A-Share Pre-Market Brief | 2026-08-25 Tuesday

Scan window: 2026-08-24 (Mon) 15:00 close → 2026-08-25 (Tue) 08:10 (Beijing time). Quote basis: East Money push2delay API, 2026-08-24 closing prices, local read time 2026-08-25 07:0x–08:0x. Overseas markets are the 2026-08-24 US close (Beijing time 8/25 04:00; timestamps verified item by item). Futures are SHFE night session (21:00–02:30) and COMEX new-session reads, each labelled with its own read time, never mixed.

⚠️ Three basis declarations, please read first:

1. The quote API's "P/E" field is dynamic P/E (half-year profit ×2, annualized), not TTM. Everywhere this report draws a valuation conclusion, TTM is computed separately and labelled. The gaps are large: Wanhua Chemical dynamic 11.38 vs TTM 13.91; Luxshare Precision dynamic 26.5 vs TTM attributable 23.3 / TTM ex-non-recurring 28.5; T&S Communications dynamic 113.0 vs TTM 117.2. Both Luxshare's and Wanhua's profits are heavily concentrated in H2 (Luxshare's H1 is only about 40% of the full year); annualizing via H1×2 systematically understates full-year profit and overstates P/E.
2. Limit-up/limit-down is always judged by each board's own cap, not by a ±10% gut feel: ChiNext/STAR ±20%, BSE ±30%. All limit-up identifications come from the exchange's official limit-up pool, with tc == len(pool) verified (limit-up 46=46, limit-down 11=11, no silent failure).
3. The interim-report season has two dedicated traps; this report screened for both stock by stock: ① the earnings pre-announcement came first — a formal report is not necessarily new information (Wanhua Chemical pre-announced on 7/7, see §6); ② retrospective restatement under common-control merger — China Shenhua's prior-year comparative has been restated from RMB 24.641 billion to RMB 27.583 billion, so comparing against the old base systematically overstates growth (see §6).

---

0. Today in One Sentence

The only genuinely new hard information inside the window is two and a half items: Luxshare Precision's (002475) Q3 earnings pre-announcement, the cash flow and profit purity in Sunrise Optoelectronics' (300502) interim report that the pre-announcement did not give, and WuXi AppTec's (603259) cybersecurity incident (negative). Everything else that looks exciting is either already priced yesterday, or propped up by a false claim that "gold hit an all-time high."

And the judgment that matters most today is not "what to buy," it is "do we chase yesterday's defensive rotation." The answer: no incremental confirmation overnight, don't chase. On 8/24 A-shares and US equities moved the same way on the same day — semis down, defensives up, precious metals up on both sides (A-share electronics sector main-force net outflow RMB 36.117 billion; Nasdaq −0.76% while the Dow +0.26%). The rotation is real and cross-market; but whether it carries into today depends on whether fresh fuel arrived overnight, and four independent pieces of evidence say it did not:

1. Domestically-priced gold and silver spiked and faded in the night session. SHFE silver continuous hit a night-session high of 17,072 (+1.36%) and closed at 16,775, −0.40% vs the 8/24 day-session close; SHFE gold continuous closed at 1,005.96, −0.09% vs the day-session close. The night session (21:00–02:30) happens to cover the main US trading hours and is domestic capital's most direct single vote on the overnight overseas move — and it voted for a pullback.
2. The Dragon-Tiger List shows yesterday's limit-up silver names being sold. Shengda Resources (000603) net buy −RMB 211 million on the list; Hunan Silver (002716) −RMB 198 million on one of its lists; among the six active names only Silver Nonferrous (601212) at +RMB 180 million is positive.
3. "Gold hit an all-time high" is wrong, and wrong by a wide margin; the truth about silver is more extreme still. Gold's 2026 peak was $5,626.8 (1/29 intraday) / $5,446.5 (1/28 close); it is now $4,726, 16.0% below. Silver's peak for the year was $121.78 (1/29 intraday), then it crashed roughly 30% in a single day on 1/30 (CME margin hike triggered forced liquidation), bottomed for the year at $55.765 on 7/16, a −54% peak-to-trough drawdown. The current $70 is a +25.5% bounce off a deep hole, still −42.5% from the peak. The Forbes headline that day was "Gold Price Hits 15-Week High" (a 15-week high, i.e. back to mid-May). Several Chinese closing wraps wrote "all-time high" — that is not only a wrong number, it turns a bounce inside a deep drawdown into a breakout. This report does not use it.
4. The catalyst itself is outside the window, and domestic silver prices barely moved on 8/24. The US Treasury's doubling of long-end buybacks was announced on 8/19 and takes effect on 9/9; the actual price jump happened on 8/20 (SGE silver +5.24%) and 8/21 (+3.29%). On 8/24 itself SGE silver rose only +0.12% and SHFE ag0 +0.43% (SGE gold +2.10% the same day). So the 8/24 limit-ups in silver names were not driven by that day's silver price; they are day 3 of pricing the 8/20–8/21 move, pure capital and sentiment relay.
5. In coal, the stocks ran ahead of the commodity. On 8/24 coal-industry stocks rose +2.80% and the coke sector +3.64%, while the same day coking coal futures front month rose only +0.16%, coke futures +1.56%, and thermal coal CCI5500 has been flat at RMB 860/tonne for four consecutive readings. The real fundamental driver happened between May and mid-August (coking coal +23.3% in a month, +46.6% YTD), and there has been no verifiable incremental supply event since 8/20.

Pre-market state assessment: a low-win-rate event waiting period. Three hard constraints:

1. This week's watershed is 8/27 (Thursday), not today. NVIDIA's FY2027Q2 report lands after the US close on 8/26, i.e. the early hours of 8/27 Beijing time, so the A-share reaction day is Thursday. NVIDIA is already −2.91% to $208.48 on 8/24 — pre-emptive de-risking has begun. Today's and tomorrow's compute-chain trades are, in essence, prepaying a premium for an event two days out.
2. 500 companies disclose interim reports today (exchange booking basis), with 8/31 the disclosure deadline — this is the densest week of earnings landmines.
3. Relay money is a net outflow. The "yesterday's limit-up incl. one-word board" concept saw main-force net outflow of RMB 753 million on 8/24, and "theme stocks" RMB 6.886 billion — the index structure looks like rotation, but the money-making effect has not followed.

The biggest self-correction in this report versus yesterday's (8/24): yesterday's top recommendation Shengyi Technology fell −4.89% that day and the entire Top 5 declined, while the precious-metals group that yesterday explicitly Passed produced 2 genuine limit-ups. Yesterday's fundamental analysis was not itself wrong; the error was treating "financial quality" as a same-day ranking device — in a cross-market style switch, statement quality does not determine that day's move. This report therefore raises "fund flow and position" to the same weight as fundamentals: the within-branch ranking in §7 is based entirely on capital structure and seal quality, not financials.

---

1. News Overview

Four "news items" this report explicitly corrects or downgrades

① "COMEX gold hit an all-time high on 8/24" — factually wrong, not used anywhere in this report. Several Chinese closing wraps wrote "COMEX gold closed at $4,621.1, +1.09%, an all-time high." The price magnitude is plausible (close to the read at the A-share close), but "all-time high" is wrong: gold's all-time high is roughly $5,589–5,600, set on 2026-01-28, about 16% above the current level; silver's all-time high is $121, versus about $69 now, about 42% below. The Forbes headline on 8/24 was "Gold Price Hits 15-Week High", and IndexBox likewise states explicitly that both metals are "still below the all-time highs set earlier this year." Why this one is the most dangerous: it turns a bounce inside a deep drawdown into a breakout to new highs. Silver falling from $121 to $69 and then bouncing back to $69 is a completely different trade from "new high."

② The overnight increment in precious metals: overseas up, the domestic night session down — this report goes with the night session and does not split the difference.

Rationale: A-share precious-metals stocks price RMB-denominated gold and silver; night-session participants overlap heavily with A-share money and the session covers the main US trading hours — the night session is the read closest to today's A-share open. Also note from the Forbes original: silver "broke above $70 for the first time since June over the weekend," and it is now around $69 — it has already fallen back below $70.

③ Coal: on 8/24 the equity side ran ahead of the commodity side, with no new fundamentals that day.

The real fundamental driver came earlier: coking coal front month +23.3% over the past month, +46.6% YTD, +36.3% YoY (8/24 close 1,584, 12-month range 1,016.5–1,591.5); coke closed 8/24 at its highest level in 12 months. The supply-side events are also all earlier: 186 coking-coal mines in Shanxi halted production from 5/23 (capacity involved 252.65 million tonnes), the 8/5 accident-related halt at Shanxi Coking Coal's Xiqu mine (approved capacity 2.7 million tonnes), and the 8/13–8/22 self-inspections at Yuyang washing plants. There has been no verifiable incremental supply event since 8/20. Assessment: the 8/24 coal rally is a catch-up to a month of commodity gains that already happened, plus defensive rotation, not new fundamentals that day.

