8 times what H-share investors pay. And on 9/8 itself the A share was +4.08% while the H share was −2.28% — the two venues have already begun to diverge.
2. PB is at the 91.3% percentile of the last 3 years. ⚠️ The basis must be made explicit: that percentile is taken from East Money's A-share-basis historical series (PB 21.09x, 3-year median 2.2, about 9.6x the median), and the series is internally consistent; if the true A+H blended basis in the table above is used instead, PB is 14.2x, about 6.5x that median. The two rulers give very different absolute values, but the direction is the same — PB is in a historically high zone. The PE percentile is only 66.9% because the denominator (E) is growing explosively and pushing PE down — this is exactly the "low PE percentile + high PB percentile" combination that typically appears at a cycle top rather than a cycle start.

⚠️ This issue does not accept the claim that "all four have PB percentiles of 83%–91%, so the whole sector has been systematically repriced." That conclusion gives neither each company's percentile value nor the source series, making it an unsourced assertion; and 长飞 alone has the two bases described above, a different ruler from the other three (pure A-share names). In the §5.2 cross-comparison table, 长飞's PB of 21.09 is on the East Money basis and is not the same ruler as 中天's 2.89, 亨通's 4.62 or 中际旭创's 26.67; readers making comparisons should use this section's blended basis of 14.2x.

3. A 60.52% single-quarter gross margin is a cycle position, not a steady state. The company itself writes in its risk section that "if artificial intelligence develops below expectations… its pull on optical fiber and cable demand may decline."

✅ 长飞's risk-item verification came back clean: no share pledges, no freezes, all shares are freely tradable with no lock-up expiry pressure, no material litigation, no inquiry letters, no penalties; goodwill of RMB 952 million is 5.8% of equity attributable to parent and is not impaired; the share-reduction plan of shareholder Changjiang Communications has expired, with only 999,985 shares actually sold (0.12% of total shares). ⚠️ But there have been 4 unusual share price movement announcements within 3 months (06-05, 06-24, 07-31, 08-07).

Hollow-core fiber (first-hand evidence, 2026 interim report text): "cumulative delivery of more than 10,000 fiber-km," minimum attenuation 0.04dB/km, longest draw 91.2 km, "deployed in more than 13 commercial and pilot projects in China and multiple overseas locations, making it the world's most widely deployed commercial hollow-core fiber supplier." ⚠️ The discounts that must be applied: "cumulative" is a cross-period measure, not the current period; "commercial and pilot" is a blended measure that is not broken out; there is no customer name, order amount or separately recognized revenue whatsoever; at the component level (Changxin Bochuang) it is explicitly still at the "sampling" stage. There is no reliable data on hollow-core fiber's contribution to that RMB 2.449 billion of 2026H1 ex-non-recurring profit, and this round of results cannot be attributed to it.

5.3 Branches 3–7: the climate theme — the catalyst is real, and the transmission is denied by the company itself

Catalyst closure: the WMO issued an alert on 2026-09-03 confirming that El Niño has formed and will continue to strengthen, with an 81% probability of reaching super strength in October–December and a near-100% probability of persisting through end-February 2027; the UN FAO assesses that global grain output in 2026/2027 faces downward pressure. Today is the 5th trading day of this catalyst.

⚠️ Commodity-side confirmation strength must be given product by product, not stated wholesale as "confirmed" or "not confirmed":

US agrochemical control group (tradable proxies, 9/8 close): Mosaic +2.63%, Nutrien +1.52%, CF Industries +0.73%, Bunge +4.36%; while ADM −0.13% and Deere −1.85%. ⚠️ The accurate description is "4 up, 2 down, with moderate gains," not a broad rally — farm machinery leader Deere actually fell 1.85%. However it is read, the magnitude is far smaller than the single-day +7.01% in A-share phosphate fertilizer and phosphorus chemicals.

⚠️ The hardest piece of falsifying evidence comes from an announcement by a company inside the theme itself. From the text of 万向德农 (600371)'s 2026-09-08 17:46 unusual-movement announcement:

"Recently the market has discussed climate topics such as a 'super El Niño' extensively; the relevant topic mainly concerns price fluctuations in agricultural products such as commodity corn, but the company's main business is the R&D, production and sale of corn seed, which sits upstream in the agricultural industry chain and belongs to a different link from the commodity corn market. At present the domestic corn seed industry as a whole is in a destocking cycle, with the market showing supply exceeding demand; intensified competition has pressured corn seed selling prices, and the gross margin of the company's corn seed business has also declined."

