A-Share Pre-Market Brief | 2026-08-26 Wednesday

Scan window: 2026-08-25 15:00 (previous session close) → 2026-08-26 07:30 (Beijing time) Previous session: 2026-08-25 Tuesday. SSE Composite 3889.44 (+0.19%), SZSE Component 13745.87 (−0.35%), ChiNext Index 3397.52 (−1.00%), STAR 50 1604.59 (+0.14%), CSI 300 4552.03 (−0.24%); combined turnover RMB 1.83 trillion (contracting). Official limit-up pool: 65 names (Eastmoney ZTPool tc=65, pool length 65, verified consistent); 9 consecutive limit-ups, 22 broken seals, seal rate 75% (CLS definition, ex-ST).

---

0. One-Sentence Summary of Today

1. The entire trading day happens before NVIDIA's earnings. NVIDIA's FY2027Q2 results are set for after the US close on 8/26 (conference call 17:00 ET = 8/27 05:00 Beijing time), so the A-share reaction day is tomorrow, 8/27, not today. Today's compute-chain trading is, in essence, prepaying a premium for an event that only reveals itself tomorrow morning.
2. The new news inside the window is overwhelmingly interim reports, and the "stunning growth rates" in this batch are for the most part old news already announced in early July. I traced each one back to its first-disclosure date: 沪电股份 (WUS Printed Circuit, 002463), 盐湖股份 (Qinghai Salt Lake Industry, 000792), 雅化集团 (Yahua Industrial Group, 002497), 杭电股份 (Hangzhou Cable, 603618), 欧科亿 (OKE Precision Cutting Tools, 688308), 当升科技 (Beijing Easpring Material Technology, 300073), 江西铜业 (Jiangxi Copper, 600362), 招商证券 (China Merchants Securities, 600999), 高德红外 (Guide Infrared, 002414) all issued earnings pre-announcements between 7/3 and 7/17. Before scoring any expectation gap, you must first ask "on what day did this number first appear?"
3. This brief's first draft put 江西铜业 (600362) in the top recommendation slot as the "only company to beat the upper bound of its forecast"; that has been overturned after verification. The 7/9 pre-announcement gave two ranges, not one: net profit attributable to parent RMB 7.550–8.500 billion, ex-non-recurring RMB 7.630–8.580 billion. Actual parent net of RMB 8.632 billion did beat the parent-net upper bound by 1.55%, but ex-non-recurring came in at only RMB 7.556 billion, below the lower bound of the ex-non-recurring range. The difference comes from +RMB 1.076 billion of non-recurring gains, the largest single item being RMB 673 million from the book revaluation of its own pre-existing 12.14% stake at the time of the SolGold acquisition — zero cash flow. And since net non-recurring gains are by definition "parent net − ex-non-recurring", stripping them out yields exactly the ex-non-recurring figure of RMB 7.556 billion — which sits just barely above the parent-net forecast floor of RMB 7.550 billion, yet clearly below the ex-non-recurring forecast floor of RMB 7.630 billion. This is not a "beat"; it is a fork of "parent net above the ceiling, ex-non-recurring below the floor" — full reasoning in §5-2.
4. The only heavyweight, entirely unpriced piece of news in the window points downward: 泸州老窖 (Luzhou Laojiao, 000568) H1 parent net −43.37%, revenue −36.35%, and because the decline did not reach 50% it never issued a pre-announcement — the interim report is the first disclosure, and on 8/25 the stock fell only −0.91%. Baijiu and broader consumer names need a separate defence today.
5. Yesterday's strongest agriculture theme forked overnight along two legs: the grain/planting leg (金健米业 (Jinjian Cereals Industry, 600127) +69.0% over 20 days) is deeply overextended yet got overnight confirmation (CORN +1.93%, SOYB +1.31%); the fertiliser leg sits far lower yet got overnight rejection (NTR −1.95%, CF −1.53%). The leg that has already run the most was confirmed, and the leg that has not run was rejected — that combination is not suitable as a fresh entry direction today.
6. Driver types: earnings (primary) + overseas industry (secondary) + capital structure (secondary). The policy side is empty within this window — the "8/25 Shanghai property policy" I checked turns out to be the equivalent 2025-08-25 document, while the 2026 "Shanghai Eight Measures" was released 8/20 and effective 8/21, both outside the window (see "three news items falsified by this brief" in §1).
7. Pre-market state assessment: overseas markets were mildly warm overnight (Nasdaq +0.66%, Philadelphia Semiconductor +1.44%, NVIDIA +2.19% ending a seven-day losing streak), but the A-share market's own follow-through is weak (the 46 limit-up names of 8/24 closed on average +1.03% on 8/25, with only 43% finishing green). Better external environment + weak internal follow-through means today favours hunting among names that were released post-close and carry zero pricing on the day, rather than chasing the consecutive limit-up ladder.

---

1. News Overview

1.1 New news within the window (2026-08-25 15:00 → 08-26 07:30)

1.2 Out-of-window but still-being-priced carry-over themes (pricing progress noted; not counted as "new news")

1.3 Three "news items" explicitly falsified or downgraded by this brief

If copied verbatim, all three would directly corrupt today's judgement. Evidence is listed item by item.

① "On 25 August six Shanghai departments released a new property policy" — mistaken identity; it is in fact the equivalent 2025 document. Search results took a notice from the Shanghai Municipal Housing Administration Bureau at fgj.sh.gov.cn/gzdt/**20250825**/... dated 25 August 2025 and treated it as a new policy of 25 August 2026. I pulled up the 2026 original directly: "Notice on Optimising the Municipality's Real Estate Policy Measures", document number Hu Jian Fang Guan Lian [2026] No. 359, dated 20 August 2026, explicitly stating "effective from 21 August 2026". Shanghai released no new real estate policy on 8/25; the limit-ups along the property chain are day 3–4 of pricing the "Eight Measures", not a fresh catalyst. Primary source: Shanghai Municipal Housing Administration Bureau⁸

② "US tech under pressure on 25 August, Nasdaq −0.76%, NVIDIA −2.91%" — that report actually describes 24 August. A report headlined "close of 25 August" says: Dow +0.26% at 53417.16, S&P −0.28% at 7652.86, Nasdaq −0.76% at 25980.19. But in CNBC quote data, the previous close corresponding to the 8/25 session is exactly those three numbers (53417.16 / 7652.86 / 25980.19) — they are the 8/24 closes. The real 8/25 closes were Dow 53577.40 (+0.30%), S&P 7677.28 (+0.32%), Nasdaq 26151.30 (+0.66%), and NVIDIA +2.19% at 213.05 (not −2.91%). The direction is exactly reversed.

Note: the figure −2.91% does exist, but it belongs to 8/24. The whole narrative is one trading day stale.

③ "百傲化学 (603360) is an agrochemical beneficiary" — its core business has nothing to do with agriculture, and its largest revenue line is already semiconductor equipment. Eastmoney's industry classification places it in "agrochemical products", and on 8/25 it did indeed limit up along with that concept. However: the company's core business is isothiazolinone industrial biocides (used in coatings, water treatment and other industrial settings — not pesticides); and in 2025 its semiconductor business (Xinhuilian) generated RMB 903 million of revenue, already exceeding the RMB 821 million from chemicals. The chain "El Niño → grain output cuts → agrochemical demand" does not transmit to it. This is a classification-label-driven limit-up, not a fundamental beneficiary. Yicai⁹ / CLS¹⁰

---

2. Strongest Catalyst Branches, Descending

Branch-by-branch Notes

Branch 1 | PCB / high-speed switches — the increment is not in total profit, it is in the slope of gross margin

Why it is a positive: the total profit in WUS Printed Circuit's interim report contains no expectation gap (parent net landed at the 54.9% percentile of the forecast range, ex-non-recurring at only the 43.0% percentile, in the lower half), but what was not in the pre-announcement is the increment:

• Across-the-board Q2 standalone acceleration: revenue +67.75% (Q1 +53.91%, accelerating 13.8pp), ex-non-recurring +81.60% (Q1 +56.02%, accelerating 25.6pp). The 7/14 pre-announcement gave only an H1 profit range and no revenue, so the market could not derive the +67.75% revenue line in advance.
• Gross margin slope break: Q2 standalone overall gross margin 41.27%, +5.64pp QoQ. The previous three quarters had been in a narrow 35.6%–35.8% range (25Q3 35.83%, 25Q4 35.67%, 26Q1 35.63%) — Q2 is an inflection point, not a continuation.

Metric note: the earlier 25Q2 figure was 37.31%, so the YoY change in the 41.27% is only +3.96pp, less striking than the +5.64pp QoQ. Please cite both numbers together.

• Business mix quantified for the first time: data communications totalled RMB 11.330 billion (82.77% of revenue, +73.46%); within that, high-speed switches and supporting routing RMB 7.139 billion (52.15%), AI servers + HPC RMB 1.828 billion (13.35%), general servers RMB 2.046 billion (14.95%). PCBs of 32 layers and above +190.83% YoY; 1.6T switches already shipping in volume, 224Gbps in small-batch supply.