④ The US-Canada tariffs (effective 8/22) and the DRC copper-cobalt export ban (signed 6/29) are both outside this window. Both are real and still developing, but neither is "something that happened last night." The DRC ban is already in week 8 of pricing. This report treats them as background variables and awards them no score.

---

2. Strongest Positive Branches, Descending

Branch-by-branch notes

Branch 1 | Precision manufacturing — this report's core finding is not in the pre-announcement, it is in a self-disclosure in the body of the interim report.

What the market sees is "H1 ex-non-recurring profit up only +6.47%," and the single-quarter split looks worse still: Q2 ex-non-recurring was −0.13% YoY (while revenue was +44.5% YoY). But this is not deteriorating operations, it is an accounting presentation mismatch — and the company itself spelled it out in the body of the interim report. Original text:

Affected by exchange-rate volatility, the company's foreign-currency asset and liability exposure produced approximately RMB 1.986 billion of FX losses in the first half … of which RMB 1.297 billion of FX risk management gains were recognized in non-recurring gains and losses. Because FX gains and losses on the company's foreign-currency monetary assets and liabilities are presented in finance costs and thus included in net profit both before and after deducting non-recurring items, while the gains and losses on the FX hedging business that hedges those FX risks are presented as non-recurring items and are excluded from ex-non-recurring net profit, the ex-non-recurring net profit fails to reflect the full effect of the FX hedge.

The finance-cost note gives hard evidence: FX gains/losses were +RMB 1.9855 billion (a loss) in 2026H1 versus −RMB 493.0 million (a gain) in 2025H1, a YoY deterioration of RMB 2.479 billion; while net interest expense actually improved by RMB 105 million (from RMB 328 million to RMB 224 million).

That is: the loss falls entirely inside the ex-non-recurring line, the hedging gain falls entirely outside it. Restated (using the H1 effective tax rate of 9.9% and an attributable share of 92.0%, an estimate, not a disclosed fact):

• Zeroing out FX in both periods: 2026H1 ex-non-recurring ≈ RMB 7.608 billion vs 2025H1 ≈ RMB 5.197 billion → +46.4%
• Restoring only the RMB 1.297 billion hedging gain (acknowledging a residual RMB 689 million net exposure loss after hedging): adjusted ex-non-recurring ≈ RMB 7.037 billion → +25.7%

Both bases point the same way: true operating ex-non-recurring growth is between +26% and +46%, and the reported +6.47% badly understates it. Meanwhile attributable net profit at +18.04%, which contains both the loss and the hedging gain, is the more honest FX-neutral read.

That is the entire investment logic of this branch: not "an inflection in growth," but "the market is pricing it off a number that accounting presentation has depressed."

But four counter-arguments must be given alongside it, or this logic becomes a one-way narrative:

1. The company's 2026 delivery-versus-guidance position has shifted systematically (see the table in §5-1): all three landed at or below the low end, and H1 ex-non-recurring even broke through its own guidance floor by 4.7% with no correction announcement issued. If that pattern persists, Q3 attributable net profit is only +10.85%, a continued deceleration from Q1 (+20.24%) and Q2 (+16.17%).
2. TTM free cash flow is −RMB 2.109 billion (TTM operating cash flow RMB 16.537 billion − TTM capex RMB 18.646 billion); H1 operating cash flow was −RMB 2.446 billion (prior-year period −RMB 1.658 billion).
3. The +40.16% revenue growth is not organic: auto electronics +274.10% comes almost entirely from consolidating Leoni (consolidated around 2025Q3), contributing RMB 23.733 billion of the RMB 50.001 billion revenue increment (47.5%). Leoni enters the YoY base from 2026Q3, so Q3 revenue growth will slow markedly.
4. The segment matching the AI narrative, "communications and data centre," is only 9.52% of revenue, and it is the only segment with a sharp gross-margin decline (−2.48pp).

One market worry that is falsified (in the bulls' favour): the itemized goodwill note in the interim report contains no Leoni entry, opening balance = closing balance, and the note states explicitly that "goodwill arose from business combinations not under common control in prior years." Goodwill is only RMB 2.239 billion, 0.64% of total assets and 2.39% of attributable net assets — the widely feared "large Leoni goodwill impairment" does not exist in the statements.

Branch 2 | Optical modules — this is not "buy the branch," it is "stratify by quality inside the branch." All four have now reported (all single-quarter figures below are computed by this report from cumulative differences):

• Eoptolink is the only one leading on all four of growth, cash flow, profit purity and valuation: non-recurring items are only RMB 18.43 million (0.24% of attributable net profit), operating cash flow is positive and +69.67% YoY, and TTM P/E of 43.8× is below Innolight's 49.8× and Shengyi's 59.1×.
• Innolight has the fastest growth, but Q2 operating cash flow was −RMB 1.568 billion (all of H1's RMB 1.800 billion came from Q1), and Q2 attributable net profit exceeded ex-non-recurring by RMB 542 million.
• T&S Communications' Q2 revenue was −0.9% YoY (Q1 was still +40.7%) — growth already stopped in Q2 — while its TTM P/E of 117.2× and P/B of 43.56 are the most expensive in the group. It fell −8.63% on 8/24, the worst of the big three, consistent with the fundamental ranking. Its interim report came out on 8/19 (outside the window), and the media diagnosis at the time was "a 77% surge in passive components cannot mask stalling active components and repeated chip shortages."

Branch 3 | Defensive / dividend — real, but be clear about its nature. On 8/24: Ping An Insurance (601318) +2.94% (P/B 0.97, below book), Kweichow Moutai (600519) +2.50%, China Pacific Insurance (601601) +2.01%, China Merchants Bank (600036) +1.75%, Shaanxi Coal (601225) +2.99%; at the concept level, below-book dividend stocks +1.02% with main-force net inflow of RMB 606 million, dividend stocks +0.77% with net inflow of RMB 577 million, scarce resources +0.67% with net inflow of RMB 999 million. The cross-market corroboration is the hardest part: the same day US consumer defensives led, energy and tech lagged, and the US 10-year fell to 4.70%. But the nature must be stated plainly: this is "passive absorption after growth stocks bled," not good news arriving at the defensive sectors themselves. Electronics saw single-day main-force net outflow of RMB 36.117 billion, while the largest inflow, industrial metals, was only +RMB 1.002 billion — the outflow is 36× the inflow; the overwhelming majority of the money left rather than switched. Durability depends entirely on when growth stocks stabilize, and 8/27 is that switch.

Branch 4 | Semiconductor equipment — this report's only case of "one piece of news, two legs pointing in opposite directions." YMTC Holdings' IPO (accepted 8/21 23:10, outside the window) raises RMB 33 billion including RMB 20.8 billion for production-line technology upgrades. In US equities it was read as negative: Micron and SanDisk fell notably on 8/24 on the grounds that "another Chinese memory chipmaker filed for an IPO in Shanghai" — what is being priced is future supply growth. The same logic splits inside the A-share market too:

Three markets, two legs, direction fully consistent. So "the YMTC IPO is good for the memory sector" is a wrong mapping; the correct mapping stops at equipment. But be honest: the RMB 20.8 billion of capex is delivered in 2027.

Branch 5 | Precious metals — the most heavily revised item in this report. After verifying the revenue mix, the conclusion went from "downgrade" to "the name of this branch is itself wrong."

Layer one: there was no "silver limit-up wave."

⚠️ East Money's "dynamic P/E" is a single-period annualization for every name in this group, and the 82.70 for Silver Nonferrous is simply wrong — it takes 2026Q1's RMB 149.98 million ×4, whereas the company's 7/15 pre-announcement already said H1 would be only RMB 64–95 million. In its own 8/25 pre-market abnormal-movement filing Silver Nonferrous states: static P/E −65.69×, dynamic P/E −88.40×. All TTM figures here are recomputed as "FY2025 − 2025 same period + 2026 same period."

Layer two: the 8/24 dividing line was not "gold vs silver" at all, it was "deeply oversold small caps vs heavyweights."

Note: Sichuan Gold, Zhaojin Gold and Western Region Gold are pure gold names yet all surged, while Hengbang Co. (gold smelting) fell 1.68%. So this is not "buy silver, sell gold," it is a rotation into previously oversold small- and mid-cap precious-metals elasticity names.

The decisive evidence — the gain is highly correlated with the drawdown from the January high:

All six peaked for the year on 2026-01-29, all six bottomed in mid-to-late July, and all six rose 18–26% over the last 5 sessions. And the two that fell the furthest (Silver Nonferrous −55.9%, Hunan Silver −45.8%) are precisely the only two that sealed a 2nd consecutive limit-up on 8/24. That is the signature of an oversold bounce, not of a fundamental ranking.

