A-Share Pre-Market Brief | 2026-08-17 Monday

0. Today in One Sentence

After tracing every one of this weekend's filings back to its "first disclosure date," the conclusion is: all the eye-catching numbers inside the AI narrative are old news — either pre-announced in early July, or disclosed as a quota in mid-July; while the information that genuinely saw daylight for the first time is lopsidedly negative, with the one exception landing on a chemicals stock outside the narrative.

On August 15 (Saturday) alone, 1,763 filings were disclosed and 136 companies released 2026 interim reports. On the surface the brightest spot is the "network side" of the AI server chain, but after tracing each one back to its first disclosure date:

Conversely, the information that was not pre-disclosed within the window and saw daylight for the first time today is this:

This is not coincidence: earnings growth above a certain magnitude must be pre-announced, so "no pre-announcement" is itself a signal. And among these "no pre-announcement" companies, the only report that is both large and clean, and genuinely free of front-running, is not in the AI narrative — it is the chemicals stock 卫星化学: ex-non-recurring RMB 6.053 billion, 2.1× that of 生益科技 (RMB 2.908 billion), ex-non-recurring/attributable 97.2%, Q2 single-quarter ex-non-recurring QoQ +109.8%, while it closed down 0.79% on 8/14 with net main-force outflow and turnover of just 1.05%. PE-TTM 9.68 (15.2 percentile over the past ten years), PB 2.24 (41.7 percentile).

Why this stock was all but destined to be missed: it belongs to "chemical raw materials" and is not in any AI supply-chain stock pool. Any screening process that "picks the industry narrative first, then picks stocks within the narrative" will systematically fail to see it — even though its ex-non-recurring profit is 2.1× the largest name inside the narrative. The correct order is to run the full-market ex-non-recurring ranking first, then cut into narrative branches. Section 6.1 of this brief publishes the complete screening methodology and the list of exclusions.

Driver type: earnings quality and structure (primary), not the magnitude of earnings — the latter was already priced by July's pre-announcements. What is being priced today is the things a pre-announcement never writes down: segment gross margins, single-quarter inflection, operating cash flow, the balance sheet. And at that layer, the news is mixed and skews negative.

Pre-market state assessment: a day of divergence, with a cautious tilt. Last Friday's leading theme, optical communications, was driven by a policy narrative — no earnings figures, no ladder of consecutive limit-ups. Of the 11 consecutive-limit-up stocks, only 共进股份 (Gongjin Electronics, 603118) is a telecom-equipment name, and it broke its seal 6 times on 8/14 with a seal-order/turnover ratio of just 0.047; no ladder has formed. Of the 6 telecom limit-ups, 5 were first-day boards, and the interim reports of the leaders will not be disclosed until 8/19–8/28. The State Council Information Office briefing on July economic data is scheduled for today at 15:00 (the closing bell), so it is not an intraday disturbance.

⚠️ The single highest-weight finding in this brief: 网宿科技 (300017) disclosed a negative-growth interim report on the night of its limit-up

网宿科技 (ChiNext, ±20%) was the single strongest stock in the entire market last Friday — closing +20.01% (limit-up), main-force net inflow RMB 2.045 billion (first among all individual stocks), Dragon-Tiger List net buy RMB 972 million (first in the market), turnover 20.30%, with 2 institutional buyers. Last week's recap already confirmed it had no company filings whatsoever on 8/13–8/14; that +20% was purely IDC/compute-hosting theme driven.

It never issued an earnings pre-announcement. Its 2026 interim report was logged at 18:19 on August 14 (about 3 hours after the limit-up) and formally disclosed on 8/15: revenue RMB 2.318 billion −1.38%, attributable net profit RMB 348 million −6.61%, gross margin 35.92%, weighted ROE just 3.24% — both revenue and profit down.

In other words: the market's most aggressive RMB 2.045 billion of main-force money bought ahead of an interim report that would announce negative growth that very evening, and nobody had a chance to react. Today is the first pricing day for that report.