One thing that must be spelled out: its main engine is high-speed switches, not AI servers. AI server + HPC revenue of RMB 1.828 billion is below the RMB 2.046 billion from general servers, accounting for only 13.35% of total revenue. Treating WUS as an "AI server PCB play" is a mismatch; the accurate positioning is "leader in high-speed network switch PCB". This directly affects its true sensitivity to NVIDIA's results.

Is it already fully expected: the profit line was priced once on 7/14 and then given back — the stock limit-up +10.00% on the pre-announcement day (124.66→137.13), and by the 8/25 close of 114.29 it had given back −16.66%, fully closing the limit-up gap. It is −5.55% over the last 10 trading days, and −1.11% on 8/25 with a turnover rate of only 2.14%. This is not an overextended-at-highs pattern; it is a continued give-back after a pre-announcement.

Risks: cash-to-profit ratio 0.39 (1.25 in the prior-year period), operating cash flow −45.14%, capex +159.46%, free cash flow −RMB 2.451 billion; gearing 52.20% (+5.74pp), interest-bearing debt up from RMB 4.556 billion to RMB 8.315 billion (+82.5%) in half a year; 64.6% of the RMB 128.8 million of non-recurring gains comes from related-party equity disposals and reversals of receivables impairment; the smart-vehicle segment grew revenue +1.11% (essentially zero growth) with gross margin −6.22pp. Two capacity-expansion projects have also been adjusted because land was not delivered on schedule or site selection was deferred.

Branch 2 | Copper / non-ferrous — operations really are accelerating, but the "beat" claim must be withdrawn

First, a retraction of an error in this brief's first draft. The first draft described Jiangxi Copper as "the only company in this batch to beat the upper bound of its forecast" and placed it in the top recommendation slot. After verifying the primary source, that claim does not hold.

The 7/9 pre-announcement (CNINFO Lin 2026-026 original¹¹) gave two ranges side by side:

The first draft only checked the parent-net range and missed the ex-non-recurring range printed on the same page. This is the point in this brief most in need of self-correction — it is not a data error but selective citation, and the half that was missed sits in the very same document.

Where the difference comes from: non-recurring gains of +RMB 1.076 billion (the company's own estimate at the time of the pre-announcement was −RMB 80 million, a swing of RMB 1.156 billion). Item by item:

The correct way to strip it out: net non-recurring gains of RMB 1.076 billion are by definition "parent net − ex-non-recurring", so removing them yields exactly the ex-non-recurring figure of RMB 7.556 billion — only RMB 6 million above the parent-net forecast floor of RMB 7.550 billion, yet clearly below the ex-non-recurring forecast floor of RMB 7.630 billion. (If instead you subtract the pre-tax gross amount of RMB 1.119 billion from the three disposals, you get roughly RMB 7.51 billion; neither convention should be written as "about RMB 7.4 billion".) Conclusion: the parent-net "beat" is entirely carried by non-recurring items, and once stripped out the company barely reaches even the floor of its own parent-net forecast. As for that RMB 673 million of "gains", the economic substance is: the company paid RMB 6.967 billion in cash for an 87.86% stake in SolGold, took on an undeveloped exploration asset that lost RMB 112 million net and burned RMB 37 million of operating cash flow during the reporting period, and then revalued its own old position on paper.

Why ex-non-recurring fell below the floor: the "Asset Impairment Provision Announcement" filed on the same day gives the answer — H1 impairment provisions totalled RMB 2.905 billion (of which inventory write-downs were RMB 2.494 billion), with an impact on parent net of −RMB 1.621 billion. When the company made its pre-announcement in early July it most likely had not fully accounted for this round of impairment.

But the positive evidence must be given just as fairly — operations really are accelerating, and this part cannot be read out of the pre-announcement:

Q2 standalone ex-non-recurring +136.4% and gross margin going from 3.56% to 6.44% exclude one-off gains — that part is real. The issue is not "whether there is improvement" but that the figure of RMB 8.632 billion cannot be used to measure that improvement.

Position in the value chain (this changes the narrative):

• Captive mine self-sufficiency is only 7.9% (captive ore 100,000 tonnes / cathode copper 1.2604 million tonnes). The interim report's "copper contained in self-produced concentrate +37.36%" is a changed definition — the numerator includes 36,400 tonnes of First Quantum attributable output while the denominator (99,300 tonnes last year) does not; on a like-for-like captive-mine basis it is 100,000 vs 99,300 tonnes, +0.7%, essentially flat (that footnote is printed right below the output table).
• Subsidiary net margins: mining 34.1%, copper smelting 1.39%, gold smelting 1.07%, trading 0.20%–0.48%; overall net margin on sales is only 3.09%, weighted ROE 10.22% (Zijin 19.60%, CMOC 18.29%).
• Copper concentrate treatment charges TC/RC are at historic extreme negatives: the company states a H1 low of −US$124.45/tonne; Guosheng Securities, citing SMM data, puts the imported copper concentrate index at −US$159.37/tonne on 7/31, implying theoretical smelting profit of −RMB 1,588/tonne on spot TC/RC and +RMB 4,497/tonne on long-term contracts.

That is: Jiangxi Copper is not a "mining beta play on rising copper prices" but a "processor operating in a high copper and high gold/silver price environment". Its moat is its long-contract share, by-product recovery and scale — not resource self-sufficiency.

Other negatives: Q2 standalone operating cash flow of only RMB 306 million (−86.8% YoY, quarterly cash-to-profit ratio 5.3%); interest-bearing debt up RMB 65.1 billion in half a year to RMB 140.4 billion; no interim dividend (RMB 4 per 10 shares in the prior-year period).

Valuation (the one uncontested cheap item): PB 1.74x vs Zijin 4.69x and CMOC 4.39x; still roughly a 29% discount after PB/ROE adjustment; A/H premium 41.45% (H shares at HKD 36.30, read at the 8/25 16:08 Hong Kong close). But its PE of 13.19x is nearly identical to Zijin's 13.24x and CMOC's 14.17x while its ROE is only half theirs — on a PE basis it is the most expensive of the three.

The one clean point on timing: at the 8/25 close the market had genuinely not seen this interim report (CNINFO records the disclosure date as 2026-08-26, board meeting 8/25), so the −3.92% that day was not a reaction to it (SHFE copper was +0.06% the same day, the non-ferrous sector −2.78%, peers broadly down). The interim report is entirely new information for the 8/26 open — it is just that this new information cuts both ways.

Branch 3 | Optical fibre & cable — the group with the strongest branch and the weakest stocks

On 8/25 the "communication cable and accessories" sector rose +4.46% with RMB 4.389 billion of net main-force inflow, making it a sector with both top gains and top inflows market-wide; Hengtong Optic-Electric had net buying of RMB 1.850 billion on the block-trade board, with very high-quality buy-side seats: Shanghai-Hong Kong Stock Connect dedicated seat RMB 1.089 billion, institutional dedicated seats RMB 656 million, CICC Shanghai RMB 421 million, Goldman Sachs (China) RMB 276 million, J.P. Morgan (China) RMB 221 million — northbound + institutions + foreign investment banks, not hot money.

But a strong branch does not mean the stocks are buyable:

• 亨通光电 (Hengtong Optic-Electric, 600487) was already +6.90% on 8/25 and has no company-level news in the window (I checked the announcements disclosed on 8/26: 亨通光电, 长飞光纤 (YOFC, 601869) and 中天科技 (ZTT, 600522) have all not released interim reports);
• 杭电股份 is the only name in this group with in-window new information, with H1 parent net of RMB 393 million landing at the 82.5% percentile of its forecast range (the second highest in this batch). But on the same day it also filed an "H1 2026 Asset Impairment Provision" announcement (amount could not be verified), on top of the private A-share placement plan (dilutive) disclosed in July and a share reduction plan by shareholders, directors and senior management; moreover the headline +938.67% growth is itself a product of the low base created by the RMB 362 million impairment provision taken in 2025 (the 2025H1 base was only RMB 37.81 million).
• In the same sector, 通鼎互联 (Tongding Interconnection, 002491) had net selling of RMB 455 million on the block-trade board on 8/25 with a turnover rate of 24.28% — the branch is already differentiating internally.

Branch 5 | Grain / planting — confirmed overnight, but the price level no longer permits it

Overnight CORN +1.93%, SOYB +1.31% and WEAT +0.59% all moved the same way and are internally consistent, confirming the output-cut trade is still running (using three commodity ETFs to cross-check each other avoids the staleness of a single text source).

But the A-share leg is in a very poor position: 金健米业 (Jinjian Cereals Industry, 600127) is +69.0% over 20 days, with a turnover rate of 46.17% on 8/25 and a sealing order of RMB 29 million against RMB 2.715 billion of turnover = a ratio of just 0.011 (an extremely thin seal); 登海种业 (Denghai Seeds, 002041) is +33.0% over 20 days and 万向德农 (Wanxiang Doneed, 600371) +34.8%. This is already trading day 3–5 of pricing, and intermittent patterns such as "5 limit-ups in 7 days" or "3 limit-ups in 6 days" themselves indicate widening disagreement.