Layer three (the most important): of the three bought as "silver stocks," two are not silver companies at all.

Layer four: the name the market treats as the leader lost money in Q2 — in a quarter when gold and silver were surging. Silver Nonferrous' 7/15 pre-announcement put H1 attributable net profit at RMB 64–95 million, while Q1 attributable net profit already reached RMB 149.98 million ⟹ a 2026Q2 single-quarter loss of RMB 55–86 million. The pre-announcement also states explicitly that profit came from "an increase in fair-value gains on embedded derivative financial instruments arising from pricing transactions" (not the core business), and flags that "asset impairment testing is still in progress." Also: gearing ratio 69.91% (highest in the group); largest shareholder CITIC Guoan has pledged 98.56% of its holding; and CITIC Group's share-reduction plan is inside its window (2026/6/8–9/7, up to 1% of total shares).

Conclusion: only one name in this branch has silver credentials that survive verification (Shengda Resources), and it carries an independent risk that has nothing to do with the silver price — the company and its actual controller Zhao Mantang were both placed under CSRC investigation for suspected disclosure violations on 2026-06-15/16, still unresolved. So this report handles the whole precious-metals branch as follows: strength stays at B, but no recommendation slot and no watch-only slot are granted, and all six go on the Pass list.

---

3. Master Ranking of Single-Stock Catalyst Strength (descending by stock)

"Catalyst level" measures how hard the news itself is; "total score" measures how actionable this stock is today — the two need not agree. P/E is always labelled with its basis.

---

4. Single-Stock Scoring Model (100 points)

---

5. Detailed Analysis of the Top 10

1️⃣ Luxshare Precision002475Priority deep-dive | Total score 74 · SZSE main board ±10%

Related news: after the close on 2026-08-24, filing 2026-090 "2026 Q3 Earnings Pre-Announcement" PDF¹; the interim report the same day PDF¹⁴.

Catalyst logic (direct): the real information is not in the pre-announcement, it is in the body of the interim report. See Branch 1 in §2 — the company voluntarily disclosed H1 FX losses of RMB 1.986 billion (inside the ex-non-recurring line) and hedging gains of RMB 1.297 billion (outside it), and the finance-cost note shows FX gains/losses deteriorating RMB 2.479 billion YoY while net interest expense actually improved by RMB 105 million. Restated, true operating ex-non-recurring growth is +26%–+46%, versus a reported +6.47%.

Single-quarter breakdown (single quarter = cumulative difference)

Basis	H1 cumulative	Q1 alone	Q2 alone	Q3 alone (implied by guidance)
Revenue	RMB 174.504bn (+40.16%)	RMB 83.888bn	RMB 90.616bn (+44.5%)	Not guided
Attributable net profit	RMB 7.843bn (+18.04%)	RMB 3.660bn (+20.24%)	RMB 4.183bn (+16.17%)	RMB 5.403–6.555bn (+10.85%–+34.49%, midpoint +22.67%)
Ex-non-recurring	RMB 5.962bn (+6.47%)	RMB 2.776bn (+15.22%)	RMB 3.186bn (−0.13%)	RMB 4.403–5.755bn (+11.67%–+45.95%, midpoint +28.81%)

Table scrolls horizontally

Prior-year Q3 base: attributable RMB 4.874bn, ex-non-recurring RMB 3.943bn

⚠️ But the pre-announcement itself carries very little information, and it must be read at the low end — this is the most important passage in this section:

Pre-announcement	Range	Actual	Landing point
2025 9M attributable	+20%–+25%	+26.92%	Above the ceiling
2026 Q1 attributable	+20%–+22%	+20.24%	Close to the floor
2026 H1 attributable	+18%–+22% (RMB 7.840–8.106bn)	RMB 7.843bn (+18.04%)	Only RMB 0.003bn above the floor
2026 H1 ex-non-recurring	+11.74%–+20.06% (RMB 6.257–6.722bn)	RMB 5.962bn (+6.47%)	Broke through the floor by 4.7%, with no correction announcement

Table scrolls horizontally

The 2026 landing positions have shifted systematically versus 2025. If that pattern persists, 9M attributable net profit is more likely to come in near RMB 13.246 billion, implying Q3 attributable growth of only +10.85% — a continued deceleration from Q1 (+20.24%) and Q2 (+16.17%), the opposite direction to the "AI-driven acceleration" narrative. Also: the company dropped the Q3 single-quarter table that was in the 2025 pre-announcement, reducing disclosure granularity; and the ex-non-recurring range translates into a 26.6%-wide band for Q3 alone.

Industry branch stage: launch phase. Luxshare fell only −1.70% on 8/24 (a day when 3,965 names declined), turnover 1.24%, and the pre-announcement came out after the close, so it was priced at zero that day.

Fundamental verification (item by item)

Item	Value
Gross margin (H1)	11.78% (prior year 11.61%, +0.17pp); Q1 11.92% / Q2 11.65%
Segment mix	Consumer electronics RMB 122.476bn (+19.27%, 70.18% of revenue, gross margin 10.25%); auto electronics RMB 32.391bn (+274.10%, 18.56% of revenue); communications and data centre RMB 16.609bn (+49.66%, only 9.52% of revenue, gross margin −2.48pp, the only sharp decline)
Operating cash flow (H1)	−RMB 2.446bn (prior-year period −RMB 1.658bn, a 47.54% YoY deterioration); FY2025 was +RMB 17.325bn, so a negative first half is seasonal
TTM free cash flow	−RMB 2.109bn (TTM OCF RMB 16.537bn − TTM capex RMB 18.646bn). Net-profit-to-cash conversion 203% (FY2024) → 104% (FY2025) → 93% (TTM), a one-way decline
Gearing ratio	67.44% (about 63.5% pro forma for the H-share raise); quick ratio 0.65 (below 0.8 for five consecutive periods)
Interest-bearing debt	RMB 134.816bn (short-term borrowings RMB 93.569bn); interest coverage ≈9.6×
Goodwill	Only RMB 2.239bn (0.64% of total assets, 2.39% of net assets); Leoni generated no goodwill — the widely feared goodwill-impairment risk does not exist in the statements
Working capital	Receivables RMB 47.666bn (+54.4%), inventory RMB 50.367bn (+47.5%), payables RMB 74.533bn (+48.4%) — all three growing faster than revenue at +40.16%
Interim dividend	RMB 1.1 per 10 shares, RMB 849mn in total (payout ratio about 10.8%)

Table scrolls horizontally

Valuation (total market cap about RMB 414.5bn, A-shares + H-shares converted from HKD; A-share close on 8/24 RMB 53.72)

Metric	Value
P/E (TTM attributable)	23.3× (TTM attributable RMB 17.799bn)
P/E (TTM ex-non-recurring)	28.5× (TTM ex-non-recurring RMB 14.532bn)
P/E (dynamic, East Money)	26.5× (= H1×2 annualized; H1 is only about 40% of the full year, so this basis systematically overstates P/E by about 25%)
P/B	4.43×; about 3.60× pro forma for the H-share raise
Weighted ROE	H1 8.76% (about 17.5% annualized); FY2025 21.10%

Table scrolls horizontally

Industry position: leader in consumer-electronics ODM / precision manufacturing, a core-army name. The H-shares translate to RMB 50.49, only a 6.0% discount to the A-shares — an extremely narrow A/H spread.

Technical sentiment: SZSE main board ±10%. The −1.70% on 8/24 was markedly resilient (the SZSE Component Index was −2.13%), turnover 1.24%, volume did not expand, no one-word board and no gap-up-then-fade. Market cap RMB 414.5bn — low one-day-wonder risk, but low elasticity too; do not hold it to theme-stock expectations.

Risks (four, by severity)

1. Customer concentration: in FY2025 the top five customers were 65.04% and the largest customer 56.68% (single source; internally consistent with the revenue back-calculation but the annual-report text was not obtained — to be verified). The consumer-electronics segment is 70.18% of revenue at a gross margin of only 10.25%.
2. FX remains the main variable for Q3: USD/CNY went from 7.1530 in 2025-08 to 6.7120 in 2026-08, roughly 6.2% of RMB appreciation over a year, with a further 1.20% inside Q3; exports are 83.70% of revenue. And because the H-share proceeds are in HKD, failing to convert them promptly would amplify the FX exposure.
3. Leoni base effect: from 2026Q3 it enters the YoY base, so the revenue growth rate and the structural gross-margin dividend fade together. Excluding Leoni, H1 organic revenue growth is about +20%–+23% (estimated).
4. Tight liquidity: quick ratio 0.65, short-term borrowings RMB 93.569bn; the company also announced the same day that it plans to register a bond issue, consistent with that picture.

