A-Share Pre-Market Brief | 2026-09-11 Friday

Scope and caliber of this edition (please judge conclusion strength accordingly) ① News window = 2026-09-10 15:00 (previous session close) → 2026-09-11 07:15 (Beijing time). All announcements are filtered by East Money's eiTime (actual publication moment), not notice_date. A total of 1,647 A-share announcements fell inside the window and each was screened one by one. ② Quotes, sector and single-stock fund flows are as of the 2026-09-10 close; limit-up / failed-limit-up / limit-down counts come from East Money's official pools (9/10: 35 limit-ups, 22 failed limit-ups, 11 limit-downs). ③ The Dragon-Tiger List is for 9/10 and was published after 18:00 last night; this edition has retrieved it and done seat-level drill-down (the parts left blank in yesterday's recap because it was generated at 15:40 have been filled in this morning). ⚠️ The same stock can appear on multiple rows under different "listing reasons"; this edition de-duplicates by seat before aggregating and does not sum overlapping rows. ④ Price limits differ by board: Shanghai/Shenzhen main board ±10%, ChiNext/STAR Market ±20%, Beijing Stock Exchange ±30%. Every stock in this report is tagged with its board. Do not read a ±20% / ±30% name with ±10% intuition. ⑤ Overnight global markets and commodities are read at 2026-09-11 06:00–07:15, U.S. equities as of the 9/10 close, commodities from the night/electronic session. ⚠️ The absolute level of international crude is contradicted across multiple text sources, so this edition uses the domestic SC crude futures night session as its primary caliber — see the note in §1. ⑥ ⚠️ This edition contains one S-grade negative (the fire at China State Shipbuilding's Beihai shipyard). It is the only S-grade event in the window, and its direction is down. Do not misread "strongest catalyst" as "strongest long opportunity."

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

I. Channels that worked

• East Money announcements np-anotice-stock, pulled across 22 pages for 2,200 records; 1,647 inside the window, on the eiTime caliber.
• push2delay.eastmoney.com/api/qt/clist/get returned today / 5-day / 10-day fund flows across three lookbacks for 496 industry and concept sectors, succeeding on the first try (the akshare East Money cluster RemoteDisconnected logged in yesterday's recap did not recur this morning, but we still went direct).
• stock_zt_pool_em returned the limit-up pools for 9/8, 9/9 and 9/10; stock_zt_pool_previous_em succeeded.
• stock_lhb_detail_em(20260910) returned 64 rows. The very same call that threw TypeError at 15:40 yesterday worked normally this morning — further proof that it was "not yet published," not an API failure (fourth hit on the memory lhb-publishes-after-1800).
• stock_lhb_stock_detail_em retrieved buy/sell seat detail name by name.
• The CNBC quote API supplied U.S. equities and commodity futures.

II. Channels that failed / had to be worked around

• ⚠️ WebFetch cannot parse East Money announcement PDFs — all three calls against pdf.dfcfw.com returned "binary, unreadable." Switched to curl download + local pypdf extraction; all succeeded.
• ⚠️⚠️ One silent failure that nearly made it into the body text: a WebFetch on Fujian Mindong Electric Power's abnormal-movement announcement returned a structurally complete, professionally worded "announcement summary" — but the PDF was actually 0 bytes. The model had confabulated boilerplate about "the company explained three aspects" from an empty file. The whole passage was discarded and replaced with the real text extracted locally. This is the same shape as the memory edgar-cik-leading-zero-empty-file (HTTP 200 + 0 bytes = the most dangerous silent failure), except this time the LLM completed on empty input, which is more dangerous than an empty return because it looks like success.
• pdftotext exists on this machine but produced empty output for this batch of PDFs; pypdf worked.

III. Five cross-checks deliberately performed on method — and they changed the conclusions

1. ⚠️⚠️ The relationship between Beihai Shipbuilding and China State Shipbuilding is the opposite of what search returned. A Jiemian News piece explicitly states that CSSC Qingdao Beihai Shipbuilding Co. is not a subsidiary and belongs to sister listed company China Shipbuilding Industry Corp — if taken at face value, the whole conclusion would become "unrelated to 600150." But that article was published in December 2024, and China Shipbuilding Industry Corp was absorbed into China State Shipbuilding via share swap in August 2025 and delisted. We finally went with the company's own announcement text at 00:30: it states in writing that the Beihai yard is a "controlled subsidiary." The memory listing-status-must-be-queried reproduced verbatim: status-type facts must be checked against a current primary source; a search summary will hand you an internally consistent wrong answer.
2. ⚠️⚠️ "Iran attacked 10 ships in a single strike" happened on 9/9 and is outside the window. The first search screen presented it as the latest development; adopting it directly would have treated an event already priced in the previous session as a fresh catalyst this morning — and 9/9 was precisely the day three tanker-shipping names hit limit-up. Memories scan-window-misses-multiday-themes + recurring-column-headlines-hide-date.
3. ⚠️⚠️ No credible consensus is available for the absolute level of international crude. TradingEconomics gives "Brent 101.25, +0.04%" (nearly identical to the 101.29 cited in yesterday's recap, strongly suggesting a silent lag), while CNBC futures @BZ.1 gives 108.99 (prior close 107.63) and @CL.1 gives 103.83. The two cannot be reconciled, so the body text does not quote an absolute international crude price and instead uses domestic SC2610 night session +6.19% (open interest 36,372 lots, volume 163,914 lots; liveness verified) as the primary caliber. Memories etf-proxy-detects-stale-spot-quotes / two-pct-figures-may-differ-by-base.
4. ⚠️ Concept-level and industry-level 10-day fund flows disagree: industry "printed circuit board" 10-day +6.527 billion, concept "PCB" 10-day −0.725 billion; industry "power" 10-day +0.869 billion, concept "green power" 10-day −1.485 billion. The constituent sets differ and neither is wrong; the body text presents both calibers side by side rather than picking the flattering one.
5. ⚠️ Overlapping Dragon-Tiger rows have been de-duplicated: Yu San Xia A appeared on 2 rows, Dunhuang Seed on 3. Summing rows would systematically overstate net buying (memory lhb-seat-lists-overlap-double-count).
6. ⚠️⚠️ A sub-agent check overturned the draft's rationale for branch #2 — the single most valuable revision in this edition. The draft rated power A− and gave Hunan Development a recommendation slot on the grounds that "PE 25.48, the only normally valued name in the branch." After pulling the interim-report originals, fundamentals-analyst found: ① East Money's PE field uses four different algorithms across these four names — 25.48/98.24/137.56/86.81 are all unusable (Huayin's true value is negative); ② the "RMB 5 trillion grid" spend goes to transmission and distribution equipment and offers zero elasticity to power generators; ③ the four companies' realized tariffs in 2026H1 were +0.8% / −1.0% / −7.0% / unavailable — not one of them delivered a "tariff increase." Acting on this: branch A− → B+, Huayin Electric Power watch-only → Pass, Mindong Electric Power's fundamentals score cut, and Hunan Development's recommendation rationale rewritten from scratch (now "the only name with both revenue and profit growth + a net-profit-to-cash ratio of 1.99"). This is exactly the "apply the same yardstick to the branch you most want to recommend" that yesterday's recap demanded — and this time it did overturn our own conclusion.
7. ⚠️ THS and East Money fund calibers must not be mixed: THS real-time net shows Hunan Development at −3.4292 million and Sinergy Clean Energy at −141 million (net outflow on a limit-up day), while East Money's main-force net inflow caliber is positive. The two use different statistical methods and cannot be subtracted from each other; this edition uses East Money as the primary caliber and flags the discrepancy in the body.
8. ⚠️⚠️ A second sub-agent check overturned our own avoidance rationale from yesterday (Wuxi Taiji Industry). Both yesterday's recap and this edition's draft said "overnight memory selloff → avoid Taiji Industry." After downloading and parsing 14 primary CNINFO announcements, fundamentals-analyst proved that causal chain simply does not hold: ① on 9/10 the A-share memory chain fell broadly (Longsys −0.30%, GigaDevice −0.32%, Unigroup Guoxin −1.16%), and Taiji Industry was the only name to limit up against the tide — the opposite direction from the memory theme; ② in its 6/26 and 8/14 abnormal-movement announcements the company states that Haitai uses "full cost + 10% of total investment" cost-plus pricing, is "little affected by industry upcycles," and "does not involve HBM products"; ③ the contract text (Lin 2025-035) contains no clause linked to DRAM prices; ④ empirical numbers: across 2024→2025, when DRAM prices swung violently, Haitai's net profit went 237.11 million → 239.06 million, a change of just +0.8%. §5.5 has been rewritten and the avoidance rationale replaced with "attribution unknown + PE at the 92.5th percentile of ten years + ex-non-recurring −31.01%"; the test name for verification point 6 in §8 has also been switched from Taiji Industry to genuine memory stocks — using Taiji to test memory would give a signal pointing the wrong way.
9. ⚠️⚠️ The PE field went wrong a second time, this time 66.22 vs 95.73. The quote software shows 66.22 for Taiji Industry, while the sub-agent's own calculation of TTM = 448.1 − 327.1 + 315.5 = RMB 436.5 million gives 95.73×, matching Tencent's and Baidu's third-party calibers exactly. Adding the four power names, this edition has found caliber problems in the same field across 5 stocks — this is no longer incidental, and we suggest making "valuation must be computed yourself" a standing discipline.
10. ⚠️ THS last.js silently omitted the 9/10 candle (the last bar stopped at 9/9); the sub-agent only obtained it after switching to the Tencent quote channel. Another instance of "no error, just stale data" (memory stale-source-silent-lag).
11. ⚠️⚠️ The price-hike evidence chain for branch #1 was revised upward — the only time this edition strengthened a conclusion. The draft carried over yesterday's recap's caliber, saying the most recent hike was Fu Chiao on 7/1, and on that basis rated logic hardness "medium-high, already running 12 months." The check found two more recent hikes had been missed: Kingboard Laminates' seventh price-hike letter of the year on 8/28 (FR-4 +10%, thin cloth below 7628 +20%) and Panasonic Electric Works' 8/6 notice effective 9/1 (up to +30%). ⚠️ This lead was not found proactively — the Taiji Industry sub-agent surfaced it while cross-checking 9/10 media closing commentary — showing that "no new news in the window" had been silently turned by me into "no need to check industry developments outside the window," which is precisely the scan-window-misses-multiday-themes failure yesterday's recap criticized. Branch logic hardness has accordingly been raised from "medium-high" to "high," and §5.1 now tabulates the timing and magnitude of all four price hikes.