Treatment in this brief: acknowledge that the overnight confirmation is real, but give it no recommendation slot. The reason is the price level, not the logic.

Branch 7 | Fertiliser / agrochemicals — the branch this brief rejects most completely, with three independent pieces of evidence

① Overseas fertiliser stocks fell as a group overnight, forking away from the grain leg. NTR −1.95%, CF −1.53%, IPI −0.21% (MOS +1.12% the only gainer), while grain ETFs rose across the board over the same period. On the same night, grain was bought and fertiliser was sold.

② 盐湖股份's interim report directly falsifies the "potash price increase" logic. This is the hardest piece of evidence, because it comes from the company's own financials:

• Realised potassium chloride prices were only +5.2% YoY (potash product revenue / potassium chloride sales volume: RMB 3,019 → 3,176 per tonne);
• The 2026 annual import large contract is US$348/tonne CFR versus US$346 in 2025, +0.6% YoY — the price-increase dividend was already fully realised by mid-2025;
• Potassium chloride output fell −15.5% YoY (1.9898 → 1.6817 million tonnes) while sales volume rose +26.4% — within the potash segment's +31.4% revenue growth, volume contributed +26.4% and price only +5.2%, so roughly 80% came from destocking, and that 565,700 tonnes of destocking is a one-off;
• The company states in its interim report: "Since May 2026 the domestic market has entered the fertiliser off-season, port potash inventories have kept building, and market transaction prices for potash have gradually weakened", and warns that "many planned potash capacities may come online successively starting in the second half of 2026".

③ Of the two names treated as fertiliser stocks, one is not a fertiliser stock at all and the other is already a lithium stock.

• 百傲化学 (Bai'ao Chemical, 603360): core business is industrial biocides (not pesticides), and in 2025 semiconductor equipment revenue of RMB 903 million already exceeded chemicals at RMB 821 million (see §1.3-③);
• 盐湖股份 (Qinghai Salt Lake Industry, 000792): in 2026H1, lithium product gross profit of RMB 4.675 billion > potash product gross profit of RMB 4.467 billion; of the RMB 5.036 billion of combined gross profit increment from the two products, lithium contributed 75.9% and potash only 24.1%. And its share price stopped tracking agriculture long ago — since 7/2 (the last trading day before the pre-announcement) it is −9.05%, versus the agriculture ETF at +5.20% and 亚钾国际 (Asia-Potash International, 000893) at +10.96%, underperforming the agriculture ETF by 14.3pp; on 8/25, the day the agriculture theme strengthened, all three potash names fell (Salt Lake −3.62%, Asia-Potash −0.32%, 东方铁塔 (Eastern Tower, 002545) −2.46%).

In one sentence: if the driving logic of the agriculture theme is "fertiliser prices are rising", then that logic is simultaneously rejected at three levels — overseas share prices, domestic financial statements, and sector price action; and of the two stocks that most look like fertiliser plays, one is not a fertiliser company and the other's profits come from lithium.

---

3. Overall Stock-Level Catalyst Strength Ranking (descending by stock)

The first criterion in this table is "on what day did this number first appear". "Forecast percentile" in the table = where the actual value lands within the July pre-announcement range; the percentile itself is neither good nor bad, but "having a pre-announcement" means the magnitude-level expectation gap was already released in July.

Ranking note: this table is sorted descending by total score; 英维克 (002837) and 金健米业 (600127) are tied at 22 pts. 泸州老窖 (000568) is a negative-catalyst name, does not take part in the scoring rank, and is fixed at the bottom of the table. The two downgrades in this table: 江西铜业 fell from the first draft's 76 pts / rank 1 to 58 pts / rank 2 (§5-2), and 英维克 from 50 pts / rank 4 to 22 pts / Pass (§5-4). In both cases verifying primary sources overturned the first draft's conclusion; neither was a fine-tuning.

---

4. Stock Scoring Model (100 points total)

Three scoring exceptions in this brief (different from yesterday, must be explained)

1. "Pre-announcement already issued" directly suppresses the expectation-gap score. The interim reports of 沪电股份 (002463), 盐湖股份 (000792), 雅化集团 (002497), 杭电股份 (603618), 当升科技 (300073), 江西铜业 (600362), 招商证券 (600999) and 高德红外 (002414) were all preceded by a July pre-announcement. The share price already reacted once on the pre-announcement day (沪电股份 limit-up +10% on 7/14), and the interim report merely fills in the numbers. As a result not a single name market-wide scored 8 pts or above on this item: although 江西铜业 (600362) exceeded the parent-net ceiling, its ex-non-recurring fell below the floor, so treating it as a "fork" it gets only 5 pts (rather than 9 pts for a "beat"); the rest get 2–5 pts, and 欧科亿 (688308) and 盐湖股份 (000792) get only 1–2 pts because their ex-non-recurring hugs the floor.
2. "No pre-announcement" adds points, but the release clock must be checked at the same time. The interim report / announcement of 泸州老窖 (000568) and 艾可蓝 (300816) really are first disclosures and not yet priced. But "first disclosure" does not equal "not yet priced", and this item has two counter-examples: ① 英维克 (002837) likewise issued no pre-announcement, yet its interim report went live at 8/24 18:15 and was traded through the full 8/25 session; ② 凯莱英 (002821) likewise issued no pre-announcement, but had already limit-up +10.00% on 8/25 — "disclosed and already priced". Neither gets any points on this item. The direction can also go either way: 泸州老窖 is a first disclosure, but its content is −43.37%.
3. A classification mismatch is penalised straight down to Pass. 百傲化学 (603360) (not agrochemical) and 盐湖股份 (000792) (in reality a lithium stock) are penalised for mismatch regardless of how much they have risen.

---

5. Detailed Analysis of Key Stocks

This section is not ordered by the scores in §3, but by "which names need to be thought through first today": the first 4 are those this brief changed most or where the conclusions carry the most weight, the 5th is the only heavyweight negative, and the last 5 are the ones most easily misread. The numbering is reading order only, not a recommendation ranking.

1️⃣ 沪电股份002463Priority deep-dive | 69 pts · WUS Printed Circuit · SZ main board ±10%

Related news: interim report disclosed post-close on 2026-08-25, with H1 revenue of RMB 13.689 billion (+61.17%), parent net of RMB 2.923 billion (+73.72%), and PCB segment gross margin of 40.52%.

Positive logic (direct, but the entry point is not "beating expectations"): This interim report is not an expectation-gap report; it is a quality-confirmation plus structure-disclosure report. Reasons: the 7/14 forecast range was RMB 2.830–3.000 billion against an actual RMB 2.923 billion, with parent net landing at the 54.9% percentile and ex-non-recurring at the 43.0% percentile (lower half); moreover the actual figures for all four of the company's historical pre-announcements have hugged the midpoint (deviation ≤2%), so forecast precision is extremely high and the ranges given are extremely narrow, meaning that on 7/14 the market could already anchor H1 profit accurately — and the stock limit-up +10.00% that day.

The four things genuinely not yet priced:

1. Q2 standalone acceleration: revenue +67.75% (Q1 +53.91%), parent net +82.69% (Q1 +62.90%), ex-non-recurring +81.60% (Q1 +56.02%). Profit accelerating faster than revenue = operating leverage being realised.
2. Gross margin inflection: Q2 standalone overall gross margin 41.27%, +5.64pp QoQ, breaking the narrow 35.6%–35.8% range of the previous three quarters (+3.96pp YoY). PCB segment 40.52% (+4.06pp).
3. Mix quantified for the first time: high-speed switches RMB 7.139 billion (52.15%) / AI servers + HPC RMB 1.828 billion (13.35%) / general servers RMB 2.046 billion; PCBs above 32 layers +190.83%; 1.6T already shipping in volume, 224Gbps in small-batch supply.
4. FX losses masked the operating improvement: net H1 FX losses of RMB 275.6 million (versus a gain of RMB 34.89 million a year earlier, a RMB 310.5 million YoY deterioration); restated, the operating YoY figure is roughly +90% (reported +73.72%).

Fundamental verification (negatives must be presented equally):

• Cash flow is the biggest negative in this report: operating cash flow RMB 1.151 billion (−45.14%), cash-to-profit ratio 0.39 (1.25 a year earlier), capex RMB 3.602 billion (+159.46%), free cash flow −RMB 2.451 billion. The company attributes this to "a full order book increasing raw material purchases". Receivables +44.7% and inventory +42.3% are both below revenue growth of +61.2%, so the ageing structure itself shows no deterioration signal — this is working capital tied up during an expansion phase.
• Leverage rising quickly: gearing 52.20% (+5.74pp), interest-bearing debt RMB 4.556 → 8.315 billion (+82.5%), net interest-bearing debt +RMB 3.761 billion. No goodwill.
• Quality of ex-non-recurring earnings: ex-non-recurring / parent net is 95.59%, a three-year low; 64.6% of the RMB 128.8 million of non-recurring items comes from a related-party equity disposal (RMB 46.10 million, the transferee being related party Huangshi Lianhong Real Estate) and a reversal of receivables impairment (RMB 37.06 million) — neither is sustainable. But Q2 standalone non-recurring items were only RMB 49 million with ex-non-recurring / parent net at 97.06%, so the low-quality portion is concentrated in Q1.
• The smart-vehicle segment is stalling and bleeding: revenue RMB 1.438 billion, +1.11%, gross margin −6.22pp.
• Thai subsidiary: H1 revenue RMB 671 million, still a loss of RMB 17.53 million; Q2 standalone revenue RMB 375 million with net profit of +RMB 43 million, a genuine turnaround, but it accounts for only 2.56% of the company's Q2 parent net, and this was already spelled out in the 7/14 pre-announcement text — the interim report merely fills in the numbers.