This is simultaneously the biggest risk signal and a direct falsification of the inertial assumption that "last Friday's strength = today's strength." See Section 6 and Section 8.4.

---

1. News Overview

Window: 2026-08-14 (Fri) 15:00 close → 2026-08-17 (Mon) 07:00. A total of 1,915 Shanghai/Shenzhen/Beijing A-share filings were scanned within the window (1,763 on 8/15, 2 on 8/16, 150 on 8/17). The "first disclosure date" column is the core of this table — it determines whether a piece of news is new information today, or the confirmation of old information.

News not adopted / downgraded (disclosed candidly):

• Multiple industry articles were found on "four rounds of CCL price increases this year," "Kingboard raising prices another 10–15% in August," "Shengyi's M9 passing NVIDIA certification," etc. — all are secondhand media paraphrases whose publication times cannot be individually confirmed, so they are not counted as catalysts. Hard evidence for price increases is instead taken from the price/volume breakdown in the interim report (see 5.1).
• Rare earths: last Friday 中国稀土 (China Rare Earth, 000831) hit limit-up and the sector rose +5.82%; a repeat search still failed to locate any new policy or company filing within the window, with all hits being the old 2025 filings No. 61/62. The driver remains unidentified (see Section 6).
• No new ministry-level industrial policy was found over the weekend. The "15th Five-Year Plan for the Safety and Emergency Equipment Industry" was released on 8/13 (Thursday), outside the window.
• Time check on the "July economic data": one search summary claimed the data would be released "in the morning"; per the State Council Information Office and the statistics bureau release calendar, it is actually 3 p.m. on 8/17 (the closing bell). The official timing has been adopted.
• The exact timestamp of the Centec filing is in doubt: the local filings API returns 8/16 18:28, while CNINFO labels it 2026-08-17 and the filing is dated "August 17, 2026." The two are inconsistent; this brief treats it as "released within the weekend window" and does not use a minute-level timestamp.

---

2. Strongest Positive Branches, Descending

Branch notes:

• The new information in item 2 (CCL) is positive, but its nature is "confirmation of acceleration," not "unexpected upside": 生益科技's profit magnitude was already announced 7/14; three pieces of structural information are new today: ① excluding 生益电子, the CCL core business's Q2 attributable net profit rose +94.0% QoQ, faster than the +83.9% on a consolidated basis; ② Q2 single-quarter gross margin 32.63% (Q1 28.10%); ③ the price/volume breakdown shows about 2/3 of revenue growth came from price increases and 1/3 from volume.
• Items 2 and 6 are a see-saw, not a joint rally: 生益科技 owns 62.50% of 生益电子 and consolidates it. 生益电子 purchases about RMB 585 million of copper-clad laminate from the 生益 group (16.26% of its purchases of that category), and intra-group eliminations doubled from −RMB 327 million to −RMB 640 million. The parent raising prices = the subsidiary's costs rising, which is precisely the mechanical explanation for why their Q2 single-quarter gross margins moved +4.53pct and −1.54pct respectively. Group profit is migrating from midstream PCB to upstream CCL.
• An important qualification on the "price increase" narrative: 生益科技's domestic gross margin rose +5.86pct while export gross margin rose only +0.57pct; 生益电子's domestic gross margin rose +13.8pct, overseas +0.32pct. If this were a global CCL price increase, export gross margins should also have expanded. That suggests the geographic base of the price increases may be narrower than the narrative implies. This is an open item; no conclusion is drawn.
• Why item 5 (switches) was downgraded from A+ to B: the mystery of 锐捷网络's gross margin now has a clear answer — it is a mix shift, not a price war (direct sales rose from 58.8% to 67.0% of revenue, direct-sales gross margin 25.72% vs distribution 43.58%, while distribution gross margin rose +3.17pct against the trend; the full interim report contains no mention of "white box/ODM"). This point is a clarification and skews positive. But the price is in the cash flow: operating cash flow −RMB 3.018 billion (prior year −RMB 314 million), receivables +RMB 1.821 billion, inventory +RMB 2.654 billion, prepayments +RMB 1.389 billion (+3,125%), short-term borrowings RMB 1.301→4.728 billion, gearing up to 62.92%. Net working capital absorption of RMB 5.86 billion, funded by short-term debt and cash burn. This is the "stock up first, pre-fund first" pattern: it magnifies profit when demand materializes, and simultaneously exposes inventory write-downs, bad debts and interest when demand weakens at the margin. The interim report cannot distinguish which one it is.
• Why item 8 (switch chips) was cut from A to C+: see 5.9 for detail. The core point is that the magnitude information in this contract was already public on 7/16 (the RMB 2.2 billion quota filing), the shareholders' meeting approved it on 8/6, and the market already reacted at the time (7/16 +5.96%, the next day 7/17 −16.5%). Moreover, in the quota filing the company itself wrote that the quota "is an upper-bound forecast on the basis of potential contracts and orders… the actual revenue recognized will be lower than this quota"; actual transactions with related parties in January–May 2026 were only RMB 181.6 million, annualizing to about RMB 436 million, below the RMB 495.1 million of full-year 2025 — that channel is actually flat to weakening, not accelerating.