The same announcement further self-discloses: a cumulative gain of 122.64% from 8/17–9/8, with 9 limit-up days and 4 unusual-movement triggers in that span; cumulative turnover rate of 165.19% from 9/1–9/8, averaging 27.53% per day; 2026 interim revenue of RMB 99.4749 million, −15.14% year on year, and net profit attributable to parent of RMB 13.8016 million, −44.47% year on year; as of the 9/8 close, dynamic PE 167.79x, static PE 835.57x, PB 8.91, and it writes directly that this "already severely deviates from fundamentals and a reasonable valuation range."

⚠️ The weight of this passage lies in the fact that the attribution (it is El Niño) and the falsification (but it does not transmit to me) come from the same first-hand document. It answers both the previous issue's two-day unanswered question of "what the agriculture catalyst is" and this issue's mandatory question of "whether this catalyst can land on seed companies' income statements."

Layered conclusions:

• Soft commodities (white sugar): catalyst real, commodity confirmation real (SR at a window high) → keep grade A−;
• Fertilizer / phosphorus chemicals: catalyst real, commodity confirmation weak (urea +2.82% over 5 days vs the sector's +7.01% in one day) → cut to B, play only low positions, not high-rung boards;
• Seeds / crop farming: catalyst real, transmission denied in writing by the company, commodity (corn) essentially unmoved → cut to C, do not participate.

5.4 Branch 4: semiconductor equipment — US and A-share directions are completely opposite, and it may in fact be a negative

Overnight US market: Applied Materials +3.98%, Lam Research +4.15%, TSMC +2.35%, Broadcom +2.98%, SMH +1.19%, Intel +9.05% (128 million shares, plans to raise PC CPU prices by 10%); but Nvidia −2.01%, Micron −1.61%.

A-shares on 9/8: semiconductors −1.34% (32 up / 152 down), electronic chemicals −1.46% (weakest of the day), main-force net outflow RMB 1.389 billion; 北方华创 (002371, Shenzhen main board) −1.93%, 中微公司 (688012, STAR Market) −2.32%, 拓荆科技 (688072, STAR Market) −3.06%, 华海清科 (Hwatsing Technology, 688120, STAR Market) −1.71%.

⚠️ This issue does not read the US equipment rally as a positive for A-share semiconductors, for a directional reason: equipment makers' revenue speaks to future supply, not current demand. Applied Materials and Lam rising while Micron (the storage demand side) fell 1.61% has a blunt meaning: "future capacity is expanding while current prices are loosening" — neutral-to-positive for the domestic-substitution narrative in local equipment, negative for storage/chip prices. Against a backdrop where A-share semiconductors have had two consecutive days of net outflow and 32 up versus 152 down, using an overseas signal of dubious direction to go long against the trend is insufficiently supported. This branch is graded B+ and given no recommendation slot.

5.5 Branch 5: pharma / CRO — one of the few directions where institutional seats were one-way buyers

• 奥浦迈 (688293, STAR Market ±20%) hit limit-up +19.9965% on 9/8, closing at RMB 68.29; institutional seats 3 buy 0 sell, net buy RMB 53.50 million, 10.05% of that day's turnover.
• Sector: medical R&D outsourcing +3.69% (27 up, 2 down) on 9/8, main-force net inflow RMB 1.793 billion (net share 7.29%); medical services +2.85% with net inflow RMB 2.078 billion; pharma & biotech net inflow RMB 2.624 billion.
• Independent corroboration: in August the focus of A-share institutional research visits shifted from tech to large consumer, with pharma & biotech research interest jumping to first among all industries, while the electronics sector's share fell from the year's high of 17.59% in June to under 15% in August (Shanghai Securities News).
• ⚠️ The position risk is explicit: 奥浦迈 is +91.13% over the last 60 days and +27.50% over the last 5 days, closing at its 60-day high (100th percentile).