Industry position: leader in high-speed network switch PCB. Not a leader in AI server PCB — AI servers + HPC account for only 13.35% of revenue, below general servers.

Valuation: closed 8/25 at RMB 114.29, total market cap RMB 219.9 billion. PE (TTM) 43.44, ex-non-recurring PE 44.78, PB 12.23. Peers: 生益科技 (Shengyi Technology, 600183) 60.3, 景旺电子 (Kinwong Electronic, 603228) 77.0.

Metric warning: 深南电路 (Shennan Circuits, 002916) (63.9) and 胜宏科技 (Victory Giant Technology, 300476) (51.4) still have 2026Q1 as their latest reporting period, so their TTM denominators do not yet include the interim report and their PEs will fall as those reports are disclosed. "WUS has the lowest PE market-wide" is not a like-for-like conclusion; the only like-for-like comparables are Shengyi and Kinwong.

Technical sentiment: SZ main board ±10%. On 8/25 it was −1.11% with a turnover rate of only 2.14% and turnover of RMB 4.734 billion, opening lower, drifting lower and closing in the lower half of the range. It is −5.55% over the last 10 sessions, −16.66% since the 7/14 pre-announcement limit-up, and −26.26% from its 6/25 high. This is not acceleration at highs; it is a low after a drawdown.

Final judgement: Priority deep-dive. What you are buying is the Q2 gross margin of 41.27% as a slope and the +67.75% revenue line — not an "earnings beat" (on the pre-announcement basis it did not beat, and ex-non-recurring is even in the lower half). The biggest offsetting item is NVIDIA tonight — but note that 82.77% of its revenue comes from data communications, of which switches are the bulk, so its sensitivity to that single company is lower than the market imagines.

---

2️⃣ 江西铜业600362Watch closely (first draft: "priority deep-dive, rank 1"; downgraded after verification) | 58 pts · Jiangxi Copper · SH main board ±10%

⚠️ This is the entry this brief changed most; the conclusion first: the first draft's pivot was wrong and has been withdrawn.

The first draft described Jiangxi Copper as "the only company in this batch to beat the upper bound of its forecast" and put it in the top recommendation slot. After verifying the primary source of the 7/9 pre-announcement: that document gave two ranges side by side, and the first draft only checked one of them.

Metric	7/9 forecast range	Actual	Landing point
Net profit attributable to parent	RMB 7.550 ~ 8.500 billion	RMB 8.632 billion (+106.77%)	Above ceiling +1.55%
Ex-non-recurring parent net	RMB 7.630 ~ 8.580 billion	RMB 7.556 billion (+73.73%)	Below floor −0.97%

Table scrolls horizontally

This is not a "beat"; it is a reverse fork of "parent net above the ceiling, ex-non-recurring below the floor", and what beat is the metric that includes one-off gains while what missed is the metric that represents operations. Ex-non-recurring / parent net = 87.54% (104.19% in the prior-year period).

Source of the difference (fully reconciled within the interim report): non-recurring gains of +RMB 1.076 billion, versus the company's own estimate of −RMB 80 million at the time of the early-July pre-announcement, a swing of RMB 1.156 billion. Within the income statement, the RMB 1.119 billion under "investment income from disposal of long-term equity investments" breaks into three items:

• RMB 673 million — the difference from remeasuring the pre-existing 12.14% stake at fair value at the time of the SolGold acquisition, a pure accounting gain with zero cash flow;
• RMB 381 million — secondary-market reduction of 10.78 million shares of Wanguo Gold;
• RMB 66 million — disposal of the joint venture Nesko.

The correct way to strip it out: net non-recurring gains of RMB 1.076 billion are by definition "parent net − ex-non-recurring", so removing them gives exactly the ex-non-recurring figure of RMB 7.556 billion — only RMB 6 million above the parent-net forecast floor of RMB 7.550 billion, yet clearly below the ex-non-recurring forecast floor of RMB 7.630 billion. The parent-net "beat" is entirely carried by non-recurring items. Moreover, the full interim report offers no explanation whatsoever of "exceeding the forecast ceiling".

Why ex-non-recurring fell below the floor: the "Asset Impairment Provision Announcement" filed the same day shows H1 impairment provisions totalling RMB 2.905 billion (inventory write-downs RMB 2.494 billion, fixed assets RMB 272 million and others), with an impact on parent net of −RMB 1.621 billion, reconciling with the RMB 2.881 billion of asset impairment losses in the income statement. One positive and one negative, together pushing parent net above the ceiling and ex-non-recurring below the floor.

But the positive evidence must also be given in full — operations really are accelerating, and this cannot be read out of the pre-announcement: Q2 standalone parent net of RMB 5.814 billion (+161.7% YoY), ex-non-recurring RMB 4.430 billion (+136.4% YoY, +41.7% QoQ), and gross margin rising from 3.56% in Q2 last year to 6.44% (Q1 was 5.07%). This portion excludes one-off gains.

Position in the value chain (this changes the narrative):

• Captive mine self-sufficiency is only 7.9%. The interim report's "copper contained in self-produced concentrate +37.36%" is a changed definition: the numerator includes 36,400 tonnes of First Quantum attributable output while the denominator (99,300 tonnes last year) does not; on a captive-mine basis it is actually 100,000 tonnes vs 99,300 tonnes, only +0.7% — that footnote is printed right below the output table.
• Subsidiary net margins: mining 34.1% / copper smelting 1.39% / gold smelting 1.07% / trading 0.20%–0.48%; overall net margin on sales 3.09%, weighted ROE 10.22% (Zijin 19.60%, CMOC 18.29%).
• TC/RC are at historic extreme negatives: the company states a H1 low of −US$124.45/tonne; Guosheng Securities, citing SMM, puts the imported copper concentrate index at −US$159.37/tonne on 7/31; on a spot TC/RC basis theoretical smelting profit is −RMB 1,588/tonne, and on a long-contract basis +RMB 4,497/tonne.

It is not a "mining beta play on rising copper prices" but a "processor operating in a high copper and high gold/silver price environment". An extremely low gross margin (5.82%) × RMB 307.1 billion of revenue = an extremely high price-elasticity amplifier, effective in both directions.

Other negatives: Q2 standalone operating cash flow of only RMB 306 million (−86.8% YoY, quarterly cash-to-profit ratio 5.3%), with almost all of H1's cash flow coming from Q1; interest-bearing debt up RMB 65.1 billion in half a year to RMB 140.4 billion, gearing 65.48% (but only +1.76pp YoY, which is seasonal and should not be overstated); no interim dividend, versus RMB 4 per 10 shares (RMB 1.385 billion) in the prior-year period.

Valuation: closed 8/25 at RMB 44.15, market cap RMB 152.880 billion. PB 1.74x (Zijin 4.69x, CMOC 4.39x), still roughly a 29% discount after PB/ROE adjustment; A/H premium 41.45% (H shares at HKD 36.30, read at the 8/25 16:08 Hong Kong close; conversion rate HKD/CNY = 0.8599, cross-checked against two independent sources — er-api quotes 0.859898 directly, and the CNBC cross of USD/HKD 7.8378 ÷ USD/CNY 6.7203 = 0.8574. The renminbi is currently in a strong range, which also corroborates the FX losses appearing simultaneously in the interim reports of both 沪电股份 and 江西铜业. There is a one-hour information gap versus the 15:00 A-share close, and at that time neither market had seen the interim report). But PE (TTM) of 13.19x is nearly identical to Zijin's 13.24x and CMOC's 14.17x while ROE is only half theirs — on a PE basis it is the most expensive of the three.

Timing (the one clean point): CNINFO records the disclosure date as 2026-08-26 with the board meeting on 8/25, so at the 8/25 close the market had not seen this interim report. The −3.92% that day was not a reaction to it (SHFE copper was +0.06% that day, the non-ferrous sector −2.78%, peers broadly down: Tongling −3.02%, CMOC −3.00%, Zijin −2.49%). So the interim report is entirely new information for today's open — it is just that this new information cuts both ways and cannot be written up as "unfairly sold off" or "the bad news is out".

Final judgement: Watch closely, not the top recommendation slot. Only three pivots still stand — the genuine Q2 standalone ex-non-recurring acceleration of +136.4%, SHFE copper at RMB 107,980/tonne (8/25 settlement) sitting 5.85% above the company's disclosed H1 average of RMB 101,817/tonne, and the 29% PB/ROE discount — and all three must be read in the same paragraph as ex-non-recurring below the forecast floor, the RMB 2.905 billion impairment, Q2 operating cash flow of RMB 306 million, and no interim dividend.