IV. What this edition did not obtain / could not establish (listed honestly)

• Three-lookback fund flow for the shipping & ports sector was not obtained (that sector name did not appear in the 496 industry sectors returned this run), so §2's fund-flow judgement on the tanker branch can only be substituted with the "oil & gas resources" concept and single-stock data, and its strength is downgraded accordingly.
• Current readings of VLCC spot rates (BDTI/TCE) were not obtained — that is the true earnings driver of the tanker branch; this edition has only crude price as a proxy, hence no recommendation slot for tanker shipping.
• The attribution of Wuxi Taiji Industry's (600667) 9/10 limit-up remains unresolved for a second consecutive session (see §6).
• No new price-hike letters dated 9/10–9/11 were found for this round of electronic cloth / copper-clad laminate, so it is still characterized as "an ongoing price-hike cycle, not a fresh catalyst this morning."
• Hunan Gold's (002155) shareholder meeting convenes today and the result will only come tonight; the verification point rolls forward.

---

0. Today in one sentence

1. The only S-grade event in the window is a negative: China State Shipbuilding (600150) announced at 00:30 today that a cargo-ship fire broke out at 11:15 on 9/10 at its controlled subsidiary CSSC Qingdao Beihai Shipbuilding, and that all 25 missing persons have been found, none with signs of life; the Ministry of Emergency Management has dispatched a working group, and both Xi Jinping and Li Qiang issued instructions. Yet the stock closed +3.26% on 9/10 with turnover of RMB 7.145 billion — because the scale of casualties was only reported at 17:30 and only carried by media at 19:06; during the session the market knew only that there was "a fire," not that "25 people died." This news has to be priced in all at once at today's open.
2. Overnight the U.S. memory chain sold off systematically: Philadelphia Semiconductor −2.66%, Micron −4.90%, SK Hynix ADR −5.20%, Lam Research −5.65%, SanDisk −4.06%. ⚠️ And the A-share memory chain was already falling yesterday (Longsys −0.30%, GigaDevice −0.32%, JCET −0.76%, Unigroup Guoxin −1.16%), with the memory-chip concept seeing 5-day fund flow of −8.602 billion and 10-day −42.100 billion — one of the most heavily bleeding directions in the whole market. The overnight external market is another kick to a chain that was already on the way down, not an inflection point.
3. The strongest long branch remains the upstream of the "electronic cloth → copper-clad laminate" price-hike chain, not because of new news but because among the branches positive on all three fund-flow lookbacks (today / 5-day / 10-day), its magnitude dwarfs the rest (components +3.017 / +20.160 / +5.369 billion; printed circuit board +3.017 / +18.283 / +6.527 billion; ⚠️ also positive on all three are utilities +2.031 / +2.897 / +1.248 and marine equipment +0.488 / +2.310 / +2.342, but their 5-day magnitude is only 1/7 to 1/9 of it), and the Dragon-Tiger List for Jin'an Guoji shows genuine institutions and Shenzhen-Connect northbound money buying, not pure hot money.
4. Power is the only industry whose limit-up count rose for two consecutive days (0→2→5), but ⚠️ this edition's fundamental check downgraded it from A− to B+: its only 2-board name, Fujian Mindong Electric Power, announced last night that "there are no material matters that should have been disclosed but were not," and the Dragon-Tiger List shows the local Ningde seat selling and Shenzhen Connect a net seller; more importantly, both public narratives supporting it fail — the "RMB 5 trillion grid investment" flows to transmission-and-distribution equipment manufacturing and offers zero elasticity to generators, while "September tariffs rising month-on-month" was not delivered in any of the four names' actual realized tariffs in 2026H1 (+0.8% / −1.0% / −7.0% / unavailable). The branch stands, but it is driven by fund flow and a valuation narrative, not by an earnings inflection.
5. Pre-market state: day 4 of the ebb phase + one downward S-grade event + one downward overnight external market. Today's first task is risk avoidance and lowering position expectations, not hunting strength. Driver types: event (negative) · overseas industry (negative) · price (crude, direction up but the mapping has failed twice).

---

1. News overview

Window: 2026-09-10 15:00 → 2026-09-11 07:15. Impact grade S = changes an industry trend or directly drives order elasticity; A = clearly positive/negative for one branch with multiple stocks affected; B = direction clear but the chain is long and needs verification; C = sentiment stimulus, weak persistence.

⚠️ The structural fact inside the window matters more than any single item: among 1,647 announcements there was not one "major price-hike letter," "large new order (>10% of revenue)" or "large earnings upgrade." This is consistent with yesterday's recap — this is a fund-flow / theme-rotation market, not an announcement-driven one. Accordingly this edition does not force a "positive" onto any limit-up stock; where nothing can be found, we say so (memory ai-attribution-stitches-old-news).

---

2. Strongest branches, descending

⚠️ This edition enforces the two hard disciplines set by yesterday's recap: Discipline A (gating up front): any branch with negative cumulative main-force net inflow on both the 5-day and 10-day lookbacks may not occupy a recommendation slot. The "three-lookback fund flow" column below is mandatory and may not be skipped. Discipline B (verification-point priority): structural verification points (consecutive limit-ups / sector leader / multi-day fund flow) outrank exogenous price ones (commodities / overseas markets). When they conflict, follow the structural one.

2.1 ⚠️ Why, with crude up +6.19%, this edition still gives tanker shipping no recommendation slot

This is the judgement this edition most needs to explain, because on the surface it is counter-intuitive.

Evidence supporting the long case (which must be listed honestly first):

• SC2610 night session +6.19%, a move significantly larger than 9/10's +3.71%;
• The "oil & gas resources" concept is positive on all three lookbacks (+0.438 / +1.608 / +2.248 billion);
• The Hormuz situation is indeed escalating: at least 75 vessels have been attacked and 21 crew killed so far, with 42 tankers waiting outside the strait;
• War-risk premiums have risen from 1–3% of hull value to 7.5–10%, and VLCC Middle East–China freight once hit a record US$423,736/day in March 2026.

But there are three counter-arguments, and they are structural (prioritized under Discipline B):

1. ⚠️⚠️ The same mapping has been rejected by the market two sessions in a row, and the second rejection came under more favorable conditions. On 9/9 three tanker names limited up (an event-driven pulse); on 9/10 crude rose another 3.71%, zero names limited up, and the whole group fell, with the shipping & ports industry limit-up count going 0 → 3 → 0. Yesterday's recap put it plainly: "the market did not pay a second time for the same piece of information." Today is the third test, and the first two tests agreed.
2. ⚠️ The driving event itself is outside the window. The most prominent search result — "Iran struck 10 ships at once, the largest attack since the war began" — occurred on 9/9, exactly the day the tankers limited up — it is an event already priced in the previous session, not incremental this morning. Inside this morning's window I could not find a single new, time-verifiable Hormuz escalation event.
3. ⚠️⚠️ I did not obtain this branch's real earnings driver. Tanker companies' profits are determined by VLCC spot rates (TCE), not by crude prices — crude is only a proxy, and that proxy has given the wrong signal two days running. This edition failed to obtain current BDTI/TCE readings, and granting a recommendation slot without the real driver amounts to betting on a proxy that has already been falsified twice.