Final judgment: priority deep-dive, but what you are buying is "repair of a misread accounting presentation + valuation," not "an inflection in growth." Three supports: ① the market is pricing it off +6.47% while the FX-restated figure is +26%–+46%; ② TTM P/E 23.3× and P/B 4.43 (3.60 pro forma) are the lowest on the whole list; ③ the market's biggest worry, Leoni goodwill impairment, is falsified by the statements. Stated equally clearly: the Q3 pre-announcement itself carries very little information, taking the midpoint may be too optimistic, and read at the floor it is decelerating. This is the point of disagreement in this recommendation that most needs cross-verification.

---

2️⃣ Eoptolink300502Watch closely | Total score 73 · ChiNext ±20%

Related news: the "2026 Interim Report" released after the close on 2026-08-24 summary text². Revenue RMB 20.90975bn (+100.34%), attributable net profit RMB 7.52917bn (+90.98%), ex-non-recurring RMB 7.51074bn (+90.92%), net operating cash flow RMB 1.61641bn (+69.67%), weighted ROE 34.86% (−3.26pp), total assets RMB 36.647bn (+41.60%). No cash dividend, no bonus shares, no capitalization issue.

Catalyst logic (direct): at the attributable-profit level there is almost no expectation gap — the 7/19 guidance was RMB 7–8bn and the actual was RMB 7.529bn, dead centre; the ex-non-recurring guidance was RMB 6.981–7.981bn and the actual RMB 7.511bn, likewise dead centre. The incremental information is not in the magnitude of profit, it is in the two things the guidance did not contain:

• Non-recurring gains and losses were only RMB 18.43 million, 0.24% of attributable net profit (ex-non-recurring / attributable 99.76%, 99.87% in Q2 alone). Innolight is 95.91% for H1 and 93.15% for Q2 alone.
• Operating cash flow of +RMB 1.616bn, +69.67% YoY, growing in step with net profit. Innolight's H1 operating cash flow was RMB 1.800bn, −44.08% YoY, with Q2 alone at −RMB 1.568bn.

Single quarter: Q2 revenue RMB 12.572bn (+96.9% YoY, +50.8% QoQ); Q2 attributable net profit RMB 4.749bn (+100.5% YoY, +70.8% QoQ); Q2 ex-non-recurring RMB 4.743bn (+100.5% YoY). Profit is growing faster than revenue, so net margin is expanding.

Industry branch stage: divergence / early ebb. The big three were hammered together on 8/24, and the communications-equipment sector saw main-force net outflow of RMB 14.074bn. The interim report came out after that decline, so today is the first reaction day.

Fundamental verification: core business is 800G/1.6T optical modules. TTM attributable net profit RMB 13.119bn → P/E 43.8×, below Innolight's 49.8× and Shengyi's 59.1×; P/B 23.64 (high, but below T&S at 43.56).

Technical sentiment: ChiNext ±20% — the −6.79% on 8/24 is only an ordinary correction on ChiNext. Turnover 3.94%, value traded RMB 20.57bn.

Final judgment: watch closely, not a priority slot. One reason, but a hard one: NVIDIA reports after the US close on 8/26, so the A-share reaction day is 8/27. Buying today = actively holding exposure to an event two days out. The statement quality supports it outperforming peers; it does not support adding ahead of the event.

---

3️⃣ Ping An Insurance601318Watch closely | Total score 66 · SSE main board ±10%

No stock-level news; it is here because of style. The cross-market defensive rotation moved the same way on the same day (US consumer defensives led, the US 10-year fell to 4.70%), and two markets moving together on the same day rules out an "isolated A-share sentiment" explanation. Fundamentals: P/B 0.97 — below book; P/E (dynamic) 5.37; total market cap RMB 994.5bn. This report has not verified its 2026 interim report and draws no conclusion on embedded value or new business value. Technical sentiment: +2.94%, turnover 1.23%, moderate volume. Final judgment: watch closely. But the nature is passive absorption — electronics outflow RMB 36.1bn versus the largest inflow sector at +RMB 1.0bn, the money is leaving, not switching; once growth stocks stabilize on 8/27, this leg gives it back first.

---

4️⃣ China Coal Energy601898Watch closely | Total score 64 · SSE main board ±10%

This is the only coal name that survives item-by-item verification. 2026 interim report (disclosed 8/22)¹⁵: revenue RMB 73.136bn (−1.8%), attributable net profit RMB 8.149bn (+5.8%), ex-non-recurring RMB 8.544bn (+11.7%), operating cash flow RMB 9.864bn (+28.6%), weighted ROE 4.95% (flat). Interim dividend of RMB 1.84 per 10 shares.

Why it is cleaner than Shaanxi Coal, Shenhua and Yankuang — one comparison table:

Company	H1 status	Attributable YoY	Ex-non-recurring YoY	Source of growth
China Coal Energy (601898)	Disclosed (8/22)	+5.8%	+11.7%	Core business; ex-non-recurring outpaces attributable
Shaanxi Coal (601225)	Pre-announcement only (7/11)	+47%–+53%	+31.76%–+38.10%	The roughly RMB 1.7bn gap comes from "investment income on selling listed-company shares held"
China Shenhua (601088)	Pre-announcement only (7/15)	−4.7%–+8.0% after restatement	+7.4%–+15.6%	Coal-chemical and transport volumes; the attribution never mentions coal prices
Yankuang Energy (600188)	Pre-announcement only (7/15)	About +53%	About +2%	Almost all of the increment comes from transferring 100% of Inner Mongolia Xintai Coal; core-business ex-non-recurring is only +RMB 100mn

Table scrolls horizontally

⚠️ China Shenhua's basis trap deserves its own paragraph: in H1 2026 the company completed the acquisition of equity in 12 target companies from China Energy Investment Group, a merger under common control, and has retrospectively restated the comparative period, with prior-year attributable net profit restated from RMB 24.641bn to RMB 27.583bn. On the restated (only comparable) basis, Shenhua's H1 attributable growth is −4.7%–+8.0%, i.e. broadly flat or slightly down. The widely quoted "+6.9%–+21.1%" compares the new statements against the old, pre-restatement base and systematically overstates growth.

Valuation: P/E (TTM) 10.71× and P/B 1.18×, both the lowest in this group. Technical sentiment: +3.57%, turnover 0.73%. Risks: revenue is still −1.8% YoY, with growth coming from cost and mix rather than volume-and-price together; it also has a retrospective restatement, so cross-period comparisons need care. Final judgment: watch closely. In a branch where the equities have already run ahead of the commodity (see §1-③), picking the one with the lowest valuation, the best ex-non-recurring quality, and an actual disclosure (not merely a pre-announcement) is the only prudent move.

---

5️⃣ Piotech688072Watch closely | Total score 62 · STAR ±20% · RMB 500 · 000 + 2-year eligibility threshold

The driver is out-of-window news (the YMTC IPO accepted on 8/21); it is here because of the 8/24 tape confirmation: on a day when 3,965 names fell, it rose +1.58% against the tape on turnover of only 2.08%. The inverse divergence between equipment and memory appeared simultaneously in A-shares and US equities (see Branch 4 in §2), so the mapping chain is verified by two markets. Domestic leader in thin-film deposition equipment, P/E (dynamic) 74.05, market cap RMB 198.9bn. Risks: the RMB 20.8bn of capex is delivered in 2027; the STAR eligibility threshold means some readers cannot participate (that is information, not a reason to exclude it). This report has not verified its most recent reporting period's profit mix and draws no conclusion on profit quality.

6️⃣ Naura Technology002371Watch closely | Total score 59 · SZSE main board ±10%

Same logic as Piotech, but the interim report is not out — it is a black box. −0.91% on 8/24, better than the semiconductor sector's −2.05%; turnover 1.07%, market cap RMB 513.7bn, P/E (dynamic) 78.56. Equipment names in the same group that have already reported generally decelerated (Hwatsing Technology attributable +11.44%, Piotech operating cash flow −92.63%), which is the biggest uncertainty. Watch closely, do not chase.