---

3️⃣ 亨通光电600487Watch closely | 51 pts · Hengtong Optic-Electric · SH main board ±10%

Related news: no company-level news in the window (I checked the announcements disclosed on 8/26; the company has not released an interim report). It is included because of flows.

Flow evidence (strongest market-wide): closed 8/25 at RMB 64.60, +6.90%, turnover rate 10.60%, block-trade net buying of RMB 1.850 billion (buys RMB 2.664 billion / sells RMB 814 million), the largest market-wide that day and far ahead of second place (英维克 (Envicool, 002837) at RMB 369 million).

Seat breakdown (this is the key; net amount alone is not enough):

Top 5 buyers	Amount
Shanghai-Hong Kong Stock Connect dedicated seat	RMB 1.089 billion
Institutional dedicated seat	RMB 656 million
CICC Shanghai Branch	RMB 421 million
Goldman Sachs (China), Century Avenue, Pudong, Shanghai	RMB 276 million
J.P. Morgan (China), Yincheng Middle Road, Shanghai	RMB 221 million

Table scrolls horizontally

Among the top 5 sellers, the Stock Connect dedicated seat sold RMB 335 million (still net buying of RMB 754 million) and CICC Shanghai sold RMB 116 million. The buy side is northbound + institutions + foreign investment banks, not hot money — this is the highest-quality flow of the day.

Branch background: on 8/25 the "communication cable and accessories" sector was +4.46% with RMB 4.389 billion of net main-force inflow; G.652.D fibre prices have risen from roughly RMB 20 per fibre-km at end-2025 to above RMB 35 in January 2026, driven by AI data-centre demand plus the long expansion cycle for optical preforms.

Risks: ① there is no stock-level news in the window, so buying today is purely chasing flows; ② it was already +6.90% on 8/25; ③ PE (TTM) about 45 (read on 8/7, may have changed since); ④ the interim report is not out, a source of uncertainty; ⑤ in the same sector, 通鼎互联 (Tongding Interconnection, 002491) had block-trade net selling of RMB 455 million on 8/25 — the branch is already differentiating internally.

Final judgement: Watch closely, no recommendation slot. The reason is simple — the branch is strong and the flows are strong, but the stock has no new information in the window, so it does not meet this brief's "new driver" inclusion threshold. If it gaps up today and is absorbed by flows of the same quality, that can serve as a confirmation signal of branch strength, not as an entry point.

---

4️⃣ 英维克002837Pass(first draft: "watch closely, rank 4"; re-judged after verification) | 22 pts · Envicool · SZ main board ±10%

⚠️ This is the second self-overturn in this brief. The first draft put it in a recommendation slot on the grounds that "the interim report had no pre-announcement and is a first disclosure + the price level is not high". Neither holds.

① The interim report was not released on the evening of 8/25; it went live at 8/24 18:15, and it is negative.

Time	Event
8/22 (Saturday)	Board approved the interim report
8/24 18:15	First release: revenue +17.24%, parent net profit −14.32%
8/25 08:51 / 08:54	China Securities Journal and Jiemian re-reported pre-market
8/25 09:30	Opened at 53.54, gapping down −2.21%, low of −3.18%
8/25 around 10:15	Sealed limit-up, closed at 60.23 (+10.01%)

Table scrolls horizontally

Cross-check: on 2026-04-21, the day this company announced its Q1 report, it went limit-down at the open with only 60,000 lots traded all day, showing that its periodic reports are consistently made public before the open. By the pre-market of 8/26, this information is already 36 hours old and has been fully traded.

② The interim report itself is negative: H1 revenue RMB 3.0167 billion (+17.24%), parent net RMB 184.8 million (−14.32%), ex-non-recurring RMB 175.5 million (−13.22%), gross margin 24.93% (−1.22pp), weighted ROE 5.19% (−1.98pp). Revenue up, profit down = negative operating leverage (operating costs +19.16%, administrative expenses +28.98%, financial expenses +2,014.88%).

③ Q2 is deceleration, not acceleration: revenue YoY went from Q1 +26.03% → Q2 +12.24% (decelerating 13.8pp, the lowest since 2024); Q2 parent net was only +5.06%. The "sharp QoQ recovery in Q2" that bulls point to (RMB 8.66 million → RMB 176 million) is a base effect — the comparison base is a Q1 that was close to zero.

④ The limit-up was sector beta, not stock alpha: on 8/25 the whole group moved together — 申菱环境 (Shenling Environment, 301018) +16.16%, 同飞股份 (Tongfei Refrigeration, 300990) +13.56%, 英维克 (002837) +10.01%, 高澜股份 (Gaolan Technology, 300499) +6.39%, 飞荣达 (Shenzhen FRD, 300602) +6.22%, and from 8/18 to 8/24 all five moved highly in sync (all falling 8%–10% on 8/19).

Metric caution: 申菱环境 and 同飞股份 are ChiNext ±20% names while 英维克 is SZ main board ±10% and had already sealed its ceiling that day, so "they rose more" cannot by itself prove that Envicool underperformed — a price-capped stock structurally cannot outrun a ±20% name. What genuinely proves "beta, not alpha" are these three points: ① two peers with no interim-report news at all rose together with it the same day, showing the driver came from the sector rather than the company; ② it gapped down −2.21% at the open with a low of −3.18% — if the interim report were positive, it would not have fallen first; ③ the seal quality was poor (see ⑧: sealing-order/turnover ratio only 0.058, with almost no volume in the afternoon).

⑤ The liquid-cooling narrative is broken at the disclosure level: the company has never separately disclosed liquid-cooling revenue or its share. The only quantifiable metric is the subsidiary Envicool Intelligent Connection (liquid-cooling piping / quick couplings) with H1 revenue of RMB 191.3 million, 6.34% of the group. Its only disclosed relationship with NVIDIA is that its UQD/MQD products are included on the MGX ecosystem partner list — ecosystem certification ≠ a supply contract, and the company has never disclosed any orders or revenue from NVIDIA. In April–May 2026 several media outlets questioned its "NVIDIA supplier" status. The "data-centre thermal management" segment that carries liquid cooling has a gross margin of 23.34%, actually below the company average of 24.93%, and with the largest decline (−2.49pp).

⑥ "Only +9.3% over the last 20 days, so the level is not high" is an illusion created by the statistical window. The real path is 79.56 on 7/1 → 45.82 on 8/7 (−42%, with two limit-downs along the way) → 60.23 now. This is not a launch after low-level consolidation; it is the third week of a rebound after a deep decline. It is still −35.6% from the all-time high of 93.515, but already +31.4% off the 8/7 low.

⑦ Valuation is the most expensive among peers: PE (TTM) 156.6x, PB 21.5x, versus Tongfei at 6.7x / 飞荣达 (Shenzhen FRD, 300602) at 6.1x / Gaolan at 6.8x — PB is 3.2–3.5x that of peers while ROE is only 5.19%. Of the three that have released interim reports, Envicool is the only one with negative profit growth (Shenzhen FRD +52.09%, Gaolan +12.22%), so the PEG denominator is negative and cannot be calculated.

⑧ Holdings: block-trade net buying of RMB 369 million, but broken down it is three hot-money seats at +RMB 395 million, Shenzhen Stock Connect net selling of RMB 108 million, and institutions with 1 buy and 3 sells; the sealing-order/turnover ratio is only 0.058 (a weak seal); and 96.7% of the volume was concentrated in the morning, with only 3.3% all afternoon.

⑨ The NVIDIA exposure is structured as "bearing the risk without sharing the upside": overseas revenue is only 14.72% of the total; the company has never disclosed any orders or revenue from NVIDIA, so their scale cannot be verified (note: "not disclosed" does not mean "zero", but it does mean this narrative cannot be confirmed). In the interim report's risk factors the company states: "Whether overseas market expansion proceeds smoothly and whether it reaches scale production are both uncertain. The company's upfront resource investment will also have a certain impact on operating results." If NVIDIA misses tonight, its financials would be essentially unaffected in substance, but the 156x PE / 21.5x PB premium rests entirely on the AI liquid-cooling narrative, so the elasticity of a valuation de-rating far exceeds its earnings elasticity; conversely, if NVIDIA beats, revenue will not be mechanically added to it either.

Final judgement: Pass. Not merely because of "no new news", but because two conditions hold at once: there is no new 8/26 information (the news is old news from 8/24), and that old news is itself negative (parent net −14.32%). The only honest bull argument is that operating cash flow turned from −RMB 234 million to +RMB 62 million, and that overseas revenue grew +71.39% at a gross margin of 48.20% — but the former still leaves a cash-to-profit ratio of just 0.33, and the latter accounts for only 14.72%.

---

5️⃣ 泸州老窖000568Avoid, and set a defence around the baijiu sector | Negative S · Luzhou Laojiao · SZ main board ±10%

This is the only heavyweight, entirely unpriced, downward-pointing piece of news in the window, and it warrants its own entry.

Related news: interim report disclosed post-close on 8/25, H1 revenue RMB 10.472 billion, −36.35% YoY; parent net profit RMB 4.339 billion, −43.37% YoY (Sina Finance³).