Conclusion: tanker shipping is rated "watch only," with one structural verification point (see §8). The falsification test is explicit: if two or more tanker names hit limit-up or clearly go multi-board before 11:30 today, the third test has overturned the first two and this judgement is wrong — follow that signal, not this report's wording. ⚠️ The reverse risk must also be stated: if tankers really do take off today, this edition will have missed the day's strongest branch by "over-learning yesterday's lesson." That is a risk this edition explicitly accepts; the memory "withdrawal-confirmed is not a permanent label" is about exactly this.

2.2 ⚠️ Mandatory table: limit-up pool industry distribution across the last 3 sessions

This table is a mandatory item added by yesterday's recap (created because the power theme was missed). Rule: any industry whose limit-up count rises two days in a row enters the candidate pool, and an explicit include-or-exclude conclusion must be given.

⚠️ This table got two things right today: ① it confirms power is the only industry whose "pricing progress is advancing"; ② it visualizes the tanker pulse shape as 0→3→0 — looking only at crude +6.19%, today would very likely repeat yesterday's mistake.

2.3 ⚠️⚠️ Power branch downgraded from A− to B+: both public narratives fail on inspection

This section changed the most conclusions in this edition's quality review. After the sub-agent pulled the 2026 interim report text and financial data for each of the four power names, both pieces of logic the branch rested on were falsified.

① The "RMB 5 trillion 15th Five-Year Plan grid investment" has nothing to do with these generators — wrong link in the chain.

Verified: State Grid's "15th Five-Year Plan" fixed-asset investment is RMB 4 trillion (+40% vs the "14th Five-Year Plan") plus China Southern Power Grid's roughly RMB 1 trillion ≈ RMB 5 trillion, directed at UHV corridors, distribution network upgrades, grid digitalization and renewable interconnection/absorption. This is capex in the transmission-and-distribution segment, and the payees are UHV / transmission-and-transformation / distribution equipment manufacturers. For generators it adds not one kWh of electricity sold and not one fen to on-grid tariffs. The only indirect link is better absorption capacity → less curtailment of wind and solar (Mindong's interim report does note "curtailment at the Baicheng wind farm"), but that is distant and cannot be quantified into 2026–2027 statements.

Judgement: for these four names, the "RMB 5 trillion grid investment" is a 4/4 pure thematic mapping with zero direct earnings elasticity. ⚠️ If you want to trade this logic you should buy the equipment side — and the equipment side (grid equipment) has 5-day/10-day fund flows of −2.986 / −4.535 billion, both negative, and is already excluded by Discipline A. Neither end works.

② "September tariffs rising month-on-month" was not delivered in any of the four names' actual realized tariffs.

This is the most powerful falsification table — realized tariffs back-computed from the electricity volumes disclosed in each interim report:

⚠️ Not one of the four validated "rising tariffs" in 2026H1. Sinergy Clean Energy goes further and names falling tariffs as the main cause of its earnings collapse: "affected by market-based power trading and tariff policy, large-industrial power prices fell materially year on year this period, and settlement tariffs for the company's customers fell accordingly."

⚠️ One more pass-through trap must be spelled out: agency purchase price = on-grid tariff + T&D tariff + system operation fees + government funds. A rise in the agency purchase price may come entirely from higher "system operation fees" (industry reporting already flagged a nationwide surge in system operation fees in 2026Q1), and that money flows to ancillary services and capacity compensation — it is not a rise in generators' on-grid tariffs. ⚠️ Moreover this edition only verified the +6.5% for September in Guangdong's Pearl River Delta, and none of the four names are in Guangdong; the September energy-component breakdowns for Fujian, Hunan and Zhejiang were not obtained.

③ So what was the power sector actually rallying on for 9/10?

Public attributions point to the Document No. 136 implementation rules, 2026 annual retail-side contracting, excess-return sharing for retail power companies, and the valuation re-rating from the "utility-ization" of capacity tariffs — ⚠️ a valuation narrative, not an earnings upgrade, and not the same thread as "tariffs rising month-on-month + RMB 5 trillion of grid spend."

⚠️⚠️ The methodological point of this section: yesterday's recap criticized "the checklist gets silently skipped on the branch you most want to recommend." Today we applied the same yardstick to the power branch we wanted to recommend, and the result was a downgrade from A− to B+ plus the overturning of the original recommendation rationale. The branch is retained (because consecutive increases in limit-ups and positive flows on all three lookbacks are facts), but readers must know: what supports it is fund flow and a valuation narrative, not verifiable earnings improvement.

---

3. Overall single-stock positive-strength ranking (descending by stock)

⚠️ Important note: the whole market is on day 4 of an ebb phase and carries one S-grade negative, so absolute scores in this ranking are low across the board. No name with a total score below 60 is given "priority deep-dive" in this edition. Board limits: main board ±10% / ChiNext · STAR Market ±20% / Beijing Stock Exchange ±30%.

⚠️⚠️ A caliber warning about PE (the single most important data trap found during this edition's checks): The PE shown by quote software is wrong for several of these power names. This edition's sub-agent recomputed TTM = FY2025 + H1_2026 − H1_2025 and cross-validated against Baidu's full-history caliber, with the following results:

Stock	Shown by quote software (East Money)	Recomputed TTM (matches Baidu)	What East Money is actually computing
Fujian Mindong Electric Power (000993)	98.24	337.9	Using static FY2025 profit
Hunan Huayin Electric Power (600744)	137.56	−193.9 (negative)	Caliber unclear, even the sign is wrong
Sinergy Clean Energy (603105)	86.81	35.0	Annualizing H1
Hunan Development (000722)	25.48	No clean reading (restatement-contaminated)	Caliber unclear

Table scrolls horizontally

One field, four different algorithms across four stocks. The body text uses the recomputed values throughout, and stocks not independently re-verified (e.g. Leshan Electric Power) are explicitly tagged "East Money caliber, not re-verified." ⚠️ This trap affects two scoring items at once — "industry position and fundamentals" and "risk deductions" — because copying the software reading would turn Huayin Electric Power from "a chronically loss-making stock with negative PE" into "a PE-137 high-valuation growth stock," and those two carry completely different risk grades.

---

4. Single-stock scoring model (100 points)

⚠️ Four scoring disciplines specific to this edition (stated up front for tomorrow's reconciliation):

1. "No new news" is not neutral, it is a deduction. Neither branch 1 nor branch 2 has a fresh catalyst this morning, so both are held below 71 points, and no name scores above 80.
2. Negative on both 5-day and 10-day = theme persistence scores 0 and no recommendation slot. Today this discipline excluded five directions — brokers, memory, semicap, grid equipment and communication cable — including memory, which had the largest inflow yesterday.
3. A company announcement self-certifying "no undisclosed material matters" = expectation gap goes to zero. Mindong Electric Power was therefore cut from B+ to B and from recommendation candidate to watch only.
4. ⚠️ Being positive on all three fund-flow lookbacks cannot override a safety accident. Marine equipment is positive on all three (+0.488/+2.310/+2.342) and is still avoided outright — this discipline is new in this edition, because yesterday's framework had no slot for "material non-financial risk."
5. ⚠️⚠️ Valuation metrics must be self-computed or cross-validated; never take the quote software at face value. This is how this edition found that East Money's PE field uses four different algorithms across the four power names (§3 note), and it is why Huayin Electric Power was cut from "watch only" to "Pass," Mindong Electric Power's fundamentals score was pushed from neutral to negative, and Hunan Development's recommendation rationale was rewritten entirely. — This is the discipline that changed the most conclusions in this edition, and we suggest keeping it permanently.
6. ⚠️ A branch narrative must land on "which link collects the money." "RMB 5 trillion of grid investment" sounds bullish for power, but the payees are transmission-and-distribution equipment makers, with zero elasticity for generators (§2.3). The scoring question is not "is this policy good for this industry" but "who ultimately gets paid."

---

5. Detailed analysis of the top 10 stocks

51 金安国纪002636the highest score in this edition, but not the safest · Jin'an Guoji · Shenzhen main board ±10%

• Related news: ⚠️ no new news inside the window, but the price-hike evidence chain is more recent and harder than the draft assumed (this item was revised upward during checking):

  Date	Party	Content
  2026-08-28	Kingboard Laminates (global CCL leader)	⚠️ Seventh price-hike letter of the year: all-thickness FR-4 uniformly +10%; PP prepreg 7628 and above +10%, thin cloth below 7628 +20%
  Issued 2026-08-06 / effective 09-01	Panasonic Electric Works	Price adjustment for circuit-board materials, with some products up as much as 30%
  Issued 2026-06-30 / effective 07-01	Fu Chiao Industrial	Standard E-glass +30%, LowDK2 +15%
  2025-09 → 2026-08	7628 electronic cloth	RMB 3.8/m → about RMB 10/m, >165% YoY
  ⚠️ Two points that must be stated clearly: ① this edition did not obtain any new price-hike letter dated 9/10–9/11, so we cannot say "a price-hike letter came out today, hence the rally"; ② but the claim that "the most recent hike was in July" is wrong — the leader issued its seventh hike on 8/28, and Panasonic's increase only took effect on 9/1, both 7 trading days ago, the same age as the power branch's policy catalyst. This is an ongoing price-hike cycle led by the industry leader, not an afterglow.