7️⃣ Shanghai Datun Energy600508Watch only | Total score 57 · SSE main board ±10%

The only one of the three limit-up coal names with new information that day: interim report disclosed 8/22 (Saturday)¹⁶, with 8/24 the first trading day after earnings. Revenue RMB 4.450bn (+27.21%), attributable net profit RMB 283mn (+38.02%), ex-non-recurring RMB 268mn (+34.28%), operating cash flow RMB 1.112bn (+550.06%, 3.9× attributable net profit), gross margin +2.17pp to 18.25%, P/B 0.78, below book. But three corrections keep it out of a recommendation slot: ① the absolute profit level is extremely low — TTM attributable net profit is only RMB 299mn, P/E (TTM) 33.49×; of the RMB 220mn earned in full-year 2025, RMB 205mn came in H1, meaning 2025H2 attributable was only about RMB 15mn, close to breakeven — this is a small-cap coal company with highly unstable earnings (market cap RMB 9.997bn). ② The gearing ratio rose from 34.12% to 41.65% within a year. ③ It completed a bonus/capitalization issue in June 2026, taking total shares from 722.7mn to 1,011.8mn, so per-share metrics are only comparable after restating for the new share count. Watch only.

8️⃣ Silver Nonferrous601212Pass(the first draft said "watch only"; downgraded after verifying the revenue mix) | Total score 22 · SSE main board ±10%

This is the most heavily revised name in this report, and the reasoning must be written out. The first draft listed it as the one name in the precious-metals group "leading on all four measures on a relative basis" and gave it "watch only," on tape-based grounds (a genuine 2nd consecutive limit-up, seal/turnover 0.208, turnover of only 2.95%, Dragon-Tiger net buy +RMB 180mn — all four of those tape facts still hold). But once the revenue mix was verified, the tape advantages stopped mattering, because it is not a silver stock at all.

① It is a copper smelter. The 2025 annual report text says its own mines "produced 323,500 tonnes of copper, lead, zinc and molybdenum metal in concentrate through the mining and dressing system, of which 90,200 tonnes of contained copper, 198,800 tonnes of contained zinc, 34,100 tonnes of contained lead and 379.52 tonnes of contained molybdenum" — there is no silver in that at all. All silver comes from comprehensive recovery from copper and lead anode slime at the smelting end. 2025 revenue by product:

Product	Revenue	Share	Gross margin
Cathode copper	RMB 34.404bn	40.26%	−0.10%
Gold	RMB 20.777bn	24.31%	6.04%
Other	RMB 13.678bn	16.01%	13.53%
Zinc	RMB 5.973bn	6.99%	−8.71%
Electrolytic silver	RMB 5.594bn	6.55%	11.88%
Copper concentrate (own mines)	RMB 2.312bn	2.71%	63.83%
Lead-zinc concentrate (own mines)	RMB 395mn	0.46%	24.66%

Table scrolls horizontally

Own mines total only 3.17% of revenue and electrolytic silver only 6.55%, with a further 20.92% being non-ferrous trading at a 0.27% gross margin. The word "Silver" in its name comes from its registered address — Baiyin City, Gansu Province.

② In a second quarter when gold and silver surged, it lost money. The 7/15 pre-announcement put H1 attributable net profit at RMB 64–95 million, while Q1 attributable already reached RMB 149.98 million ⟹ a Q2 single-quarter loss of RMB 55–86 million. The pre-announcement states plainly that profit came from "an increase in fair-value gains on embedded derivative financial instruments arising from pricing transactions" — not the core business — and flags that "asset impairment testing is still in progress." It is the only company in this group that still has not disclosed its 2026 interim report.

③ The widely quoted "P/E 82.70" is wrong. That is Q1's RMB 149.98 million ×4. TTM attributable net profit is negative, so P/E is meaningless. What the company itself wrote in its 8/25 pre-market abnormal-movement filing is: static P/E −65.69×, dynamic P/E −88.40× (the 89-peer average is 113.90 static, 108.38 dynamic).

④ Three structural risks unrelated to the share price: gearing ratio 69.91% (highest in the group); largest shareholder CITIC Guoan has pledged 98.56% of its holding (29.95% of total shares); and CITIC Group's share-reduction plan is inside its window (2026/6/8–9/7, up to 1% of total shares).

Final judgment: Pass. Its two limit-ups on 8/24 were genuine on the tape and its seal quality was the best in this group — but that was an oversold bounce in the small cap that had fallen the furthest (−55.9% from the January high, deepest in the group), unrelated to the silver price and unrelated to its own statements.

9️⃣ Shengyi Technology600183Watch only | Total score 54 · SSE main board ±10%

Yesterday's (8/24) top recommendation in this series, and it fell −4.89% that day — the call was wrong. The fundamentals remain solid: Q2 attributable net profit RMB 2.129bn (+147.0%), ex-non-recurring RMB 1.825bn (+123.1%), H1 operating cash flow RMB 2.983bn (+53.44%). But one item missed yesterday must be added today: TTM attributable net profit RMB 5.195bn → P/E 59.1×, the highest in the optical-module / compute-materials group (Eoptolink 43.8×, Innolight 49.8×), P/B 16.58. Ranking it first yesterday on "the only one with improving cash flow" without putting TTM valuation on the same table for comparison was the most substantive omission in yesterday's call. No in-window news (the interim report came out 8/15). Watch only.

🔟 Wanhua Chemical (600309, SSE main board ±10%) | Pulled from the recommendation slot → watch only | Total score 43

This is the most heavily revised name versus the first draft, and the reasoning must be written out. The first draft put it in a recommendation slot with the note "first disclosure date to be verified" — that has now been verified, and the conclusion reverses:

① A positive earnings pre-announcement was already issued on 7/7 (original¹⁷): attributable net profit RMB 9.8–10.4bn, ex-non-recurring RMB 9.6–10.2bn.

The expectation gap is not "small," it is neutral-to-negative. And the market gave it no reward on the 7/7 announcement day either: it gapped up to 73.64 and closed at 71.05 (+0.04%, a 3.5% fade from the high), then −1.45% on 7/8 and −1.63% on 7/9.

② Ex-non-recurring growth is only +54.96%, 9.4pp below attributable at +64.35% — the media generally report only the attributable figure. Non-recurring gains were +RMB 387mn (versus −RMB 122mn in 2025H1), and that single item manufactures roughly RMB 500mn of YoY "inflation."

③ Q2 single-quarter operating cash flow halved YoY. Alongside Q2 attributable net profit of RMB 6.345bn (+108.7% YoY), Q2 operating cash flow was RMB 4.927bn, −50.5% YoY, with OCF/attributable falling to 0.78× (3.27× in the prior-year period). The cause is clear: from the end of Q1 to the end of H1, inventory went from RMB 28.724bn to RMB 32.936bn (+RMB 4.212bn) and receivables from RMB 15.043bn to RMB 19.605bn (+RMB 4.563bn), tying up RMB 8.775bn in the quarter. A substantial part of Q2's high profit has not yet been converted to cash, and the inventory was built at Q2's high prices.

④ The company itself says prices have already peaked. From the operating-data filing accompanying the interim report: "prices for the polyurethane series overall rose first and then fell" (the Q1 filing said "showed an upward trend"). That is, prices already peaked and turned down within Q2, and the report released on 8/24 reflects a peak that has already passed. Moreover, polymeric MDI, the largest-volume product, saw H1 average prices down −1.2% YoY (a slight decline); the price-increase story was carried mainly by TDI (+31.5%) and flexible-foam polyether (+26.9%) — precisely the high-elasticity, low-barrier products. Feedstock rose harder still in Q2 (pure benzene ≈+58% QoQ, propane ≈+40%), and in that combination a 4.3pp Q2 gross-margin expansion is most plausibly explained by inventory gains from selling low-cost stock at high prices — and inventory gains are one-off, turning into write-down pressure when prices fall back.

⑤ Q3 has real output disruptions: Fujian MDI (down 6/23, restarted 8/1), BC in Hungary (down 7/15), Yantai MDI (down 8/5).

⑥ The one positive structural change (not a reason to recommend it today, but worth remembering): the capex cycle is receding — construction in progress fell from RMB 59.813bn in 2024H1 to RMB 33.460bn (−44%), capex fell −36.9% YoY to RMB 10.740bn, and free cash flow turned from −RMB 6.502bn to +RMB 1.044bn; it also paid its first interim dividend of RMB 8.10 per 10 shares (RMB 2.536bn in total, 25.2% of H1 attributable net profit, a 1.11% dividend yield).

Valuation: P/E (TTM) 13.91× (not the dynamic 11.38×), P/B 2.017×. P/B sits at the 5.3rd percentile over the past 10 years, P/E (TTM) at the 33.2nd percentile, and the 6.8th percentile over the past 3 years. ⚠️ But watch out for the reverse reading of that percentile combination: a 3-year P/E percentile of only 6.8% is precisely the product of a denominator that has just been filled by the Q2 peak. If Q2 is this cycle's earnings peak, P/E (TTM) will drift passively higher as the denominator rolls forward. A low P/B percentile is not itself a margin of safety.