⚠️ The branch logic of this section holds, but its representative stock has been withdrawn after fundamental verification. See §6.1: within 奥浦迈's +230.34% growth in net profit attributable to parent, roughly 60% comes from a fair-value gain on a financial liability that arose because the company's share price fell (RMB 60.9858 million), while the 9/8 share price is already 63.8% above the 6/30 level, meaning that gain will reverse in the second half; ex-non-recurring net profit rose only +66.87%, and the ex-non-recurring PE is 239x; there were no operational announcements at all within the window.

This branch therefore keeps its B+ grade and its place among "today's three strongest branches," but §9.1 gives it no recommendation slot, and no stock that this issue has not independently verified (such as 康龙化成) is substituted in. This is a self-imposed constraint of this issue: a valid branch does not entitle you to grab any stock to represent it.

---

6. Master ranking of single-stock positive intensity

⚠️ The ranking variable of this list is not a single variable. The "primary ranking basis" column for each stock states explicitly which dimension got it onto the list, so readers can select according to their own framework. ⚠️ Scores are not return forecasts. The previous two issues have proven repeatedly: fundamental quality determines "how long you can hold," not "whether it rises today."

⚠️ The ranking follows strictly from the component sums of the §7A scoring model, and every row has been re-checked for addition. The order has nothing to do with "yesterday's gain" — most of the stocks that hit limit-up on 9/8 rank in the bottom half.

⚠️ This list deliberately excludes the high-rung limit-up boards of 9/8. 亚盛集团 (600108), 万向德农 (600371), 百大集团 (Baida Group, 600865) and 金正大 (002470) all go into the §7 Pass list, with reasons given there.

61 Fundamental verification — 奥浦迈688293⚠️ the verification result directly overturned its recommendation slot

The data in this section were pulled independently from East Money's three financial statements, Tonghuashun's financial summary, the 2026 interim report and announcement texts, and Baidu Stock valuation. The verification conclusion is the opposite of the surface signal "institutions 3 buy 0 sell," so this issue downgrades it from a recommendation slot to "watch only."

First, the genuine supports (this part is solid):

Metric	2026H1	Note
Cell culture product revenue	RMB 208 million, +34.00% year on year	45.85% of revenue, gross margin 67.81%; the media business is genuinely ramping
Client pipelines already at commercial stage	18 (plus 381 with pilot-scale processes fixed)	Once a culture medium is written into a commercial production process, switching requires redoing comparability studies, so switching costs are extremely high
Net operating cash flow	RMB 106 million = 2.14x ex-non-recurring net profit	Cash-to-revenue ratio 1.06, receivable turnover days 108.7→65.2
Cash / interest-bearing debt	RMB 979 million / effectively zero	Debt-to-asset ratio 16.02%

Table scrolls horizontally

⚠️ But three hard problems, each of which points to "this limit-up should not be bought on fundamentals":

① The profit growth was bought and computed, not earned. PharmaLegacy (澎立生物) was consolidated at 100% in January 2026. Nearly half of the +155.24% revenue growth comes from consolidation; net profit attributable to parent rose +230.34%, while ex-non-recurring net profit rose only +66.87% (RMB 49.3572 million). The RMB 74.6754 million difference has been reconciled item by item against the interim report's non-recurring gains and losses schedule.

② ⚠️⚠️ The most critical point: that RMB 60.9858 million fair-value gain arose because the share price was falling in the first half — and now the share price is rising.

From the interim report text: "the fair-value gain on financial liabilities formed by the acquisition increased profit by RMB 60.9858 million," arising from "the financial liability relating to the share consideration for the latter two tranches not yet issued in the purchase of 100% of PharmaLegacy's equity via share issuance and cash," and it is "affected by changes in the company's share price" and "does not involve current-period operating cash flow."

The value of that liability ≈ number of shares to be issued in future × share price, so when the price falls the liability shrinks and a gain is recorded; when the price rises the liability grows and a loss is recorded. Fact check: the close was RMB 49.00 on 2025-12-31 → RMB 41.69 on 2026-06-30 (−14.9%), which is the source of the H1 paper gain. And the 2026-09-08 close of RMB 68.29 is already +63.8% above the 6/30 level.

This means that the more the share price rises, the worse second-half reported profit becomes — a reflexive structure that the market is clearly underpricing.

Cross-corroboration (this is hard evidence, not an estimate): the company disclosed diluted EPS of only RMB 0.45, half of basic EPS of RMB 0.91. Accounting standards require reversing this paper gain in the diluted calculation — which means the company's own statements already concede that this RMB 60.9858 million is not sustainable earnings.