  Table scrolls horizontally
• Positive logic: direct. CCL is the first link in the pass-through of electronic cloth price hikes, and what is affected is revenue and gross margin, not sentiment.
• Branch stage: day 2 of acceleration, and spreading upstream. On 9/10 the printed circuit board industry saw main-force net inflow of RMB 3.017 billion (1st across all industries), with 5-day RMB 18.283 billion and 10-day RMB 6.527 billion.
• Fundamental check: PE (TTM) 36.33, PB 12.98, total market cap RMB 55.66 billion. ⚠️ PB 12.98 is the number this edition most needs to flag — it means the market has already paid a steep premium for this round of price hikes.
• Industry position: CCL second-in-command (not the leader; the leader is Shengyi Technology).
• Technicals and sentiment: limit-up on 9/10 with turnover of RMB 4.791 billion, the highest among all limit-up stocks; ⚠️ but the seal-to-turnover ratio was only 0.060, the seal broke twice intraday, and it only sealed for good at 13:44 — this is a weak-seal pattern.
• ⚠️ Dragon-Tiger seat drill-down (new in this edition; yesterday's recap could not obtain it because it was generated at 15:40): Dragon-Tiger net buy of RMB 549 million (buy 890 million / sell 341 million). ⚠️ The buy and sell lists contain duplicate seats; the following is already de-duplicated by seat:

  Seat	Net amount	Nature
  Institutional seat ①	+RMB 167 million (buy 186 / sell 18)	Institution
  Shenzhen Connect dedicated seat	+RMB 140 million (buy 288 / sell 148)	Northbound
  Guotai Haitong · Shanghai Songjiang Zhongshan East Rd	+RMB 131 million	Hot money
  Guosheng Securities · Ningbo Sangtian Rd	+RMB 72 million	Hot money
  GF Securities · Zhengzhou Nongye Rd	+RMB 21 million	Hot money
  Institutional seat ②	+RMB 23 million	Institution
  Institutional seat ③	+RMB 12 million	Institution
  East Money · Lhasa Financial City South Ring Rd	−RMB 17 million	Hot money
  The three institutional seats together net bought about RMB 202 million and Shenzhen Connect net bought RMB 140 million; together they account for 62% of net buying.
  ⚠️ This is the only name in this edition where institutions and northbound money net bought at the same time — the quality of the buying is clearly better than in the power branch (Huayin Electric Power's top five buyers contain zero institutions, and Mindong Electric Power saw Shenzhen Connect as a net seller).

  Table scrolls horizontally
• Final judgement: watch closely. ⚠️ The reason for withholding "priority deep-dive" is the poor seal quality + PB 12.98 + no fresh catalyst; the reason for the high score is that the quality of the buying is the hardest on the whole list. If it gaps up more than +4% today, the risk of chasing clearly exceeds the reward (yesterday's counter-examples of chasing gap-ups were extremely dense).

52 湖南发展000722recommendation rationale rewritten after the sub-agent check · Hunan Development · Shenzhen main board ±10%

⚠️ The draft rationale for this section was "PE 25.48, the only normally valued name in the power branch." On inspection that PE is unusable and the rationale had to be rewritten. The recommendation slot is retained but the argument changed — and the process itself is worth showing readers.

• Related news: no new company-specific news. ⚠️ The branch's two public narratives (RMB 5 trillion of grid spend, September month-on-month tariffs) were falsified point by point in §2.3.
• ⚠️ Why the PE is unusable: in April 2026 the company completed the acquisition of 85% of Gaodian, 90% of Tongdian, 90% of Qingdian and 88% of Xiaodian — four hydropower companies — through share issuance plus cash (CSRC Permit [2026] No. 551), which constitutes a business combination under common control and requires retrospective restatement. Comparable 2025H1 figures were adjusted from the roughly RMB 170 million originally disclosed to RMB 407 million, whereas the RMB 349 million / 72 million in the 2025 annual report are unrestated figures. Subtracting a restated H1_2025 from an unrestated FY2025 artificially shrinks the numerator — three sources give three different PE (TTM) readings of 43.61 / 25.48 / 61.7, and none is credible. This edition marks it "cannot be confirmed." ⚠️ The same cause produces a common misreading: the comparison "H1 net profit RMB 185 million > FY2025's RMB 72 million" is invalid — the two numbers are not on the same company caliber.
• Rewritten recommendation rationale (the fundamentals themselves are hard):
  • The only one of the four power names with both revenue and profit growth in 2026H1: revenue RMB 533.8 million (+31.20%), net profit attributable to parent RMB 185.0 million (+77.45%), ex-non-recurring RMB 123.9 million (+202.9%), all on the adjusted comparable caliber;
  • Cash-flow corroboration: operating cash flow RMB 367.9 million, net-profit-to-cash ratio 1.99, the best of the four;
  • Gross margin 64.82% (+14.7pp YoY), with the clean-energy segment at 69.93%;
  • Controllable installed capacity raised from 245 MW at the start of the year to 745 MW (+204.06%);
  • Water inflows well above average: Gaotan +66.05%, Tongwan +34.87%, Qingshuitang +29.98%, Mangtangxi +25.97%, Xiaoxi +21.90%, Zhuzhou Hangdian +15.77% (only Niaoerchao at −5.31%);
  • PB 2.0–2.5 (67.1st percentile), gearing 40.24%.
• ⚠️⚠️ Three deductions that must be stated at the same time:
  1. The growth was bought, not organic — gearing jumped from 9.62% to 40.24%, with long-term borrowings of RMB 1.590 billion;
  2. Non-recurring gains of about RMB 61 million, 33% of attributable profit (ex-non-recurring / attributable = 67%), and this edition did not obtain their specific source; if they are one-off items related to the acquisition, the sustainable portion of the +202.9% must be reassessed, and 2027 will face a high base;
  3. ⚠️ Its growth has nothing to do with the "rising tariffs" branch narrative — the real causes are asset injection + abundant inflows. Falsifying the branch narrative does not affect its fundamentals, but it will affect its share price today.
• Technicals and sentiment: limit-up on 9/10, seal-to-turnover ratio 0.152, volume ratio 2.33, turnover rate 6.88% (the lowest in the branch). ⚠️ But the THS fund caliber shows a net outflow of RMB 3.4292 million on its limit-up day (different from East Money's caliber; the two cannot be subtracted, so both are shown for reference).
• Final judgement: watch closely. ⚠️ The risk warning needs to be stated more forcefully: fundamental quality determines the floor and the position size, not the ranking of today's move — yesterday's recap had just taken this loss on Suntak Technology (every balance-sheet argument right, yet the wrong direction for the day). The reason to buy it is not "it will go up today" but "if the power branch can extend today, it is the only name in that branch with genuine earnings support."

53 ⚠️ 闽东电力000993cut from recommendation candidate to watch only · Fujian Mindong Electric Power · Shenzhen main board ±10%

This one gets its own section because it is the name whose conclusion changed most in this edition.

• The status yesterday's recap gave it: the power branch's only 2-board name, with verification point 3 stating explicitly that "if it fails to reach 3 boards tomorrow and fewer than 2 of the 7 limit-up names go multi-board, this is a sector-level pulse rather than a new theme." It was supposed to be today's single most important object of observation and a recommendation candidate.
• ⚠️ Last night's 17:42 abnormal-movement announcement directly weakened it. The three core sentences:
  • "The company's principal business is currently operating normally, the principal business has not changed, and there has been no material change in the internal or external operating environment";
  • "The company, the controlling shareholder and the actual controller have no material matters under planning";
  • "No media reports or market rumors requiring clarification or response have been identified, and no other material events likely to have a significant impact on the share price have been identified."
• ⚠️ The board resolution at 17:45 the same evening was an announcement about "adjusting the membership of the board's specialized committees" — no project, no investment, no amount. I deliberately read it because the headline "extraordinary board meeting" looks a lot like a catalyst inside the window; having read it, I can confirm it is not.
• ⚠️⚠️ The Dragon-Tiger seat drill-down provides a third independent piece of negative evidence:
  • Shenzhen Connect net sold RMB 11 million (buy RMB 22 million / sell RMB 33 million);
  • The top selling seat was Huafu Securities' Ningde Tianhu East Rd branch, net selling RMB 90 million — Mindong Electric Power is registered in Ningde, Fujian, so this is the local seat selling;
  • The buying seats are all hot money (Galaxy Guangzhou Yuehai, Kaiyuan Xi'an Xidajie, Changjiang Jiangyin Hongqiao South Rd, Huayuan Jiangsu), with not one institutional seat;
  • Total Dragon-Tiger net buying was only RMB 62 million, read as "main force day-trading."
• Final judgement: watch only. Three independent pieces of evidence (company self-certification of no matters / local seat selling / northbound net selling) point the same way, and its PE is as high as 98.24. As a branch verification signal it remains extremely important (see §8), but as a holding it does not qualify.