Why it is unpriced — this is the key reasoning in this brief: I pulled the full table of 2026 interim earnings pre-announcements for both Shanghai and Shenzhen (4,876 records) and checked them one by one; Luzhou Laojiao is not on the pre-announcement list. The reason is the rule: main-board companies must pre-announce only when net profit is negative, when they swing to profit, or when the change exceeds 50%, and −43.37% does not reach the 50% threshold, so the company had no obligation to pre-announce and indeed did not. Conclusion: this is a genuine first disclosure. The stock fell only −0.91% on 8/25, with Moutai −0.05%, 五粮液 (Wuliangye, 000858) +0.06% and 山西汾酒 (Shanxi Fenjiu, 600809) −0.38%: the entire baijiu sector closed completely unguarded.

Transmission scope: the company describes it as "volume contraction under a control-supply, support-price policy, with reduced output of mid- to high-end liquor". That is the language of an industry-wide demand problem rather than a company-specific case, which puts pressure on the whole baijiu sector. Also note: on 8/25 养元饮品 (Yangyuan Beverage, 603156) hit limit-down at −9.99% with block-trade net selling of RMB 53.62 million — cracks have already appeared inside the broader consumer space.

Final judgement: Avoid baijiu and related consumer names today. This is not "one stock's bad news" but an industry reading of −43% magnitude landing at a level where nobody expected it.

---

6️⃣ 盐湖股份000792Watch only (classification mismatch) | 28 pts · Qinghai Salt Lake Industry · SZ main board ±10%

Related news: interim report post-close on 8/25, H1 parent net RMB 6.169 billion (+137.88%). The 7/3 pre-announcement had given RMB 6.000–6.300 billion, so parent net landed at the 56.3% percentile; ex-non-recurring of RMB 6.001 billion landed at the 0.2% percentile — almost exactly on the floor.

The value of this entry is not the recommendation but that it falsifies an entire branch's logic (see §2 Branch 7 for details):

• Lithium product gross profit of RMB 4.675 billion > potash product gross profit of RMB 4.467 billion; of the RMB 5.036 billion of combined gross profit increment, lithium accounts for 75.9% and potash 24.1%; lithium gross margin is 83.46% (+29.3pp) versus potash at 62.59% (+2.7pp).
• Potassium chloride output −15.5%, sales volume +26.4%, realised price only +5.2% → within the potash segment's +31.4% revenue growth, roughly 80% comes from one-off destocking (565,700 tonnes drawn down, with inventory falling from 1.1716 million tonnes at end-2025 to an estimated normal level of about 606,000 tonnes).
• The 2026 import large contract is US$348/tonne CFR vs US$346 in 2025 — +0.6% YoY.
• Operating cash flow grew only +1.38% (against +137.9% profit growth), and operating cash flow / net profit fell from 2.16 to 0.87, mainly because notes receivable surged by RMB 1.790 billion (company note: increased use of bank acceptances in lithium carbonate sales).
• The consolidation of Minmetals Salt Lake contributes only 7.6% of the increment (organic 92.4%) — this is a counter-intuitive clarification: as a business combination under common control, Minmetals has already been retrospectively included in last year's base, so the "consolidation padding" claim does not hold; the genuinely wrong reading is dividing 6.169 by the un-restated 2.515 to get +145%.

Valuation: closed 8/25 at RMB 27.45, market cap RMB 145.2 billion. PE (TTM) about 12.0, PB 3.04; cash on hand RMB 32.271 billion, net cash of about RMB 34.3 billion (23.6% of market cap; definition = cash RMB 32.271 billion + other current assets RMB 3.398 billion (mainly large-denomination certificates of deposit) − interest-bearing debt of about RMB 1.37 billion), gearing of only 14.07%, no goodwill. The financial statements are extremely clean.

Pattern warning: a low PE (12.0) + a relatively high PB (3.04, the highest among potash peers) is the classic reading for a cyclical whose earnings are at a cyclical peak, with annualised ROE of about 26.7%. This is the textbook "low PE trap" pattern — the low PE comes from a cycle-peak denominator, while the high PB exposes the true valuation. Treat it with caution.

Final judgement: Watch only. The statements are hard and clean, but it is classified by the market as "potash / agriculture" while in reality it is driven by lithium; if someone buys it today because of the agriculture theme, the attribution is roughly 75% wrong. It is −9.05% since 7/2 and has underperformed the agriculture ETF by 14.3pp — the share price has in fact long been priced on lithium; only the classification table still says potash.

---

7️⃣ 杭电股份603618Watch only | 38 pts · Hangzhou Cable · SH main board ±10%

Related news: interim report post-close on 8/25, parent net RMB 393 million (+938.67%), landing at the 82.5% percentile of the 7/4 forecast range (RMB 360–400 million) — the second-highest landing point in this batch. The stock was +4.39% on 8/25.

Three things that must be spelled out:

1. The +938.67% is a product of a low base. The 2025H1 base was only RMB 37.81 million, because the company took RMB 362 million of asset impairment provisions in FY2025. The media's own framing is "yesterday's 'impairment burden' has flipped into a 'profit cash cow'".
2. On the same day it filed another "H1 2026 Asset Impairment Provision" announcement (confirmed to exist on the 8/26 disclosure list, but this brief could not verify the specific amount). Alongside a set of financials whose high growth rate comes from an impairment reversal, a fresh provision appearing again is a signal that warrants separate tracking.
3. There is a placement and share reductions overhead: in July the company disclosed a "2026 Plan for the Private Placement of A Shares" (dilutive) and an "Announcement of Share Reduction Plans by Shareholders, Directors and Senior Management".

Final judgement: Watch only. The high landing percentile is real, but with four items stacked — low base + a new provision + placement dilution + reductions by directors and senior management — the risk/reward is asymmetric.

---

8️⃣ 艾可蓝300816Watch only (needs verification) | 34 pts · Aecoblue · ChiNext ±20%

Related news: announced post-close on 8/25 the signing of a RMB 1.109 billion compute-resource services contract.

Why it must carry a question mark:

• The company's core business is diesel engine exhaust after-treatment / environmental protection equipment, so compute-resource services are a cross-sector business;
• The contract value of RMB 1.109 billion ≈ 102.6% of its full-year 2025 revenue of RMB 1.081 billion — a single contract larger than a full year of revenue;
• 2025 net profit was only RMB 95 million and gearing is 59.39%, so whether it has the financial capacity to procure compute assets is questionable;
• This brief could not obtain the original announcement, so the counterparty, payment schedule, performance period and whether there is any prepayment are all currently unconfirmable.

Final judgement: Watch only, and not suitable for a recommendation slot today. A contract value this large relative to market cap is a double-edged sword: if genuine it is a major positive, but if it is a framework agreement the substance is limited. Until the original announcement is in hand, any conclusion is a guess.

---

9️⃣ 登海种业002041Watch only | 35 pts · Denghai Seeds · SZ main board ±10%

2 consecutive limit-ups, +10.01% on 8/25 with a turnover rate of 9.64% and block-trade net buying of RMB 167 million (appearing on the board under "cumulative deviation of 20% over three consecutive trading days"). Overnight CORN +1.93% provides branch confirmation.

But: it is already +33.0% over the last 20 days and branch pricing has entered trading day 3–5; the catalyst (the UK Met Office forecast of 8/21) is outside the window. The logic is right, the price level is not.

---

🔟 汉森制药 (Hansen Pharmaceutical, 002412, SZ main board ±10%) | Pass | 14 pts

5 consecutive limit-ups, +76.1% over the last 20 days, turnover rate 21.27% on 8/25.

The decisive evidence for Pass is on the block-trade board: although the stock limit-up on 8/25, it had block-trade net selling of RMB 102 million; the sell-side seats include two institutional dedicated seats (net selling RMB 22.02 million and RMB 21.60 million respectively), while the three institutional dedicated seats among the top five buyers together bought a net of only about RMB 29.63 million. That is: on the fifth limit-up, institutions were net sellers.

This is the classic pattern in which "the limit-up is sealed" and "capital is distributing" hold at the same time.

---

6. Pass List

---

7. Intra-Branch Ranking

Branch 1 | PCB / high-speed switches

Hardest stock: 沪电股份 (002463) (the only one with an interim report and the only one with Q2 acceleration). Note: on 8/25 the printed circuit board sector saw net main-force inflow of RMB 2.365 billion with the sector +1.32%, while the leader WUS was actually −1.11% — money went into the sector, the leader did not rise, and then the leader released its interim report after the close. That sequence deserves attention.

Branch 2 | Copper / non-ferrous

Hardest stock: Jiangxi Copper is still the only one with in-window financials, but its "hardness" is now down to the single item of Q2 standalone ex-non-recurring. The comparison this group most needs to remember: Jiangxi Copper PE 13.19x / ROE 10.22% / PB 1.74x versus Zijin PE 13.24x / ROE 19.60% / PB 4.69x — near-identical PE with double the ROE, which is the permanent valuation gap between "smelting assets and mining assets" and is not something this interim report can change. The PEs of Tongling Nonferrous and Yunnan Copper are lagging TTM figures built on "FY2025 + 2026Q1" and cannot be compared directly with Jiangxi Copper; they can only be read as "how far pure smelters are being squeezed in a negative-TC environment".