⚠️ The implied scale of the H2 reversal accrual is roughly RMB 150–220 million, but that estimate depends on unknown inputs such as discounting, cap/floor clauses and the measurement base date, so this issue does not cite a specific amount — only the direction and the magnitude judgment of "large enough to swallow the full-year profit of the core business."

③ The valuation is a textbook "low PE percentile + high PB percentile" divergence. PE (TTM) of 75.75x sits at the 22.8% percentile since listing (seemingly cheap), while PB of 3.09x is at the 72.7% percentile; excluding non-recurring items, PE reaches 239x. The PE percentile is low because the denominator has been lifted by that paper gain — this is exactly the textbook form of the "low PE trap."

④ No operational announcements within the window. Day-by-day scan of 9/1–9/9: 9/1 investor relations record, 9/2 board resolutions and routine equity incentive vesting, 9/3 announcement of shareholder reduction results, 9/8 business registration change at a wholly owned subsidiary (change of business scope and reissued license, filed 9/7 18:29). No event driver can be found at the company level for the 9/8 limit-up, which should be attributed to sector fund flows.

⑤ ⚠️ One basis cited in §4 of this issue must be corrected. After seat-by-seat verification of the Dragon-Tiger List:

• "The 3 institution-dedicated seats bought a combined RMB 53.5029 million and there were no institution-dedicated seats on the sell list" is entirely true, so the point that institutions bought one-way holds;
• but the whole-list net buy was only RMB 24.4814 million, or 4.60% of total turnover, not 10.05%. The reason is that the Shanghai Connect dedicated seat appeared on both the buy and sell sides (bought RMB 48.2654 million / sold RMB 40.9170 million, net inflow only RMB 7.3485 million), and the sell list contained Guotai Haitong HQ at −RMB 17.57 million, Goldman Sachs (China) Shanghai at −RMB 12.38 million, CICC Shanghai at −RMB 11.34 million, and CITIC Shandong Qingdao at −RMB 10.96 million.
• Implication: the point that the institutional seat direction is clean holds, but the overall absorption intensity is far below what 10.05% implies. Both numbers are correct on different bases, and it is inappropriate to cite only the higher one.

⑥ Risk timetable: on 2026-10-14, 9.2360 million shares = 6.50% of total market cap come off lock-up (about 5 weeks away, private-placement institutional allotment shares); on 2027-02-12 a further 10.86% unlocks. Pre-IPO shareholder Shanghai Panxin already sold 2.3212 million shares between 6/2 and 9/1 for RMB 122 million, with an upper reduction price of RMB 67.79 that almost coincides with the 9/8 close of RMB 68.29, and its stake has fallen to 4.26% (below 5%), which materially lightens its pre-disclosure obligations for further sales and constitutes low-transparency selling pressure. There is also acquisition goodwill of RMB 584 million = 18.6% of net assets attributable to parent.

⑦ The label "domestic culture media leader" cannot be confirmed. Nowhere in the interim report does the company describe itself as number one in market share or as a leader; on the contrary it writes that in "the domestic cell culture media market… imported manufacturers have long held a relatively high market share," and positions itself as "narrowing the gap." On this basis this issue does not use the word "leader."

Conclusion of this section: the culture media core business is real and the financial health is real, but the combination of "institutional buying + net profit +230%" is misleading — 49.2% (RMB 60.9858 million / RMB 124.0326 million) of net profit attributable to parent is an accounting paper gain, and the condition that produced that gain (a falling share price) has already reversed. 奥浦迈 is downgraded from a recommendation slot to "watch only," ranking 14th in the §6 master list.

---

7. Pass list

---

7A. Single-stock scoring model (total 100)

Component definitions: source authority 0–15 / positive directness 0–20 / earnings elasticity 0–15 / industry position and fundamentals 0–15 / expectation gap 0–10 / theme durability 0–10 / A-share trading attributes 0–10 / risk deduction 0 to −15.

⚠️ Two basis notes, please read first: ① The seven positive components sum to a maximum of 95 pts, not 100 (15+20+15+15+10+10+10). "Total 100" is the model's nominal scale; the actual maximum attainable is 95, before the risk deduction. This table is not normalized. ② All components take values within their own domain and are never negative; negative factors such as an overly high position or chasing highs go uniformly into the "risk deduction" column rather than as negative values in positive components. ③ Every row below has been re-checked component by component for addition.