Final judgment: pulled from the recommendation slot, cut to watch only. The information was released on 7/7; the incremental information added on 8/25 is — (a) ex-non-recurring landing at the guidance floor, (b) Q2 single-quarter operating cash flow −50.5%, (c) the first interim dividend, (d) capex and construction in progress continuing to recede — of which (a) and (b) are negative and (c) and (d) positive, netting out to neutral-to-negative. Treating this as an "earnings beat" event has no factual basis.

---

6. Pass List

---

7. Within-Branch Ranking

Branch 1 | Precision manufacturing · AI hardware

This branch has only one name with substantive evidence. No padding to fill the list — Goertek, Lens Technology and other peers had no reliable new news in this window, and forcing them in would mean writing "I could not find it" as "it benefits too."

Branch 2 | Optical modules (ranked by statement quality, not by price move)

Hardest: Eoptolink. Highest visibility: Innolight (market cap RMB 1.02 trillion, the branch's core army). The follower: Shengyi Technology. The Pass: T&S Communications — most expensive + the only one stalling, both true at once.

Branch 3 | Coal (ranked by statement quality and valuation, not by price move)

The one sentence that matters most in this branch: of the three coal stocks that hit the limit on 8/24, one (Meijin) is a victim of this price rise in supply-chain terms, one (Dayou) is a loss-making name with a deteriorating balance sheet, and only one (Shanghai Datun Energy) genuinely had new information that day — and its TTM earnings are only RMB 299 million.

Branch 4 | Precious metals (ranked by whether the silver credentials are real, which is this branch's first criterion)

The three sentences that matter most in this branch:

1. The 8/24 dividing line was not "gold vs silver," it was "deeply oversold small caps vs heavyweights" — Sichuan Gold, Zhaojin Gold and Western Region Gold are pure gold names and all surged, while Hengbang Co. (gold smelting) fell 1.68%; the gain is highly correlated with the drawdown from the 1/29 high, and the two that fell furthest, Silver Nonferrous (−55.9%) and Hunan Silver (−45.8%), are precisely the only two with a 2nd consecutive limit-up.
2. Only one of the six has silver credentials that survive verification of the revenue mix (Shengda Resources), and it carries a compliance risk entirely unrelated to the silver price.
3. The one with the highest trading visibility (Silver Nonferrous) has the worst fundamentals in the group; the one with the best fundamentals (Shengda Resources) did not even hit the limit. The two are completely dislocated — and that dislocation is itself the evidence that this is not a fundamentals-driven move.

Branch 5 | Semiconductor equipment

---

8. Verification Signals at Today's Open

Auction signals

• Luxshare Precision: a +2%–+4% gap up with volume is ideal. A gap of more than +6% means the market is pricing an "earnings beat" — and since the pre-announcement itself carries little information and reads as deceleration at the floor, the risk/reward of chasing deteriorates fast.
• Eoptolink: ChiNext ±20%, do not read its gap size with a ±10% intuition. A gap of +3% or less that holds = the market is pricing cash-flow quality; a fast fill = it is being treated as "good news fully out."
• Silver Nonferrous: one-word board or a weak open is today's single most important either/or. Sealed one-word → a 3rd board, sentiment has not broken; a gap down, or a gap up that fades → the entire precious-metals line from yesterday's 2nd boards settles up today.
• WuXi AppTec: the size of the gap down is the market's pricing of the cybersecurity incident. A gap down of less than −2% = the negative is seen as manageable; a gap down of more than −5% on volume = the whole CXO sector is untouchable today.
• Wanhua Chemical: a gap up of more than +3% would be mispricing a report that was pre-announced on 7/7 and whose ex-non-recurring landed at the floor — that is a window to reduce, not to buy.

Sector signals

• Precious metals: whether 3 or more names hit the limit quickly. If only Silver Nonferrous seals (a lone survivor), that confirms the ebb — yesterday the whole sector had only 2 genuine limit-ups, and one fewer means the ladder has broken.
• Defensive / dividend: whether Ping An, China Coal Energy and China Merchants Bank can post a second consecutive day of main-force net inflow. A single day of inflow is noise (yesterday's inflow was only 1/36 the size of the electronics outflow).
• Semiconductor equipment: whether Piotech and ACM Research Shanghai can outperform the semiconductor sector for a second consecutive day. That says more about whether the "expansion benefits equipment" mapping holds than whether equipment stocks go up.
• Optical modules: whether Innolight can stop falling. It is the branch's core army, and the branch has no chance until the core army stops falling; Eoptolink strong alone while Innolight keeps falling = a single-stock statement move, not a branch move.
• Coal: watch coking coal futures, not coal stocks. Yesterday the stocks were +2.80% while coking coal futures were only +0.16% — if coking coal futures stay flat today while coal stocks rise again, the divergence widens, which is a late-stage catch-up signal.

Stock-level signals

• Luxshare Precision: whether main-force flow turns from near-neutral yesterday to clearly positive; whether turnover can expand from 1.24% to 2%–4% (but not above 4%).
• Eoptolink: whether it gets blocked by Innolight — Innolight up while Eoptolink is not means money still only recognizes the leader, not the statements.
• China Coal Energy: whether turnover of 0.73% can stay low while money flows in.
• Shanghai Datun Energy: how the day after the limit-up is absorbed. With TTM earnings of only RMB 299 million, any volume expansion without price progress should be treated as a signal to take profit.

Risk signals (if any two appear at once, give up on offense today)

1. Gap-up dive: the three major indices gap up and turn red within half an hour.
2. Lone limit-up: Silver Nonferrous is the only precious-metals name to seal, with a seal/turnover ratio below 0.1.
3. The core army does not follow: Innolight falls more than −3% while Eoptolink also fails to close green.
4. Yesterday's strong themes ebb: the "yesterday's limit-up incl. one-word board" concept posts main-force net outflow again today (yesterday was already −RMB 753mn).
5. Turnover falls below RMB 1.9 trillion (8/24 was about RMB 2.01 trillion) — in a stock-game market, shrinking volume means relay money is withdrawing further.
6. WuXi AppTec gaps down more than −5% and drags CRO into a second leg down, with pharma heavyweights weighing on the index.

---

9. Final Recommendations

① The 5 stocks most worth watching today

Screening rule (carried over from previous reports and tightened here): a recommendation slot goes only to names that satisfy all three of ① a new driver inside the window, ② a driver that comes from the company's own financial basis or verifiable primary material, ③ not priced yesterday. "It hit the limit yesterday" is not a reason to be included; neither is "good fundamentals but no new news." Only 2 names fully qualify in this report, and the bar was not loosened to fill five slots — the other 3 are explicitly labelled "secondary slot, one notch weaker on evidence." The first draft had Wanhua Chemical (600309) in the 5th slot; it was pulled after verifying that a positive pre-announcement was issued on 7/7 and that ex-non-recurring landed at the 13th percentile of the guided range, see §5-10.

② The 3 strongest branches today

③ Directions not worth chasing today

1. The entire precious-metals branch (all six Pass) — this is the most thoroughly rejected item in this report, on six layers of evidence:
   • ① The catalyst is outside the window, and the silver price did not move that day. The US Treasury's doubling of long-end buybacks was announced 8/19 and takes effect 9/9; the price jump happened on 8/20 (SGE silver +5.24%) and 8/21 (+3.29%); on 8/24 itself SGE silver rose only +0.12% and SHFE ag0 +0.43%. The 8/24 limit-ups are a capital relay on day 3 of pricing.
   • ② The overnight night session faded. SHFE silver continuous spiked to 17,072 and closed at 16,775, −0.40% versus the 8/24 day close; SHFE gold −0.09%.
   • ③ "All-time high" is wrong, and especially so for silver. Gold is 16.0% below $5,626.8 (1/29 intraday); silver is 42.5% below $121.78 (1/29 intraday), and on 1/30 it crashed roughly 30% in a single day (a CME margin hike triggered forced liquidation), bottoming at $55.765 on 7/16 for a peak-to-trough drawdown of −54%. What is happening now is a +25.5% bounce off a deep hole, not a breakout.
   • ④ Two of the names treated as "silver stocks" are not silver companies at all. Silver Nonferrous' own-mine output contains no silver, electrolytic silver is only 6.55% of revenue, cathode copper is 40.26% at a gross margin of −0.10%, and "Silver" comes from its registered address, Baiyin City, Gansu; Hunan Silver's silver revenue is 70.26% of the total but at a gross margin of only 5.05%, smelting externally purchased feed, with own mines at only 2.94%. The other three (Sichuan Gold, Western Region Gold, Zhaojin Gold) are pure gold names with zero silver.
   • ⑤ The leader lost money in the quarter when gold and silver surged. Silver Nonferrous guided H1 attributable net profit of RMB 64–95mn while Q1 was already RMB 149.98mn ⟹ a Q2 single-quarter loss of RMB 55–86mn; TTM net profit is negative, and the company self-reports a static P/E of −65.69; impairment testing is incomplete.
   • ⑥ The ranking of gains = the ranking of oversoldness, not of fundamentals. All six peaked for the year on 1/29 and bottomed in mid-to-late July, and the two that fell furthest, Silver Nonferrous (−55.9%) and Hunan Silver (−45.8%), are precisely the only two with a 2nd consecutive limit-up; while the only one that passes on fundamentals, Shengda Resources (mined silver 62.14% of revenue, gross margin 78.37%, TTM P/E 32.3×, lowest in the group), did not even hit the limit, and the company and its actual controller have been under CSRC investigation for suspected disclosure violations since 6/15, still unresolved.