---

3. Overall Ranking of Stock-Level Catalyst Strength

In descending order of total score. The board column directly determines the daily price limit; do not read ±20%/±30% stocks with a ±10% intuition. Prices and fund flows are as of the 2026-08-14 close (snapshot timestamps 15:34–16:14, verified as not stale). PE bases are labeled throughout.

---

4. Stock Scoring Model

Total of 100 = news-source authority (0-15) + directness of the catalyst (0-20) + earnings elasticity (0-15) + industry position and fundamentals (0-15) + expectation gap (0-10) + theme durability (0-10) + A-share trading attributes (0-10) + risk deductions (0 to −15)

Notes on scoring basis (seven items, all hard rules this issue):

• Rule one: the expectation gap must deduct the earnings pre-announcement. Scoring on "small gain last Friday = large expectation gap" is a tautology — the interim reports were only disclosed on Saturday, so Friday could not possibly have reflected them. The correct criterion is "where the actual figure landed relative to the pre-announced range": only landing at or above the upper bound counts as a beat; landing at the midpoint is zero surprise. 生益科技 landed at the 94th percentile (6 pts); 中微半导 beat by 4.2% (7 pts); 锐捷 at the 63rd percentile (3 pts); 星网 exactly at the midpoint (5 pts, the points coming from the acquisition); 生益电子 at the 53rd percentile (2 pts).
• Rule two: 盛科通信's expectation gap gets only 2 pts. It is easily read as "the only positive news on the list that has not been priced," but that comes from looking only at the weekend contract filing and not tracing the 7/16 quota filing it cites (no. 2026-021). That quota is RMB 2.2 billion, 191.22% of comparable business, and was approved at the extraordinary shareholders' meeting on 8/6, and the market reacted at the time (7/16 +5.96%, 7/17 −16.5%). "The share price didn't move" does not equal "the information wasn't disclosed."
• Rule three: seal quality is always measured by the seal-order/turnover ratio, never the absolute value. The widely quoted "亨通光电's seal order of RMB 844 million" is an absolute-value measure, and inferring "best seal quality" from it does not hold: ① the largest seal order in the market was 中石科技 (Zhongshi Technology, 300684) at RMB 1.043 billion; ② 亨通's seal-order/turnover ratio is only 0.044, near the bottom among that day's limit-up stocks (中石科技 3.641, 华正新材 0.248, 阿莱德 0.380); ③ 亨通 did not seal for the first time until 13:59:06, broke the seal once, and only finally locked at 14:49:09, placing it squarely in the high-risk "afternoon seal" category named in risk item 6 of §8.4 of this brief. This brief uses the ratio throughout and labels the first-seal time; on that basis 亨通 scores 61.
• Rule four: Dragon-Tiger List "institutional net buying" must be broken out by seat and cannot be replaced by total net buying. In the frequently quoted "3 institutions net bought RMB 242 million" for 中瓷电子, RMB 242 million is the Dragon-Tiger List total net buy: broken out, institution-only seats account for just about RMB 74 million, and the largest single buyer is hot money (Kaiyuan Securities Xi'an Taihua Road branch +RMB 98 million), with Shenzhen Stock Connect +RMB 76 million. Calling this "institution-led" does not hold. On the same basis, 中国稀土's institution-only seats net bought about +RMB 253 million, higher than 中瓷电子's. (Also note: 中瓷电子's Dragon-Tiger List success rate is 42.33%; it is easily confused with the 41.97% on the 中国稀土 row.)
• Rule five: PE must be labeled as either "annualized" or "TTM." The default "PE (dynamic)" in quote software is H1 annualized, which differs substantially from the standard PE-TTM: 生益科技 annualized 52.91 / PE-TTM 66.97; 生益电子 43.72 / 47.32; 锐捷网络 annualized about 115 / PE-TTM 171x. The differences reach 21%–49% and the two cannot be mixed.
• Earnings elasticity is always measured ex-non-recurring, with a check of the ex-non-recurring/attributable ratio. 星网锐捷 was downgraded on this basis (attributable +75.9% but ex-non-recurring only +45.91%, non-recurring items 26.4%).
• Risk deductions now add weights for "operating cash flow" and "PB historical percentile." 锐捷's OCF of −RMB 3.018 billion and 星网's −RMB 3.566 billion are the main new deduction items this issue.