⚠️ The "positive directness" column is the dimension that moved most in this issue, and the one readers should watch most closely. 中天's and 亨通's totals fell from the first draft's 76/66 to 40/38, almost entirely from this column — not because their fundamentals got worse (the industry position column still gives them 10 pts), but because this issue found that tonight's positive very likely does not transmit to them. ⚠️ But see the basis warning in §5.2: those shares are 2025FY data, and 2026H1 segments are undisclosed, so the judgment itself is pending verification.

---

7B. Intra-branch ranking

7B.1 Optical fiber branch

Hardest single stock: 长飞光纤 (the only name in this issue's sample with both high purity and hard financials). Highest trading recognizability: 长飞光纤. Who follows: 烽火 (real segment data, but dragged down by the overall statements). Who to Pass: 中天, 亨通 (the logic does not hold).

7B.2 Tanker shipping / oil & gas branch

Hardest single stock: 中远海能 (highest tanker transport purity, directly matching the tanker attack event). Highest trading recognizability: 招商轮船. Who follows: 中曼石油 and 和顺石油 (Heshun Petroleum) (driven by the oil price rather than freight rates).

7B.3 Agriculture / climate branch (ranked by commodity confirmation, not by price gain)

⚠️ This branch is deliberately ranked in the opposite order to the gainers list. The three strongest performers on 9/8 (金正大 at limit-up, 亚盛集团 at its 4th limit-up board, 万向德农 +5.04% (no limit-up)) all rank last, while 云天化 and 新洋丰, which did not hit limit-up, rank at the top — the ranking variables are "commodity confirmation + ex-non-recurring quality," not "how much it rose yesterday." This simultaneously corrects the previous issue's two opposite errors of "downgrading the whole agriculture line" and "chasing the high-rung boards."

---

8. Opening verification signals for today

⚠️ Lesson from the previous issue: the accuracy rate of the verification-point system is far higher than that of the recommendation conclusions themselves (5 of 7 judged correctly). This section is therefore placed before the recommendation conclusions, and every item must be "judgeable intraday and falsifiable." ⚠️ An even more important lesson: a verification point must test the variable that loses money. Each item below is labeled with "what risk it tests."

8.1 Branch-level verification points

8.2 Auction and single-stock signals

• Auction: if 招商轮船 (601872) and 中远海能 (600026) gap up more than 3% without a matching expansion in auction volume, that is the classic event-driven one-day-wonder pattern; a gap-up of 1%–2% with expanding volume is in fact the healthier pattern.
• ⚠️ Reverse signal (the previous issue got this right once): if 长飞光纤 (601869) opens lower or gaps up by less than 0.5%, the optical fiber positive has not been accepted — the previous issue used the same criterion to correctly reject optical modules at 09:25. ⚠️ Note: 中天科技/亨通光电 are deliberately no longer used as criteria here; item 3 of §8.1 explains why (their optical communication exposure is too low, so their moves have nothing to do with the optical fiber upcycle), and this issue does not resurrect a voided criterion here.
• Sector signals: if tanker shipping/oil services can produce 3 or more fast limit-ups before 10:00, judge the branch to be established; if only names that already hit limit-up yesterday, such as 中曼石油 (603619) and 和顺石油 (Heshun Petroleum, 603353), seal alone, judge it relay rather than diffusion.
• Risk signals: ① if yesterday's 73 limit-up stocks broadly gap up and fade today, sentiment is weakening; ② the failed-limit-up rate has already risen from 17.0% to 33.6%, so relay difficulty is at a high level; ③ two new stocks list today (洛轴股份 301699 on ChiNext, 百迈科 920268 on the Beijing Stock Exchange) — ⚠️ first-day turnover in new listings pollutes the fund-flow readings of their sector, so it must be stripped out when reading machinery/bearings sector flows today.