   In one sentence: the strongest name on the tape has the worst fundamentals, and the best name on fundamentals did not hit the limit and carries an open investigation — that dislocation is itself the proof that this is not a fundamentals-driven move.

2. The limit-up coal stocks (Meijin Energy, Dayou Energy) — this is the item this report most wants to block. Meijin Energy (000723) is a coking company, downstream of coking coal: coking coal is +46.6% YTD while coke is only +30.6%, so the coking spread is compressed and the price rise is a cost negative for it; the company's 7/15 pre-announcement guided an H1 loss of RMB 450–650mn after two full consecutive loss-making years. Dayou Energy (600403) reported attributable net profit of −RMB 643mn on the evening of 8/24, attributable net assets −16.01% over the half year, gearing 80.94%, operating cash flow turned negative, and P/B 5.66× is the most expensive of the three limit-up coal names; moreover with coking coal +46.6% its gross margin is only 7.37%, showing its coal grade / cost structure did not benefit at all.
3. Memory design and modules (GigaDevice, Longsys, Biwin Storage) — the YMTC IPO is negative for them, not positive. What the market prices is future supply growth: the same day Micron and SanDisk fell notably in the US on "another Chinese memory maker filing for an IPO in Shanghai," and these A-share names were all hammered too (−4.14% to −6.70%). Expansion benefits equipment, not memory prices.
4. CXO / innovative drugs (WuXi AppTec and its chain) — the bad news is not fully out and the information is incomplete. The DEL platform investigation is still under way, and "no evidence has been found that any material was made public or misused" is a statement that characterizes nothing yet. On 8/24 the CRO concept already saw main-force net outflow of RMB 2.939bn, monoclonal antibodies −RMB 2.072bn and AI pharma −RMB 1.932bn, while WuXi AppTec's −4.31% happened before the filing — this negative has not been priced.
5. Buying Wanhua Chemical as an "earnings beat" — a positive pre-announcement was already issued on 7/7 (attributable RMB 9.8–10.4bn, ex-non-recurring RMB 9.6–10.2bn), and the actual ex-non-recurring of RMB 9.676bn lands at the 13th percentile of that range; Q2 single-quarter operating cash flow was −50.5% YoY; the company itself says polyurethane prices "rose then fell," so Q2 was the top; and polymeric MDI, its largest-volume product, still had H1 average prices down −1.2% YoY.
6. T&S Communications (300394) — most expensive + the only one stalling, both true at once: TTM P/E 117.2× and P/B 43.56 are the most expensive on the whole list, and Q2 revenue was −0.9% YoY (Q1 was still +40.7%). The interim report was disclosed on 8/19 and is not new news.
7. Yesterday's sentiment-driven consecutive boards (Chutian Dragon, Eastcompeace, Nuode Investment) — the "yesterday's limit-up incl. one-word board" concept saw main-force net outflow of RMB 753mn on 8/24 and "theme stocks" RMB 6.886bn; relay money is a net outflow. Chutian Dragon's P/E (dynamic) of −132.1 means it is loss-making.
8. Holding a heavy compute-chain position before 8/27 — NVIDIA reports after the US close on 8/26, so the A-share reaction day is 8/27 (Thursday), and NVIDIA itself was already −2.91% on 8/24. Today's and tomorrow's compute-chain trades are, in essence, prepaying a premium for an event two days out.

④ Final judgment in one sentence

There are only two and a half genuinely new items today, and they point to opposite actions: Luxshare Precision is worth a deep dive, but what you are buying is an ex-non-recurring number depressed by an FX presentation mismatch and the lowest 23× TTM valuation on the list — not a Q3 pre-announcement whose range spans 34 percentage points and which, on the company's own 2026 record across three pre-announcements, should be read at the floor; Eoptolink has the cleanest statements in the group, but its interim report carries no expectation gap in magnitude, and NVIDIA is two days away. As for the best performers of yesterday — of the two "silver stocks" that sealed a 2nd board, one produces no silver at all from its own mines and lost money in the second quarter, and the other's silver gross margin is only 5%; of the three limit-up coal stocks, one is, in supply-chain terms, a victim of this price rise; and the catalyst itself was announced on 19 August and does not take effect until 9 September, while on 24 August the domestic silver price rose all of 0.12%.

When the strongest name on the tape in a sector has the worst fundamentals, and the best name on fundamentals did not hit the limit and carries a CSRC investigation, the right action is not to reach for a third board — it is to keep the position for 27 August.

---

Self-discipline record for this report (for review comparison):