Branch 3 | Optical fibre & cable / optical communications

Hardest stock: none simultaneously has "in-window news + fundamentals + a good price level". A strong branch with weak stocks is the accurate description of this group. Who gets passed: 通鼎互联 (002491) (capital is clearly exiting).

Branch 5 | Grain / planting (El Niño)

This branch gets no recommendation slot at all. The overnight confirmation is real (CORN +1.93%), but the A-share leg has already run most of its course.

---

8. Opening Verification Signals for Today

Call-auction signals

• 江西铜业 (600362): this is the name that most needs the call auction to be characterised today. The market has two ways of reading this interim report, and the auction will tell you which one it picked:
  • Reading "parent net RMB 8.632 billion, +106.77%" → gap up of +3% or more;
  • Reading "ex-non-recurring RMB 7.556 billion below the forecast floor + RMB 2.905 billion impairment + no dividend" → flat or lower open. Do not participate on a gap up of +4% or more (that is paying for a zero-cash-flow book revaluation); only a flat to +2% open that holds is pricing the Q2 standalone ex-non-recurring of +136.4%. At the same time watch whether 紫金矿业 (Zijin Mining, 601899) (interim report already out 8/22) moves in the same direction — if Zijin does not rise while Jiangxi Copper rises alone, that means people are buying the news, not copper.
• 沪电股份 (002463): a gap up of +2% to +4% with the turnover rate expanding to 3%–5% is ideal. Give it up above +6%. The key is whether it holds after the open — holding with volume = the market is pricing the Q2 gross margin slope; gapping up then fading = the market is pricing the 0.39 cash-to-profit ratio and the +82.5% rise in interest-bearing debt.
• 泸州老窖 (000568): watch how far it gaps down, and whether Moutai / 五粮液 (Wuliangye, 000858) / 山西汾酒 (Shanxi Fenjiu, 600809) follow it down. If all three leaders gap down more than 1% together, the negative is spreading to the sector and the broader consumer space should be avoided today.
• 亨通光电 (600487): do not chase a gap up. Watch whether yesterday's high-quality flows (northbound / institutions) keep absorbing supply today.
• 英维克 (002837): already re-judged to Pass; today it serves only as a reverse indicator — if it gaps up and then weakens quickly, that is a leading signal of "the liquid-cooling concept beta receding"; watch at the same time whether 申菱环境 (301018) and 同飞股份 (300990) pull back in sync.

Sector signals

• Strong: ① three or more copper names turn green with Jiangxi Copper leading → the branch holds; ② the PCB sector records a second consecutive day of net main-force inflow (8/25 was already +RMB 2.365 billion) with WUS leading → the leader is back in place; ③ the communication cable sector records a second consecutive day of net inflow (8/25 +RMB 4.389 billion).
• Weak: ① the baijiu sector as a whole gaps down more than 1%; ② agriculture gaps up and fades (especially if Jinjian breaks its seal); ③ non-ferrous metals continues the RMB 5.729 billion net outflow of 8/25.

Stock-level signals

• Absorption volume in the first 5 minutes, whether the seal is quick, whether volume expands without breaking the seal, and whether large-order net inflow turns positive;
• Jiangxi Copper: whether it gets outflanked by peers (if Tongling Nonferrous limit-ups while Jiangxi Copper does not rise, capital is trading the branch and not endorsing the company);
• WUS: whether it gets outflanked by 深南电路 (002916) / 胜宏科技 (these two have not released interim reports, so they carry the relative advantage of "expectations not yet delivered").

Risk signals (if any two appear together, offence should be abandoned today)

1. The baijiu sector opens lower, drifts lower and spreads to food and beverage as a whole — meaning Luzhou Laojiao's negative is being read as an industry read;
2. Any two of 金健米业 (600127) / 汉森制药 (002412) / 深中华 A (000017) breaking their seals — the sentiment ladder is fracturing (of the 8/24 limit-up names, only 43% closed green on 8/25);
3. Jiangxi Copper gapping up +4% or more — meaning the market read only parent net and not ex-non-recurring, and that kind of gap up has to be paid back;
4. WUS gapping up +4% or more and then falling back to flat — meaning the market is reading cash flow, not gross margin;
5. Non-ferrous metals continuing large net outflows (8/25 was already −RMB 5.729 billion);
6. The index opens lower and turnover contracts further from the RMB 1.83 trillion of 8/25;
7. The compute chain weakening as a group into the close — capital reducing exposure ahead of NVIDIA tonight.

---

9. Final Recommendation Conclusions

① Stocks most worth watching today

Screening rule (retained and tightened): recommendation slots are reserved only for names that satisfy all three of ①a new driver within the window ②the driver comes from the company's own financial metrics or verifiable primary material ③not priced yesterday. After re-checking primary announcements one by one, this brief has overturned two of the first draft's recommendation slots: 江西铜业 is downgraded from rank 1 to a secondary slot (§5-2), and 英维克 is re-judged to Pass from rank 4 (§5-4). In the end only 1 name fully satisfies all three conditions, with 2 others in secondary slots. The remaining two places are left blank rather than padded.

② The three strongest branches today

③ Directions not recommended for chasing today

1. Baijiu and broader consumer — this is the item most in need of a defence today. 泸州老窖 (000568) H1 revenue −36.35%, parent net −43.37%, and because the decline did not reach 50% it never issued a pre-announcement, so the interim report is the first disclosure; at the 8/25 close the entire sector was completely unguarded (Moutai −0.05%, 五粮液 (000858) +0.06%, Fenjiu −0.38%). The company describes "volume contraction under a control-supply, support-price policy", which is industry-level language. On the same day 养元饮品 already hit limit-down at −9.99%.
2. The entire fertiliser / agrochemical branch — rejected by three independent pieces of evidence at once: ① overnight NTR −1.95%, CF −1.53%, forking away from rising grain ETFs; ② 盐湖股份's interim report directly falsifies the price-increase logic (realised potash prices only +5.2% YoY, the large contract at US$348 vs US$346 i.e. +0.6%, potassium chloride output −15.5%, with about 80% of revenue growth coming from one-off destocking); ③ 百傲化学 is not an agrochemical company at all (core business industrial biocides, with 2025 semiconductor revenue already exceeding chemicals), and 盐湖股份 is in reality a lithium stock (lithium gross profit RMB 4.675 billion > potash gross profit RMB 4.467 billion).
3. Chasing grain / planting higher — the logic was confirmed overnight (CORN +1.93%), but the price level does not permit it: 金健米业 (600127) is +69.0% over 20 days with a turnover rate of 46.17% and a sealing-order/turnover ratio of only 0.011. Acknowledging the branch is real and acknowledging the price is too expensive are not contradictory.
4. Chasing 英维克 and the whole liquid-cooling concept higher — this is the second self-overturn in this brief. Envicool's interim report was already released at 8/24 18:15 (not the evening of 8/25), and parent net was −14.32%, ex-non-recurring −13.22%, with Q2 revenue growth falling from +26.03% to +12.24%. The 8/25 limit-up was sector beta rather than stock alpha: on the same day 申菱环境 (301018) +16.16% and 同飞股份 (300990) +13.56%, neither of which had any interim-report news, yet both rose more. On valuation, PB of 21.5x is more than 3x that of peers (Tongfei 6.7x, 飞荣达 (300602) 6.1x) while ROE is only 5.19%, and of the three that have released interim reports it is the only one with negative profit growth. Liquid-cooling revenue has never been separately disclosed by the company, and its relationship with NVIDIA is only the inclusion of UQD/MQD products on the MGX ecosystem list — ecosystem certification is not a supply contract.
5. Buying 江西铜业 as an "earnings beat" — this is the first self-overturn in this brief. The 7/9 pre-announcement gave two ranges: parent net RMB 7.550–8.500 billion and ex-non-recurring RMB 7.630–8.580 billion; the actual parent net of RMB 8.632 billion beat the ceiling, but ex-non-recurring of RMB 7.556 billion fell below the floor. The difference is +RMB 1.076 billion of non-recurring items, the largest being RMB 673 million from the book revaluation of its own pre-existing 12.14% stake at the time of the SolGold acquisition, with zero cash flow. Stripping it out gives exactly the ex-non-recurring RMB 7.556 billion, below the ex-non-recurring forecast floor. On the same day it also took RMB 2.905 billion of impairment provisions (impact on parent net −RMB 1.621 billion), and there is no interim dividend.
6. Consecutive-limit-up sentiment names (汉森制药 (002412), 深中华 A (000017), 新华百货 (600785)) — ① of the 46 limit-up names of 8/24, only 43% closed green on 8/25, averaging +1.03%, so follow-through is weak; ② 汉森制药 had block-trade net selling of RMB 102 million on its fifth limit-up, with institutions as net sellers; ③ 深中华 A sealed a fourth limit-up on a day when the precious metals sector was −5.25%, which is pure capital behaviour.
7. Treating 万华化学 as a positive — the Hungarian MDI/TDI units completing maintenance and resuming production means supply is returning, which is pressure on prices rather than a positive. Do not read the direction backwards.
8. Chasing 杭电股份 higher — the +938.67% is a product of the low base created by the RMB 362 million impairment provision taken in 2025; a new impairment provision announcement was filed the same day; and there is still a placement plan and a reduction plan by directors and senior management overhead.
9. Going heavily into the compute chain before 8/27 — NVIDIA's FY2027Q2 results come out at 8/27 05:00 Beijing time. The entire trading day today is before the event, so any compute-chain position is prepaying a premium for tomorrow morning's outcome. NVIDIA's overnight +2.19%, ending a seven-day losing streak, has already repaired part of that expectation in advance.
10. 艾可蓝's compute contract — do not participate until the original announcement is in hand. The contract value of RMB 1.109 billion ≈ 102.6% of 2025 revenue, while the company's core business is diesel exhaust after-treatment, 2025 net profit was only RMB 95 million and gearing is 59.39%; both its ability to perform and its counterparty are currently unconfirmable.