54 ⚠️ 中国船舶600150the first name to deal with today · China State Shipbuilding · Shanghai main board ±10%

• Related news: company announcement at 00:30 today (primary). At 11:15 on 9/10 a foreign-flagged cargo ship caught fire during repairs at the company's controlled subsidiary CSSC Qingdao Beihai Shipbuilding; of the 42 people on board, 12 evacuated safely, 5 were injured and hospitalized, and all 25 missing persons were found with no signs of life; the Ministry of Emergency Management dispatched a working group; Shandong Province, Qingdao City and CSSC have mobilized resources to respond.

• ⚠️ The shareholding relationship must be made explicit, because public search returns the opposite answer: media reports found say Beihai Shipbuilding is "not a subsidiary of China State Shipbuilding but belongs to China Shipbuilding Industry Corp" — that statement is based on the 2024 structure. China Shipbuilding Industry Corp was absorbed into China State Shipbuilding via share swap in August 2025 and delisted, after which China State Shipbuilding controls seven backbone yards: Jiangnan, Dalian, Waigaoqiao, Guangzhou Shipyard International, Wuchang, CSSC Chengxi and Beihai. In today's announcement the company itself states it is a "controlled subsidiary," and that is authoritative.

• ⚠️ Pricing status: entirely unpriced, and the timeline can be checked point by point. On 9/10 China State Shipbuilding closed +3.26% on turnover of RMB 7.145 billion, the leading gainer in that day's shipbuilding concept. The key is that every disclosure about the scale of casualties came after the close:

  Time	Event
  11:15	Fire breaks out (during the A-share session)
  14:30	Open flames extinguished (during the A-share session)
  15:00	⚠️ A-share close, China State Shipbuilding +3.26%
  17:30	Qingdao Fire Department's statement: 20 of the missing had been recovered, none with signs of life
  19:06 / 19:23	Tencent News and Jiemian News publish
  9/11 00:30	Company announcement (all 25 missing persons found, none with signs of life)

  Table scrolls horizontally

  What was known during the session was "a fire"; what became known after the close was "25 people dead." That explains why the stock rose instead of falling that day, and it means the entire information set has to be digested at today's open.

• Risk calibration: ⚠️ this is a major industrial safety accident with 25 deaths and has already triggered intervention by a Ministry of Emergency Management working group. The announcement does not mention a production halt, nor does it quantify the operating impact — that is not "impact under control," it is "impact not yet assessed." Possible follow-ups include safety rectification campaigns, delays to ships under construction or repair, and accountability at the group level.

• Final judgement: avoid. ⚠️ And note: peers in the same branch — CSSC Offshore & Marine Engineering (600685, +2.67% on 9/10) and China CSSC Power Group (600482) — may be dragged along by sentiment, while the marine equipment sector was positive on all three fund-flow lookbacks (+0.488 / +2.310 / +2.342 billion) — which is exactly the point this edition most needs to make: a branch with a beautiful fund-flow picture must be avoided today for a reason the fund-flow picture cannot explain.

55 ⚠️⚠️ 太极实业600667attribution still blank, but the "memory" attribution has been definitively falsified · Wuxi Taiji Industry · Shanghai main board ±10%

Why it matters: limit-up on 9/10 with main-force net inflow of RMB 1.891 billion at a 50.52% net share, the largest main-force net inflow in the entire market. Yesterday's recap explicitly required that "a point-by-point retrospective must be done pre-market, and it may not enter a recommendation slot until attribution is established." This edition ran its deepest check of all on this name (the sub-agent downloaded and parsed 14 primary CNINFO announcements).

① Company-level catalyst: confirmed zero. A CNINFO query returns 0 announcements from 8/28 to now. The most recent is the 8/27 interim report and supporting documents (note: the disclosure date is 8/27, not the 8/26 recorded in yesterday's recap — 8/26 was the board meeting date). The limit-up analysis articles found were published on 7/27, 7/31 and 8/11, all old news, and have been discarded.

② ⚠️⚠️ Decisive counter-evidence: on 9/10 the A-share memory chain fell, and Taiji Industry was the lone exception.

Name	9/10 change	Name	9/10 change
太极实业 (Wuxi Taiji Industry)	+10.02%	江波龙 (Longsys)	−0.30%
深科技 (Shenzhen Kaifa)	+4.32%	兆易创新 (GigaDevice)	−0.32%
德明利 (Demingli)	+2.09%	香农芯创 (Shannon Semiconductor)	−0.77%
普冉股份 (PUYA Semiconductor)	+1.41%	长电科技 (JCET)	−0.76%
华天科技 (Tianshui Huatian)	+1.30%	紫光国微 (Unigroup Guoxin)	−1.16%
佰维存储 (Biwin Storage)	+0.50%	通富微电 (Tongfu Microelectronics)	−0.02%
澜起科技 (Montage Technology)	+0.02%	上证指数 (Shanghai Composite)	−0.43%

Table scrolls horizontally

⚠️ This table overturns the attribution "Taiji Industry = memory stock" itself. The overnight U.S. memory selloff was transmitted normally along the same A-share chain (memory and packaging/testing fell broadly), while Taiji Industry limited up against the tide, moving opposite to the memory theme. Among the 35 names in the 9/10 limit-up pool it was classified under "engineering consulting," the only one that day. The real themes that day were copper-clad laminate/PCB + power/gas + banks, and it touches none of them.

③ ⚠️⚠️ The premise that "memory price moves affect its profit" was written off by the company itself in its own announcements.

The "operating risk" sections of the two abnormal-movement announcements, Lin 2026-039 (6/26, first statement) and Lin 2026-046 (8/14, restatement), read verbatim:

"…engineering technical services accounted for more than 80% of operating revenue in 2025; the company's semiconductor business currently does not involve HBM products. Among them, controlled subsidiary Haitai Semiconductor… uses a profit model of 'full cost + agreed return (10% of total investment + excess return)', and is little affected by industry upcycles."

Corroboration from the contract text (Lin 2025-035, the Fourth-Phase Back-End Process Service Contract, 2025-07-01 to 2030-06-30): service fee = full cost + agreed return; agreed return per year = total investment × 10% per year + excess return, where "excess return" is annualized compensation of 10% on the portion of borrowings above US$225 million (essentially interest compensation), not a performance bonus. There is no clause anywhere linked to DRAM prices — wafers are supplied by SK Hynix, so DRAM prices do not even pass through the cost side.

Empirical numbers (the strongest single piece of evidence):

	2024	2025	2026H1
Haitai net profit	RMB 237.11 million	RMB 239.06 million	RMB 117.02 million
Net margin	6.58%	6.07%	6.05%

Table scrolls horizontally

DRAM prices swung violently from 2024 to 2025, and Haitai's net profit moved by just +0.8%. That is the textbook signature of a cost-plus model. ⚠️ The one cyclical linkage that does exist points the wrong way: the incentive clause (US$2.3 million–10 million) is tied to whether SK Hynix's annual operating profit is positive, with a cap worth about 8.7% of attributable net profit; SK Hynix is currently at a profit peak and the bonus is most likely already at the cap, so an upcycle adds nothing while a downcycle subtracts — a one-sided downward exposure.

④ Fundamentals: the "flat" income statement is propped up by a one-off reversal.

Metric	2026H1	YoY
Operating revenue	RMB 13.144 billion	−14.88%
Net profit attributable to parent	RMB 315.5 million	−3.55%
Ex-non-recurring net profit attributable to parent	RMB 223.5 million	⚠️ −31.01%
Gross margin	7.08%	−0.78pp
Weighted ROE	3.63%	−0.14pp
Gearing	73.30%	—

Table scrolls horizontally

⚠️ The 27.5pp gap between attributable −3.55% and ex-non-recurring −31.01% comes entirely from a RMB 264.34 million reversal of receivables impairment provisions (closure of the legacy Baohua Shanghai property matter, Lin 2026-048). Strip it out and real operating profit is down about a third year on year.