---

9. Final recommendation conclusions

9.1 Stocks most worth watching today (3 names + 2 slots left empty)

⚠️ This section enforces the previous issue's remediation requirement: "limit the number of recommendation slots driven by any single variable to no more than 3." The final result is 3 names across 2 mutually unrelated drivers: geopolitics 2 slots (招商轮船, 中远海能), global optical fiber β 1 slot (长飞光纤); climate theme 0 slots, institutional rebalancing 0 slots. Slots 4 and 5 are publicly left empty, for the reasons given below the table.

⚠️ Why slots 4 and 5 are left empty — this is the action readers most need to notice in this issue.

The previous issue's recap set a remediation requirement for this slot of "limit the number of recommendation slots driven by any single variable to no more than 3," while an earlier issue wrote that "forcing in a stock with no catalyst to fake diversification is more dangerous than admitting concentration." In executing this, fundamental verification vetoed three of the originally selected names in a row:

The remaining candidates (招商南油, 中曼石油, 烽火通信) all fall inside the two drivers that are already full, so filling them in would only raise concentration while disguising it as diversification.

This issue therefore gives only 3 recommendation slots and publicly leaves slots 4 and 5 empty. The empty slots are themselves information — they say that after verification, only two genuinely defensible opportunities exist today, and that thinning a position into five parts is not the same as being diversified. Identifying risk is not the same as constraining it; this section is the constraint.

⚠️ One piece of self-criticism that has to be said: the first draft placed 中天 and 亨通 in slots 2 and 4 on the grounds of "lowest position, lowest PB." After verification, the reason their positions are low is precisely that the copper and grid businesses making up 70%–80% of their revenue are weakening, and optical fiber at 8%–14% cannot rescue the income statement. Using "it fell a lot" as a bullish reason without first asking "which business fell" is the biggest error this issue nearly committed.

9.2 Today's three strongest branches

9.3 Directions not recommended for chasing today

1. ⚠️ Seeds and high-rung one-word limit boards (万向德农 600371, 亚盛集团 600108, 百大集团 600865). Three independent reasons: ① 万向德农 itself announced a static PE of 835.57x, a gain of 122.64%, "severe deviation from fundamentals," and denied any transmission between El Niño and its own core business; ② 亚盛集团's institutional seats were 0 buy 2 sell, with a volume ratio of 0.09 and turnover of RMB 89 million, so the seal quality is an illusion produced by evaporated liquidity; ③ 百大集团 had turnover of RMB 59 million and its catalyst has gone unidentified for three consecutive days.

2. ⚠️ Chasing optical modules (中际旭创 300308, 新易盛 300502, 天孚通信 300394). The FCC's draft ban on imports of Chinese-made new-model optical modules is intended to take effect within the year, and the regulatory risk points precisely at this sub-branch rather than at optical fiber; and on 9/8 optical modules had zero institutional seats on the list. ⚠️ But the other side must be stated: 中际旭创 had main-force net inflow of RMB 1.349 billion on 9/8, second highest of the day, so money has not left; this item is "do not chase highs," not "declared dead."

3. PCB / copper-clad laminate across the board. On 9/8 the sector had a main-force net outflow of RMB 1.304 billion while the price was +0.20%, a distribution pattern; institutional participation is down to 中京电子 alone, and within its institutional seats it is 2 net buy 1 net sell with a net amount of only 0.94% of turnover.

4. The "sell the news" game in foldable phones / the Apple chain. Apple's autumn launch runs from after today's close to 01:00 tomorrow morning, so the whole of today is a pure expectation game while the outcome is known before tomorrow's open — the timing risk of "sell on delivery" is explicit.

5. Bottom-fishing semiconductors and electronic chemicals against the trend. Even though US equipment stocks rallied overnight, that signal speaks to future supply rather than current demand (Micron was −1.61% over the same period), while A-share semiconductors were 32 up / 152 down on 9/8 with a main-force net outflow of RMB 1.389 billion.

6. ⚠️ 上海电影 (601595) — listed separately, because it is the biggest unresolved contradiction in this issue. It simultaneously has "the largest institutional net buy of the day (RMB 193 million, 2 buy 0 sell, verified seat by seat)" and "the controlling shareholder denying every material matter in writing." ⚠️ One additional piece of counter-evidence found in seat-by-seat verification: on the same day's sell list, Shanghai Connect net sold RMB 14.57 million, Goldman Sachs (China) Shanghai net sold RMB 10.15 million, and Guotai Haitong HQ net sold RMB 4.05 million — institution-dedicated seats were buying one-way while foreign and large brokerage seats were net selling. ⚠️ But the magnitudes on the two sides are severely unequal: institution-dedicated seats bought RMB 193 million one-way, while the above counter-direction seats net sold about RMB 28.77 million in total, less than 15% of the former. The accurate formulation is therefore "institutions are buying heavily while some foreign and large brokerage seats distribute modestly," not two pieces of equally weighted contradictory evidence. The real reason this issue does not participate is: the controlling shareholder has denied every material matter item by item, so the institutional buying has no fundamental landing point.