1. Own yesterday's (8/24) calls first: the top recommendation Shengyi Technology fell −4.89% that day, and the whole Top 5 declined (WUS Printed Circuit −4.65%, Naura Technology −0.91%, Innolight −7.72%, Shennan Circuits −4.08%); while the precious-metals group explicitly Passed yesterday produced 2 genuine limit-ups. Yesterday's fundamental analysis was not wrong; the error was treating "financial quality" as a same-day ranking device. This report therefore raises fund flow and position to the same weight as fundamentals (the ranking basis in §7 is entirely capital structure and seal quality).
2. Adding one thing missed yesterday: when ranking Shengyi Technology first on "the only one with improving cash flow," TTM P/E was not put on the same table for comparison — at 59.1× it is above Eoptolink's 43.8× and Innolight's 49.8×. Recorded in §5-9.
3. Three first-draft judgments were overturned before finalization (all triggered by sub-agent verification):
   • Wanhua Chemical was pulled from the 5th recommendation slot: verification found a positive pre-announcement already issued on 7/7 (attributable RMB 9.8–10.4bn, ex-non-recurring RMB 9.6–10.2bn), with the actual ex-non-recurring of RMB 9.676bn landing at the 13th percentile of the range; plus Q2 single-quarter OCF −50.5% YoY and the company's own statement that prices "rose then fell." The first draft wrote it as "first disclosure date to be verified" and still gave it a recommendation slot — that is interpreting uncertainty in the favourable direction, and it was the closest this report came to issuing a wrong recommendation.
   • Luxshare Precision's core thesis was rewritten wholesale: the first draft wrote the Q2 ex-non-recurring figure of −0.1% as "deteriorating profit quality," when it is in fact an FX presentation mismatch (RMB 1.986bn of losses inside the ex-non-recurring line, RMB 1.297bn of hedging gains outside it), voluntarily disclosed in the body of the interim report and verifiable in the finance-cost note (FX gains/losses deteriorated RMB 2.479bn YoY while net interest expense improved RMB 105mn). The direction was inverted, and it has been corrected throughout.
   • Shaanxi Coal was cut from recommendation candidate to watch only: guided attributable growth of +47%–+53% but ex-non-recurring of only +31.76%–+38.10%, with the roughly RMB 1.7bn gap coming from "investment income on selling listed-company shares held"; and Q1 attributable was still −12.38%, with the interim report not yet disclosed.
4. Four counter-arguments running against the recommendation were given for Luxshare at the same time, to avoid a one-way narrative: all three 2026 pre-announcements landed at or below the low end (H1 ex-non-recurring broke its own floor by 4.7% with no correction issued), TTM free cash flow −RMB 2.109bn, Leoni entering the YoY base from 2026Q3, and the AI segment at only 9.52% of revenue with gross margin −2.48pp. Read at the floor, Q3 attributable growth of +10.85% is a continued deceleration — that directly conflicts with the recommendation rationale, but it has to be written.
5. One widely circulating false fact was falsified: several Chinese closing wraps claimed "COMEX gold hit an all-time high on 8/24." Cross-confirmed as wrong by three sources — the Forbes original ("15-Week High"), IndexBox ("still below the all-time highs set earlier this year"), and the historical highs themselves ($5,589–5,600 set on 2026-01-28; silver $121). This is the main basis on which this report blocks chasing.
6. When two bases conflict, no averaging: COMEX new session had gold +0.50% / silver +0.89%, while the SHFE gold/silver night sessions were −0.09% / −0.40%. The night session was explicitly chosen with the reason stated, rather than fudging it with a midpoint.
7. Limit-ups were price-verified stock by stock rather than judged by percentage move: Sichuan Gold's +9.97% looks like a limit-up, but it closed at 56.60 against a limit price of 56.62 — 2 cents short of sealing, and not in the official limit-up pool. Zhaojin Gold and Western Region Gold likewise. The so-called "silver limit-up wave" was in fact 2 names.
8. Both the limit-up and limit-down pools were verified with tc == len(pool) (46=46, 11=11). The first request for the limit-down pool returned tc=11 with an empty pool, and was successfully re-fetched using sort=fund:asc.
9. Market breadth used two-way pagination merged and de-duplicated, covering 5,545/5,901 = 94.0%, avoiding the tail bias of one-way sorted pagination; the result was 1,460 up / 3,965 down / 120 flat, corroborating the media's "nearly 4,000 decliners." The 356 not covered are all in the middle of the distribution; both tails are complete.
10. All single-quarter figures were split out by this report from cumulative differences, and labelled "single quarter" in the tables. Luxshare's Q2 ex-non-recurring of −0.13%, T&S's Q2 revenue of −0.9% and Wanhua's Q2 OCF of −50.5% are all invisible on a cumulative basis.
11. One basis trap was caught and written into the body: because of a common-control merger China Shenhua has retrospectively restated the prior-year comparative (RMB 24.641bn → RMB 27.583bn), so the widely quoted "+6.9%–+21.1%" compares the new statements against the old base; restated, it is actually −4.7%–+8.0%, broadly flat or slightly down.
12. One counter-intuitive supply-chain direction was written into the do-not-chase list: Meijin Energy is a coking company, downstream of coking coal, so a coking-coal price rise is a cost negative for it (coking coal +46.6% YTD vs coke +30.6%, spread compressed) — and it hit the limit on 8/24. Following the "coal prices up → coal stocks benefit" intuition alone would get the direction backwards. Same family: the YMTC IPO is negative for memory design and positive only for equipment.
13. A "commodity vs equity" dimension was added to the coal branch: on 8/24 coal stocks were +2.80% and the coke sector +3.64%, while coking coal futures were only +0.16%, coke futures +1.56%, and thermal coal CCI5500 flat for four consecutive readings. The equities are running ahead of the commodity, so it is characterized as catch-up plus defensive rotation rather than new fundamentals that day, with the verification point set as "watch coking coal futures, not coal stocks."
14. Three out-of-window items were explicitly downgraded rather than treated as new catalysts: the US-Canada tariffs (effective 8/22), the DRC copper-cobalt ban (signed 6/29, week 8 of pricing), and T&S Communications' interim report (8/19). The pre-announcement dates for Shaanxi Coal (7/11), Shenhua (7/15), Yankuang (7/15), Meijin (7/15) and Dayou (7/14) are all labelled too.
15. A second overturning before finalization: the whole precious-metals branch went from "downgrade" to "all six Pass," the second directional correction in this report.
    • The first draft listed Silver Nonferrous (601212) as the one name in the group "leading on all four measures on a relative basis" and gave it "watch only", on tape-based grounds (2nd board, seal ratio 0.208, turnover 2.95%, Dragon-Tiger +RMB 180mn). Verifying the revenue mix revealed it is a copper smelter — the 2025 annual report's own-mine metal output contains no silver, electrolytic silver is only 6.55% of revenue, cathode copper is 40.26% at a gross margin of −0.10%, and "Silver" comes from its registered address, Baiyin City, Gansu Province; plus H1 guidance of RMB 64–95mn vs Q1 already at RMB 149.98mn ⟹ a Q2 single-quarter loss. The first draft treated "I have not verified the revenue mix" as an item that could be scored first and filled in later, when it was in fact the entire story of this stock — the second closest this report came to issuing a wrong conclusion (the first being Wanhua Chemical).
    • A misjudgment in the opposite direction was corrected at the same time: the first draft Passed Shengda Resources (000603) on the single ground of "Dragon-Tiger net buy −RMB 211mn." On verification it is the only genuine silver mining equity in the group (mined silver 62.14% of revenue and 71.51% of mining/dressing profit, silver concentrate gross margin 87.08%, operating cash flow 1.47× attributable, TTM P/E 32.3×, lowest in the group). It is still a Pass, but for a completely different reason: the company and its actual controller Zhao Mantang were both placed under CSRC investigation for suspected disclosure violations on 2026-06-15/16, still unresolved. Passing a stock for the right reason and passing the same stock for the wrong reason have completely different value at review time.
    • The catalyst's chronology was corrected too: the US Treasury's doubling of long-end buybacks was announced 8/19 and takes effect 9/9 (the first draft, following media paraphrase, recorded it as an overnight 8/24 event — it is outside the window); and on 8/24 SGE silver rose only +0.12%, so that day's limit-ups had nothing to do with the silver price and were day 3 of pricing the 8/20–8/21 move.
    • Silver's drawdown depth was corrected as well: the first draft wrote "about 42% below $121"; verification added the mechanism — after peaking at $121.78 on 1/29, it crashed roughly 30% in a single day on 1/30 (a CME margin hike triggered forced liquidation), bottomed at $55.765 on 7/16, a −54% peak-to-trough. It is currently a +25.5% bounce off a deep hole. "42% below the high" and "just went through a halving-style crash" are two completely different risk descriptions.
    • The other three (Sichuan Gold, Western Region Gold, Zhaojin Gold) were verified as pure gold names with zero silver, and their 8/24 moves were collateral sector drift; of these, Western Region Gold's self-produced gold is only 7.3% of revenue with the rest being purchased-gold pass-through at a 2.37% gross margin, and Zhaojin Gold's ex-non-recurring / attributable is only 0.46 (RMB 135mn being a one-off debt-restructuring gain) with P/B 23.01.
16. One methodological takeaway, recorded: the real dividing line in precious metals on 8/24 was "the drawdown from the 1/29 high," not "gold or silver" — the two that fell furthest are precisely the only two 2nd boards, while pure gold names surged just as hard and a gold smelter (Hengbang Co.) actually fell. Framing the day as "gold vs silver" makes the tape impossible to explain.
17. The Abnormal Share Price Movement Announcements from three precious-metals names pre-market on 8/25 are new in-window information, and are in the news overview (#4b); in its filing Silver Nonferrous self-reports a static P/E of −65.69 and a dynamic P/E of −88.40, which is the company's own corroboration of this report's judgment that East Money's dynamic P/E of 82.70 is wrong. Western Region Gold's 8/25 pre-market interim report + impairment (−RMB 45.27mn) is likewise new in-window information (#4c).
18. Two incomplete verifications are explicitly acknowledged, and no conclusions are drawn from them:
    • Luxshare Precision's FY2025 customer concentration (largest customer 56.68%, top five 65.04%) — this relies on search summaries and the internal consistency of a revenue back-calculation; the annual report text was not obtained, and it is marked "to be verified" in §5-1.
    • The time series for Qinhuangdao 5,500-kcal thermal coal — multiple search sources give mutually contradictory historical prices (RMB 995 in 2026-01 versus the RMB 638 average price for 5,000-kcal in 2026Q1 disclosed first-hand by Wanhua, at −0.31% YoY, which are incompatible; and "RMB 854–861 in 2025-08" versus "RMB 860 in 2026-08" are nearly identical, suggesting an evergreen page with the year swapped). The "year to date" and "year on year" changes for thermal coal are therefore marked as unconfirmable and not used; coking coal and coke use futures series instead, which carry explicit timestamps.
19. The recommendation slots were not padded to five: only 2 fully satisfy the three conditions, and the other 3 are explicitly labelled "secondary slot" with the specific condition each fails. §7 Branch 1 likewise explicitly refuses to pad the list.
20. WuXi AppTec's filing is dated 25 August 2026 (not 8/24) and has been labelled by its filing date; the report also notes that its −4.31% on 8/24 happened before the filing, i.e. the negative is not yet priced.
21. Data-fetch failures: push2.eastmoney.com and 82.push2 timed out intermittently, so push2delay.eastmoney.com was used as the primary host throughout with multi-host fallback; the Cailianshe newsflash API (nodeapi/updateTelegraphList) returned 404 and the signature workaround also failed, so that source was not obtained this time, and the news coverage instead used the exchange filing feeds (24 pages scanned page by page, 207 hits) + WebSearch; stooq's XAUUSD/XAGUSD endpoints failed, so the gold and silver all-time highs were cross-confirmed via Forbes/CBS/StoneX; East Money's commodity K-line endpoint (101.GC00Y) returned empty, so real-time snapshots + the Kitco web page were used; on the sub-agent side the East Money sector and spot endpoints all returned RemoteDisconnected, so "coal sector main-force net inflow of RMB 853mn" was not independently verified, and this report does not use it as a basis for any ranking.

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