⚠️ Valuation once again proves the quote software untrustworthy: the sub-agent computed PE (TTM) = RMB 41,789 million / (448.1 − 327.1 + 315.5) = 95.73×, matching Tencent's and Baidu's third-party calibers exactly, while the quote software shows 66.22. Ex-non-recurring PE is 130.8×; PB is 4.76×. PE sits at the 92.5th percentile of the last ten years and PB at the 92.3rd. — An ROE of just 3.63% and negative ex-non-recurring growth, yet carrying a PE and PB at the 92nd percentile of a decade.

⑤ ⚠️ Trading pattern: 17 limit-ups year to date, and the two with the largest turnover were followed by the worst drawdowns.

• 17 sessions in 2026 closed up ≥+9.8% (7/27, 7/31 and 8/11 are only three of them), and the company has issued 6 abnormal-movement announcements (including 1 triggered by a decline).
• After a limit-up, the next day was up 8 out of 16 times — a coin flip.
• ⚠️⚠️ The two worst limit-ups of the year (7/1 and 7/9) were followed by −35.17% and −38.22% over the next 10 days, with maximum drawdowns of −35.70% and −48.04% — and those were exactly the two limit-ups with the highest turnover of the year (RMB 12.988 billion and RMB 8.692 billion). From a 7/1 high of 32.10 to a 7/20 low of 15.17, 13 trading sessions, −52.7%, including four consecutive limit-downs.
• Currently RMB 19.98, +142.5% year to date, but still −37.8% from the 7/1 high.

⑥ ⚠️ Four facts on the fund-flow and governance side:

• Of the RMB 1.891 billion main-force net inflow on 9/10, RMB 1.882 billion (99.5%) was extra-large orders, with large orders contributing only RMB 10 million; the 8 minutes before the seal (09:30–09:37) saw RMB 2.766 billion of turnover, 73.2% of the whole day, while the volume ratio was only 1.04 — what is abnormal is the concentration, not the total.
• It did not appear on the 9/10 Dragon-Tiger List (amplitude 9.42%, turnover rate 9.13%, below the disclosure threshold), so the identity of the buyers cannot be verified.
• Wuxi Gaotou, a person acting in concert with the controlling shareholder, announced on 7/10 that its holding had fallen from 33.52% to 31.22%; the general manager sold between RMB 18.10–19.61 from 8/5 to 8/7 (almost exactly today's price of 19.98).
• Margin balance of RMB 1.607 billion, 4.23% of free-float market cap (the market norm is 1–2%).

⑦ What the real semiconductor transmission chain would be (if one insists on finding it): it is fab capex, not DRAM prices — 十一科技 (SDRI) signed new contracts worth RMB 33.762 billion in 2026H1 (+195.88%), of which electronics/high-tech was RMB 29.72 billion (+373.66%). ⚠️ But there are three limits: ① this is public information already disclosed in the 8/27 interim report, not news on 9/10; ② it is a cumulative figure with no single-quarter breakdown (memory cumulative-period-figures-are-old-news); ③ engineering general contracting carries a gross margin of only 1.96%, and new contracts take 1–2 years to convert into revenue.

Final judgement: Pass (Pass type = logic does not hold).

⚠️ But the wording here must be precise, because two conclusions are easy to conflate: ① "Why did it rise on 9/10" — still unknown. Six avenues were checked (announcements / memory theme / that day's sectors / fundamental inflection / valuation / seats) and all came back negative; the strongest statement available is "fund-driven, attribution unknown." "Cannot find news" does not equal "there is no news" (memory no-news-found-is-not-no-news), and this remains the largest cognitive gap in this edition. ② "Does the memory thread affect its profit" — this one now has an answer: no. Confirmed by three independent strands of evidence: the contract clauses, the company's own statements, and two years of profit figures. So the overnight U.S. memory selloff has no material effect, positive or negative, on its current earnings — ⚠️ yesterday's recap's reasoning that "overnight memory crashed, so avoid Taiji" has been overturned by this edition itself; the correct reason to avoid it is attribution unknown + valuation at the 92nd percentile + ex-non-recurring −31%, not memory. "Pass" here means "we do not endorse it," not "we expect it to fall."

56 中远海能600026招商南油 (Nanjing Tanker, 601975, Shanghai main board ±10%) · COSCO Shipping Energy · Shanghai main board ±10%

• Related news: SC2610 crude night session +6.19%.
• Positive logic: ⚠️ indirect, and the transmission chain has been falsified twice. Tanker profits are set by VLCC spot rates; crude price is only a proxy.
• Fundamentals: COSCO Shipping Energy PE 12.82 / PB 2.42; Nanjing Tanker PE 13.92 / PB 1.85 — both sit in the lowest valuation band on the whole list, and there is nothing wrong with the fundamentals themselves.
• Technicals and sentiment: on 9/10 COSCO Shipping Energy was −0.61% (the high of the day was the open) and Nanjing Tanker +0.42% (open→close −1.02%). The shipping & ports industry limit-up count went 0→3→0.
• Final judgement: watch only. Full argument in §2.1. If two or more names in this branch hit limit-up before 11:30 today, this judgement is overturned; follow the tape.

57 超声电子000823Sino Ultrasonic Electronics · Shenzhen main board ±10%

• CCL/PCB, limit-up on 9/10. PE 49.02 / PB 2.00 is the cheapest valuation in branch 1.
• ⚠️ But the technical pattern is among the worst in the branch: turnover rate 25.93%, seal-to-turnover ratio only 0.023 (among the market's weak seals), Dragon-Tiger read as "main force day-trading" (buy RMB 509 million / sell RMB 417 million, net buy only RMB 92 million).
• Judgement: watch only. A cheap valuation did not offset the fact that a quarter of the float changed hands in one day.

58 国际复材301526九鼎新材 (Jiuding New Material, 002201, Shenzhen main board) · CPIC International · ChiNext ±20%

• The very top of the fiberglass / electronic cloth chain; on 9/10 they were +6.79% (net inflow RMB 697 million) and limit-up respectively.
• ⚠️ The time structure of the branch's fund flow is contradictory: fiberglass manufacturing was +1.633 billion today, but only +0.025 billion over 5 days and −1.798 billion over 10 days. Today's money just arrived; it is not persistent.
• ⚠️ Glass futures +2.08% cannot be used as corroboration — the futures contract tracks flat (construction) glass, a different product from electronic cloth / fiberglass yarn. This is an easily misused spurious correlation.
• Judgement: CPIC International watch closely; Jiuding New Material watch only (PE 77.36, market cap only RMB 6.07 billion).

59 生益科技600183not a holding, but a signal · Shengyi Technology · Shanghai main board ±10%

• The industry leader in copper-clad laminate. On 9/10 it was −0.81% with main-force net outflow of RMB 499 million, while second-tier names in the branch rose broadly.
• ⚠️ Verification point 2 in yesterday's recap, verbatim: "if the sector still rises tomorrow while both leaders keep seeing net outflows, judge it second-tier rotation with leader distribution, and downgrade this branch from 'acceleration' to 'divergence'."
• Judgement: observe. This is branch 1's most important health indicator today (see §8).

510 ⚠️ 华银电力600744strongest money, worst quality; cut from "watch only" to "Pass" · Hunan Huayin Electric Power · Shanghai main board ±10%