7. ⚠️ 中天科技 (600522) and 亨通光电 (600487) — not recommended for participation under the "optical fiber" label. Optical communication is only 14.0% and 8.4% of their revenue, with main businesses in submarine cable, power grid and copper conductors. If you want to buy them, it should be on submarine cable or grid logic with separate research, not on tonight's Corning news. This issue's first draft listed them in recommendation slots 2 and 5 and withdrew them after verification — this item is listed here so that the correction is visible to readers.

9.4 Final one-sentence judgment

The hardest information generated inside the window today and not yet priced by A-shares is the geopolitical supply shock (US forces destroying 5 Iranian tankers on 9/8, Iran threatening tankers at Kuwaiti and Bahraini ports), and it supports the first two recommendation slots; the appeal of the second line, optical fiber, does not lie in tonight's Corning news — that contract is a bilateral procurement deal between Corning and Verizon in which A-share companies play no part, and the direction of its marginal impact is unclear — but in the industry-wide price increase verified by two mutually independent interim reports from 长飞 and 烽火, though that evidence was already public on 21 August and today is merely lagging confirmation; while yesterday's hottest theme, agriculture, finally had its catalyst identified in this issue (the WMO's 9/3 El Niño alert), only to have the transmission path denied on the very same day by an announcement from a company inside the theme.

This issue therefore gives only 3 recommendation slots and publicly leaves 2 empty: fundamental verification vetoed three originally selected names in a row — optical communication is only 14.0% and 8.4% of 中天科技's and 亨通光电's revenue, so they are simply not on this chain; and 60% of 奥浦迈's +230% net profit is an accounting paper gain created by a falling share price that is now reversing. The empty slots are not an omission; they are the most honest conclusion this issue can offer: only two genuinely defensible opportunities exist today.

---

---

10. ⚠️ Known limitations of this issue (please read alongside the conclusions)

So readers can discount each conclusion appropriately, the verifications that could not be completed and the known defects are listed honestly below:

---

⚠️ Risk disclosure

1. This list is only a compilation and observation of publicly available pre-market information; it does not constitute investment advice, a securities recommendation, or an offer to buy or sell in any form. The labels "priority deep-dive / watch closely / watch only / Pass" used in the text are this issue's internal information-strength grades, not investment ratings, and carry no implication about position size, entry/exit timing or target price.
2. Data reading timestamps: A-share quotes are at the 2026-09-08 close; US equities and international commodities are at the 2026-09-08 close or the early hours of 09-09, as labeled in each place; the Dragon-Tiger List is the complete 2026-09-08 dataset. This issue was generated between 07:00 and 08:00 on 2026-09-09, and all prices and flow data will change after the open.
3. Incomplete-coverage statement: this issue verified only the 15 stocks and 8 branches listed in §7A; the vast majority of the more than 5,000 listed names across the A-share market were not scanned. Not being mentioned does not mean there is no opportunity, and being mentioned does not mean it has been sufficiently researched. Section 10 lists 12 known limitations.
4. Stock-specific risks: the stocks in this text broadly carry risks of results missing expectations, shareholder selling, lock-up expiries, regulatory inquiries, abnormal share price movements and insufficient liquidity; the STAR Market (±20%) and Beijing Stock Exchange (±30%) additionally impose investor suitability thresholds such as RMB 500,000 in assets, and some readers are not eligible to trade them. Price limits are tiered by board — please do not read ±20% / ±30% names with a ±10% intuition.
5. This report is generated by an automated process, and may contain deviations caused by information lag, industry-chain mapping errors or data interface failures; it cannot be used directly as a basis for trading. Exchange and company official disclosures prevail.
6. Generator holdings statement: the generator of this report holds none of the names described herein and has no interest relationship with the companies described.