④ Final one-sentence judgement

The genuinely new news in today's window is almost entirely interim reports, and most of the "stunning growth rates" in this batch were already announced once in early July — 沪电股份 (002463), 盐湖股份 (000792), 雅化集团 (002497), 杭电股份 (603618), 当升科技 (300073), 江西铜业 (600362), 招商证券 (600999) and 高德红外 (002414) without exception, all with pre-announcement dates falling between 7/3 and 7/17, and WUS even limit-up 10% on the day of its pre-announcement before giving back 16.66% since. On that basis the first draft of this brief picked out 江西铜业 as "the only one to beat its forecast ceiling" and placed it first — and after verifying the primary source, that pivot collapsed: the pre-announcement of 9 July gave two ranges side by side, parent net RMB 7.55 to 8.5 billion and ex-non-recurring RMB 7.63 to 8.58 billion; the actual parent net of RMB 8.632 billion did beat the parent-net ceiling, but ex-non-recurring was only RMB 7.556 billion, below the ex-non-recurring floor. The difference is RMB 1.076 billion of non-recurring gains, the largest single item being RMB 673 million from a book revaluation of the 12.14% stake it already held when it spent RMB 6.967 billion in cash to acquire SolGold — not one cent of cash was generated. And since net non-recurring gains are by definition "parent net minus ex-non-recurring", stripping them out leaves exactly that ex-non-recurring figure of RMB 7.556 billion — only six million above the floor of the parent-net forecast, yet solidly below the floor of the ex-non-recurring forecast. On the same day it also took RMB 2.905 billion of impairment provisions, and in a half-year in which profit doubled it cancelled the interim dividend.

So only one name genuinely stands up today: 沪电股份 (002463). And what makes it worth looking at is precisely not an "earnings beat" — on the pre-announcement basis it did not beat, and ex-non-recurring even landed in the lower half of the range — what makes it worth looking at is that its Q2 standalone gross margin jumped from the narrow 35.6% to 35.8% range of the previous three quarters straight to 41.27% (+5.64pp QoQ, +3.96pp YoY), together with the 67.75% revenue growth rate; neither number was in the 14 July pre-announcement, the stock still fell 1.11% at yesterday's close, and the interim report only came out after the close. As for the downward-pointing item, it is equally unpriced: 泸州老窖's revenue −36.35% and parent net −43.37%, and because the decline did not reach the 50% pre-announcement threshold there was never a pre-announcement from beginning to end, so at yesterday's close the entire baijiu sector was still standing exactly where it was.

In this brief I overturned myself twice — the first time on 江西铜业, where I looked at only one of two ranges, and the second time on 英维克, where I treated an interim report released on the evening of 24 August with parent net −14.32% as a positive new item of 25 August. Neither was a case of taking the wrong data; both were cases of not taking enough of it: once I missed the other half of a sentence printed on the same page, and once I missed the clock on an announcement. So what is most worth taking away today is not a particular stock but the action itself — before scoring the expectation gap on any interim report, first open the original July pre-announcement, read both ranges to the end, and then take one more look at exactly what day and what hour it was released.

---

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

---

Data this brief could not obtain (blanks explained):

• The 8/25 night-session closing data for domestic agricultural futures (corn, soybean meal, strong-gluten wheat, sugar) was not obtained, so the overnight verification for the agriculture branch relies solely on CBOT agricultural ETFs (CORN/WEAT/SOYB) and overseas fertiliser-stock proxies — domestic order-book confirmation is missing.
• The specific amount in 杭电股份 (603618)'s "H1 2026 Asset Impairment Provision" was not obtained.
• The original announcement text for 艾可蓝 (300816)'s RMB 1.109 billion compute contract (counterparty, performance period, payment schedule) was not obtained.
• 江西铜业 (600362)'s latest August 2026 SMM imported copper concentrate index was not obtained (that page requires login); the −US$159.37/tonne cited in this brief is the 2026-07-31 reading, relayed via a Guosheng Securities weekly report. The claim that "the 2026 long-contract TC/RC benchmark is US$0/tonne" has no primary source in this brief and therefore has not been written into the main text.
• The precise disclosure time of Jiangxi Copper's interim report was not obtained (CNINFO records the disclosure date as 2026-08-26 and the board meeting date as 2026-08-25), so this brief can only confirm that "the market had not seen it at the 8/25 close" and cannot judge whether there was any intraday front-running.
• The cause of action and amount of the "arbitration-frozen" cash in Envicool's interim report could not be broken out separately (restricted cash totals RMB 107.5 million).
• There is no publicly citable quote for Vancouver FOB / Brazil CFR potash spot prices, so the potash price judgement in §2 Branch 7 rests on the import large-contract price (US$348 vs US$346/tonne) and the domestic reference price from ChemNet.

Internal Notes (not sent to clients)

Data-retrieval failures

• Eastmoney's push2.eastmoney.com and push2his.eastmoney.com refused this machine's IP (RemoteDisconnected / JSONDecodeError), so full A-share quotes, individual K-lines and continuous futures contracts could not be fetched directly. Rerouted: quotes via Sina quotes.sina.cn, sector fund flows via push2delay.eastmoney.com, the limit-up pool via push2ex.eastmoney.com (those two work).
• Domestic continuous futures contracts failed on both push2delay and push2his, which is why the agriculture branch lacks domestic night-session verification.
• akshare stock_zh_a_hist returned repeated RemoteDisconnected errors (same IP rate limiting), so everything was switched to Sina daily bars.
• The exact advance/decline counts for the full A-share market could not be obtained because clist pagination timed out, so this brief does not cite that figure (to avoid repeating the "one-sided pagination bias"); limit-up counts were switched to the official ZTPool with len(pool)==tc==65 verified.
• All four sub-agents (沪电股份 (002463), 盐湖股份 (000792), 江西铜业 (600362), 英维克 (002837)) returned and have all been written back. Two of them directly overturned the first draft's conclusions:
  • 江西铜业 (600362): first draft "the only one to beat the forecast ceiling" → in fact "parent net above the ceiling, ex-non-recurring below the floor". The first draft only checked one of the two ranges in the 7/9 pre-announcement, and the one it missed was printed on the same page. Score cut from 76 to 58, rank from 1 to 2 (secondary slot).
  • 英维克 (002837): first draft "interim report with no pre-announcement, a first disclosure" → in fact already released at 8/24 18:15 with parent net −14.32%. Score cut from 50 to 22, re-judged from "watch closely, rank 4" to Pass. Neither case was a sub-agent adding detail; in both, the sub-agent falsified the main conclusion. The lesson "research sub-agents are worth waiting for" paid off twice today.

Methodology notes (mapped to historical lessons)

• The most valuable lesson today: "trace it back to the first-disclosure date" requires doing two things at once — read every metric range in the pre-announcement to the end (Jiangxi Copper), and read the clock on the announcement (英维克 (002837)). The first draft's two errors happen to be one omission of each.
• The core method in this brief is "trace it back to the first-disclosure date" — all 11 interim reports were checked back against stock_yjyg_em(date="20260630") for pre-announcement dates and ranges, and landing percentiles were computed. This directly overturned the surface reading of "high interim-report growth = positive".
• Two sources that were stale by a day or by a year were caught: ① the "8/25 US close" report actually describes 8/24 (confirmed by back-checking the previous closes); ② the "8/25 Shanghai property policy" is actually the 2025-08-25 document (confirmed by the document number and dateline in the original on the government site). Both were found by back-checking, not by reading.
• 百傲化学 and 盐湖股份 are two instances of the same class of error: industry classification label ≠ actual business. Same root as "Baiyin Nonferrous's captive mines produce no silver".
• The combination of "limit-up + block-trade net selling + institutions as net sellers" at 汉森制药 (002412) is a positive application of "divide sealing orders by turnover, and break institutional net buying down by seat".
• 泸州老窖's "no pre-announcement" was derived through a full-table pre-announcement comparison + the 50% threshold rule, not found by searching. That piece of reasoning is the key to this brief's expectation-gap judgement.