• Limit-up on 9/10 with main-force net inflow of RMB 464 million at a 34.74% net share and Dragon-Tiger net buying of RMB 222 million, the most concentrated money in the power branch.
• ⚠️ But the seat drill-down is negative: the top five buying seats contain not one institutional seat — they are Guoxin Securities Beijing Branch (+RMB 91 million), Shanghai Connect (+RMB 78 million), Kaiyuan Securities Xi'an Xidajie (+RMB 63 million), Orient Securities Lhasa Jinzhu West Rd (+RMB 51 million, a classic hot-money seat) and Guoxin Securities Beijing Zhongguancun Ave (+RMB 22 million). ⚠️ The two Guoxin Securities branches together account for +RMB 113 million, and two seats from the same broker is a hallmark of a single pool of money split across seats, not institutional consensus.
• ⚠️⚠️ The fundamental check came back far more damning than the draft, which is why this name was downgraded:
  • PE (TTM) is actually −193.9 (negative), not the 137.56 shown by quote software; ex-non-recurring PE is −1,470. TTM attributable profit = 81.57 (FY25) + 51.68 (H1_26) − 206.56 (H1_25) = −RMB 73.31 million, i.e. a loss of RMB 125 million in 2025H2 alone dragged TTM negative.
  • ⚠️ Ex-non-recurring net profit was positive in only 2 of the last 12 years (+RMB 6 million in 2016, essentially zero, and +RMB 149 million in 2025). This is not cyclical fluctuation; the core business simply does not make money over the long run.
  • ⚠️ Undistributed profit of −RMB 6.146 billion. At the 2025 earnings level of RMB 82 million it would take about 75 years to work off the accumulated losses — dividend capacity is zero for the foreseeable future.
  • ⚠️ Gearing of 92.82%, and even that number is understated — the roughly RMB 510–560 million gap between attributable equity of RMB 1.726 billion and "net assets per share 0.5977 × 2.031 billion shares = RMB 1.214 billion" can only be other equity instruments (perpetual bonds); EPS 0.021 × 2.031 billion = RMB 42.65 million ≠ attributable profit of RMB 51.68 million, and the RMB 9 million difference is exactly half a year of perpetual-bond interest (about 3.5% annualized). A market cap of RMB 14.2 billion rests on RMB 1.214 billion of common shareholders' net assets.
  • PB 11.71, at the 95.2nd historical percentile — ⚠️ this is not a premium, it is a collapsed denominator: net assets have been eroded to a sliver by cumulative losses, mechanically inflating PB.
• ⚠️ Its elasticity is real, but it points the other way: the thermal on-grid tariff pass-through mechanism does hold, and a tariff rise of about 1.7% would pull TTM back to breakeven, with roughly RMB 43 million added for each additional 1%. But in 2026 its binding constraint is not tariffs, it is volume — the interim report says "abundant inflows in Hunan and a large increase in hydro generation… priority absorption of clean energy squeezed coal-fired generation space," with output −6.67%. A 1% tariff increase (+RMB 43 million) roughly only offsets a 1% volume decline (about −RMB 39 million); the two legs point in opposite directions. ⚠️ And the hydro squeezing it is exactly Hunan Development's seven stations with inflows +16%~+66% — two sides of the same coin; you cannot be long both at once.
• ⚠️ One cross-observation worth logging: Guoxin Securities Beijing Branch appeared the same day as the top buyer in both Huayin Electric Power and Chongqing Zaisheng Technology for almost exactly identical amounts (RMB 90.9748 million / RMB 90.9736 million) — two stocks in different branches, the same seat, the same magnitude. That looks more like a unified programmatic/managed-account operation than a judgement on power fundamentals.
• Judgement: Pass (Pass type = position/valuation too high + logic does not hold).

---

6. Pass list

⚠️ As required by yesterday's recap, this edition adds a "Pass type" column, distinguishing "logic does not hold" (we do not endorse it, but that is not a bearish call) from "position/valuation too high" (direction may still be up, but risk/reward is poor) and "exogenous risk." Conflating the three produces systematic misreadings.

---

7. Intra-branch ranking

7.1 Branch 1: electronic cloth → copper-clad laminate → PCB (⚠️ first establish which end is rallying today)

Yesterday's lesson, written here: on 9/10 the order of gains along this chain was "upstream copper foil/fiberglass (+6%+15%) > CCL (+2.8%+10%) > PCB manufacturing (−1%~+1%)," and both of yesterday's recommendation slots were at the very bottom. Today the ranking runs strictly from upstream down.

• Hardest name: Jin'an Guoji (best quality of buying).
• Highest trading visibility: Jin'an Guoji (top turnover in the market).
• Who is chasing: Jiuding New Material, Yihao New Material (small caps, high PE).
• Who gets a Pass: Suntak Technology, Ellington Electronics (direction reversed; they are the cost side).

7.2 Branch 2: power / green power / gas

⚠️ The ranking basis for this table was rewritten after checking: PE is no longer used (the field is unreliable for this branch, see the §3 note) and is replaced by four items — "H1 earnings direction + cash-flow corroboration + PB historical percentile + whether the tariff pass-through mechanism holds."

• Hardest name: Hunan Development (the only one with genuine earnings growth corroborated by cash flow).
• ⚠️ The mismatch most worth noting separately: Sinergy Clean Energy has the strongest tariff pass-through mechanism in the branch and the lowest valuation percentile (PB 8.2nd), yet its realized tariff fell the most in the branch (−7.0%), and the company is actively converting new projects to fixed-tariff contracts, giving up the option on rising tariffs. — The name with the best mechanism is voting with its own operating behavior to say "I do not believe tariffs will rise." That is the strongest single vote against the whole branch narrative.
• Who gets a Pass: Huayin Electric Power, Leshan Electric Power, Meineng Energy, Shuifa Gas.
• ⚠️ The branch's structural defect: of 7 limit-ups only 1 went multi-board, and that multi-board name's own announcement says it has no undisclosed matters. This is the same shape as the tankers on 9/9 (3 limit-ups, zero the next day).

---

8. Verification signals at today's open

⚠️ Per Discipline B, this table is marked with priorities. Structural (multi-board / sector leader / multi-day fund flow) > exogenous price (commodities / overseas markets). When they conflict, follow the structural one. ⚠️ Yesterday's lesson: verification point 1 (commodity price) passed with a +3.71% margin and the trade still lost money; verification point 2 (zero multi-board names) failed, and the market followed verification point 2.

---

9. Final recommendations

① Stocks most worth watching today

⚠️⚠️ This edition fills only 3 slots and deliberately leaves 2 blank, with reasons given. The reason for leaving them blank is not that no names could be found, but that: the only S-grade event in the window is a negative, the overnight external market is a negative for yesterday's strongest fund direction (memory), the market is on day 4 of an ebb phase with four consecutive declines in limit-up counts, and neither long branch has a fresh catalyst this morning. Forcing 5 slots under that combination would put "the list must have 5 names" above "the list must make money." Leaving the 5th slot blank yesterday proved right.

② The three strongest branches today

③ Directions not recommended for chasing today

1. ⚠️⚠️ 中国船舶 (China State Shipbuilding, 600150) and the shipbuilding chain. A fire at a controlled subsidiary killed 25 people and the Ministry of Emergency Management is involved, while the stock was +3.26% on 9/10 — this news is entirely unpriced. ⚠️ The branch is positive on all three fund-flow lookbacks (+0.488/+2.310/+2.342 billion); please do not let a pretty fund-flow picture make you overlook this.
2. ⚠️⚠️ Memory / semiconductor packaging & testing / semicap. Overnight Philadelphia Semiconductor −2.66%, Micron −4.90%, SK Hynix −5.20%, Lam Research −5.65%; memory chips 5-day −8.602 billion / 10-day −42.100 billion, semicap 5-day −3.966 billion / 10-day −11.464 billion.
3. ⚠️ Tanker shipping (not oil). Do not buy tankers because crude was +6.19% in the night session — 9/9 and 9/10 already ran this experiment twice, with shipping & ports limit-ups going 0→3→0.
4. ⚠️ PCB manufacturing (downstream). Suntak Technology and Ellington Electronics are the cost bearers of CCL price hikes; the direction is reversed.
5. ⚠️ Indiscriminate follow-through on yesterday's limit-up stocks — for a fourth straight day. The 48 limit-up stocks from 9/9 had a median return of −1.04% on 9/10, with only 45.8% rising.
6. ⚠️ Chasing gap-ups. Yesterday's counter-examples were dense: Baida Group gapped up +8.80% → open→close −10.89%; Nanjing Port opened at limit-up → open→close −6.56%.
7. ⚠️ Weak-seal limit-up stocks (seal orders / turnover <0.05): 超声电子 Sino Ultrasonic Electronics (0.023), 再升科技 Chongqing Zaisheng (0.024), 美能能源 Meineng Energy (0.020, seal broken 5 times), 水发燃气 Shuifa Gas (0.039), 游族网络 Youzu Interactive (0.018), 新力金融 Xinli Finance (0.048, seal broken 6 times), 京基智农 Kingkey Smart Agriculture (0.035).
8. ⚠️ Grid equipment, communication cable, agriculture/fertilizer, brokers — 5-day and 10-day fund flows are both negative, excluded under Discipline A. 8b. ⚠️⚠️ Do not use "RMB 5 trillion of grid investment" as a reason to buy generators. That money goes to UHV / distribution networks / digitalization, and the payees are transmission-and-distribution equipment makers; for generators it adds not one kWh and not one fen of tariff (§2.3). ⚠️ And the equipment side itself is negative on both the 5-day and 10-day lookbacks (−2.986 / −4.535 billion), so neither end works — this narrative has no buyable landing point in A-shares today. 8c. ⚠️ Nor should you treat "September tariffs rising month-on-month" as a delivered earnings positive. The four power names' actual realized tariffs in 2026H1 were +0.8% / −1.0% / −7.0% / unavailable, not one delivered; and any rise in the agency purchase price may come entirely from "system operation fees," which flow to ancillary services and capacity compensation and never reach generators' pockets. 8d. ⚠️ 华银电力 (Hunan Huayin Electric Power, 600744) and 湖南发展 (Hunan Development, 000722) cannot both be long ideas. The "abundant inflows in Hunan and sharply higher hydro generation" squeezing Huayin's output is the same event as Hunan Development's seven stations seeing inflows +16%~+66% — two sides of one coin.
9. ⚠️ The lone high-board name 桂林旅游 (Guilin Tourism, 000978) at 4 boards — it belongs to no fund-flow theme, Dragon-Tiger net selling of RMB 32 million, and the previous lone name (Baida Group) went open→close −10.89% the next day.
10. ⚠️ Not participable: 优彩资源 Youcai Resources (002998, one-word board on resumption, cannot be bought), 信达证券/东兴证券 Cinda Securities / Dongxing Securities (9/14 is the last trading day), *ST 清越 *ST Qingyue (suspended for delisting).

④ Final one-sentence judgement

Today is not a day for hunting strength; it is a day for defusing two landmines first. The only S-grade event in the window is a major safety accident at China State Shipbuilding's controlled subsidiary (25 dead) with yesterday's share price entirely unpriced, while the overnight external market landed squarely on memory, the market's largest fund inflow yesterday — both point down, and both act on what looked strongest yesterday. On the long side essentially only the "electronic cloth → copper-clad laminate" upstream remains (it has the largest magnitude among branches positive on all three fund-flow lookbacks, 7–9× comparable branches; Jin'an Guoji is the only name on this list with both institutional seats and northbound money putting real cash in); ⚠️ while the power thread was downgraded by this edition's own fundamental check — its rising limit-up count and positive flows on all three lookbacks are facts, but of the two public narratives supporting it, one targets the wrong link in the chain and the other failed to show up in any of the four names' actual realized tariffs; what is really driving it is a valuation narrative rather than earnings improvement, and readers must treat it as "a fund-driven theme," not "a cyclical inflection." Crude's +6.19% in the night session is the most eye-catching price signal, but the same mapping has been falsified two days running, so this edition chooses not to follow it and has written "if two or more tanker names limit up today, this judgement is wrong" explicitly into the verification points. Day 4 of the ebb, limit-up counts down four sessions in a row, failed-seal rate 38.6% — three recommendation slots and two publicly blank ones are the honest answer today.

---

⚠️ Risk warning: this list is a pre-market information review and observation only and does not constitute investment advice. A-share volatility risk is extremely high, and automatically generated content may contain stale information or errors in industry-chain mapping; it must not be used directly as a basis for trading.

Supplementary notes (please read alongside):

1. Meaning of the labels: "priority deep-dive / watch closely / watch only / Pass" in the text are pre-market information-strength grades only and carry no implication about position size, entry timing or target price. ⚠️ In particular, "Pass" is used for three different situations (logic does not hold / position and valuation too high / exogenous risk), which §6 separates via the "Pass type" column; "we do not endorse it" does not mean "we are bearish on it."
2. Data read times: quotes, sector and single-stock fund flows and the limit-up / failed-limit-up / limit-down pools are on the 2026-09-10 close caliber; the Dragon-Tiger List is for 9/10, published after 18:00 last night; overnight U.S. equities are as of the 9/10 close; commodity futures are from the 9/11 night/electronic session around 23:00; the announcement window is 9/10 15:00 → 9/11 07:15.
3. ⚠️ This edition has eight known evidence gaps; please discount the related conclusions accordingly: ① three-lookback fund flow for the shipping & ports sector was not obtained, so §2's fund judgement on tankers had to substitute concept and single-stock data; ② VLCC spot rates (TCE) were not obtained, and they are the tanker branch's true earnings driver — one of the direct reasons this edition gives tankers no recommendation slot; ③ the attribution of Wuxi Taiji Industry's (600667) 9/10 limit-up remains unresolved for a second consecutive session; ④ no new price-hike letter dated 9/10–9/11 was obtained for CCL / electronic cloth, so it is characterized only as "an ongoing price-hike cycle"; ⑤ multiple text sources contradict each other on the absolute level of international crude, so this edition uses only the relative move in domestic SC crude futures; ⑥ ⚠️ the energy-component breakdown and month-on-month direction of September 2026 agency purchase prices for Fujian, Hunan and Zhejiang were not obtained (the summary page is an image and cannot be parsed); this edition verified only Guangdong's Pearl River Delta at +6.5%, and none of the four power names are in Guangdong — this single item could on its own falsify or confirm the entire tariff narrative and is this edition's highest-priority open question; ⑦ the specific source of Hunan Development's (000722) roughly RMB 61 million of non-recurring gains was not obtained; if they are one-off items tied to the acquisition, the sustainable portion of the +202.9% ex-non-recurring growth must be reassessed; ⑧ the capacity breakdown by power source and the exact perpetual-bond amount for Huayin Electric Power (600744) were not obtained — the roughly RMB 560 million of perpetual bonds is inferred from the gap between attributable equity and net assets per share, and needs confirmation from the "other equity instruments" note in the interim report. 3b. ⚠️⚠️ This edition found one data trap capable of changing the direction of a conclusion; please take note: the PE shown by quote software (East Money) is wrong for several power names, and it uses four different algorithms for the same field (see the §3 note). Huayin Electric Power's true PE (TTM) is negative at −193.9 while the software shows 137.56 — copying the software reading would turn a chronically loss-making stock with undistributed profit of −6.146 billion, ex-non-recurring profit positive in only 2 of 12 years and gearing of 92.82% into "a high-valuation growth stock." The body text now uses recomputed values throughout, cross-validated against Baidu's full-history caliber; stocks not independently re-verified are explicitly tagged "East Money caliber, not re-verified."
4. ⚠️ "Cannot find news" does not equal "there is no news." The "attribution unknown" verdict on Taiji Industry in §5.5 is missing information, not negative evidence; please do not read it backwards as "no catalyst, therefore it should fall."
5. ⚠️ Statements about 中国船舶 (China State Shipbuilding, 600150) follow the company's 2026-09-11 00:30 announcement text. Public search contains the claim that "Beihai Shipbuilding is not a subsidiary of China State Shipbuilding"; that claim is based on the ownership structure before China Shipbuilding Industry Corp was absorbed via share swap and delisted in August 2025 and no longer applies. The company's announcement clearly identifies it as a "controlled subsidiary." The company has not yet quantified the accident's impact on operations — this edition makes no estimate of the magnitude.
6. Price limits differ by board: Shanghai/Shenzhen main board ±10%, ChiNext/STAR Market ±20%, Beijing Stock Exchange ±30%. Do not read a ±20% / ±30% name with ±10% intuition. The STAR Market and the Beijing Stock Exchange also impose investor suitability thresholds such as RMB 500,000 in assets, and some readers are not eligible to trade them.
7. New-listing note: Enflame Technology (688801) lists on the STAR Market today, and a new listing contaminates its sector's fund-flow and price-change readings on day one, so today's readings for those sectors cannot be used directly for cross-sector comparison.
8. Incomplete-coverage statement: this edition screened 1,647 announcements inside the window one by one, but deep single-stock checks covered only about 40 names, and the vast majority of listed companies were not studied individually; not being mentioned does not mean there is no opportunity, and being mentioned does not mean it has been studied sufficiently.
9. Stock-specific risks: the stocks discussed are generally exposed to earnings misses, shareholder selldowns, lock-up expiries, regulatory inquiries, trading suspensions for verification, abnormal price movements and insufficient liquidity. 湖南黄金 (Hunan Gold, 002155)'s major asset restructuring still requires approval at today's shareholder meeting (it was rejected on 7/29), review by the Shenzhen Stock Exchange and registration with the CSRC, and both approval and timing remain materially uncertain.
10. This report was produced by an automated process and may contain stale information, industry-chain mapping errors or deviations caused by data-interface failures; it must not be used directly as a basis for trading. Please rely on official exchange and company disclosures.
11. Position statement by the producer: the producer of this report holds none of the securities discussed and has no interest in the companies mentioned.

---

Data sources: East Money company announcements (np-anotice-stock, eiTime caliber, 1,647 inside the window), East Money quote and fund-flow APIs (push2delay.eastmoney.com, today/5-day/10-day across three lookbacks for 496 industry and concept sectors), East Money's official limit-up / failed-limit-up / limit-down pools (9/8, 9/9, 9/10), East Money Dragon-Tiger List and seat detail (9/10), Shanghai International Energy Exchange crude futures and the Zhengzhou/Shanghai exchanges' night session (akshare), CNBC quote API (U.S. equities and commodities). Primary announcements: China State Shipbuilding fire accident announcement¹, Fujian Mindong Electric Power abnormal-movement announcement³, Tongguang Cable tender-win announcement². News and industry information: Jiemian News (Beihai Shipbuilding fire)⁶, Science and Technology Daily⁷, Al Jazeera (Hormuz crisis)⁸, CNBC (Middle East crisis and tanker rates)⁹, Sina Finance (Enflame Technology listing)⁵, 21jingji (Shenzhen Kaifa expansion)⁴, China News Service (RMB 300 billion special treasury bonds)¹⁰.