---

5. Detailed Analysis of Key Stocks (descending by total score)

1卫星化学002648Shenzhen main board ±10% | Total 77 | Priority deep-dive · Satellite Chemical

This is the largest ex-non-recurring profit in the window, and the one most easily screened out by the "AI narrative." Why it is easily missed is explained at the end of this section.

• Related news: 2026-08-15 interim report. Revenue RMB 30.713 billion (+30.92%), attributable net profit RMB 6.227 billion (+126.94%), ex-non-recurring RMB 6.053 billion (+109.01%), gross margin 28.09% (full-year 2025: 22.31%). Filing¹
• ✅ It is the only name in this window meeting all four conditions simultaneously:
  1. No earnings pre-announcement → the information is genuinely new (confirmed by a full cross-check with stock_yjyg_em);
  2. Ex-non-recurring/attributable 97.2% → profit contains almost no non-recurring items, one of the cleanest bases in the window;
  3. Q2 single-quarter ex-non-recurring RMB 4.099 billion, QoQ +109.8% (Q1 RMB 1.954 billion) → the single quarter is accelerating; this is not a cumulative-basis illusion;
  4. Genuine zero front-running: 8/14 close −0.79%, main-force −RMB 17 million, turnover just 1.05% — more thoroughly so than 生益科技 (+1.04%) or 生益电子 (+5.59%).
• Magnitude comparison: Ex-non-recurring RMB 6.053 billion is 2.1× 生益科技 (RMB 2.908 billion), 5.4× 生益电子 (RMB 1.111 billion), and 8.9× 锐捷网络 (RMB 677 million). If one reads the weekend interim reports only along the "AI network side," the genuinely largest report is not on the list at all.
• Valuation (an extreme contrast with the AI chain): market cap RMB 85.09 billion. PE-TTM 9.68, at the 15.2 percentile over the past ten years; PB 2.24, at the 41.7 percentile (ten-year median 2.32).

  ⚠️ The "low-PE trap" must be actively tested: low PE paired with a high PB percentile is often the shape of a cycle top. But 卫星化学's PB percentile is 41.7%, in the lower-middle of its own ten-year range, with no divergence of low PE and high PB. So this classic warning does not hold here — that is the conclusion of an active test in this brief, not an assumption that it is cheap.

• ⚠️ But it is a cyclical, and that must be stated up front:
  • Full-year 2025 attributable net profit −12.54% (RMB 5.311 billion); the +126.94% in 2026H1 is a spread reversal, not a sustained growth curve.
  • The core business is ethylene from ethane cracking (C2) and propane dehydrogenation (C3), with profit highly dependent on the ethane-ethylene spread, which is jointly determined by U.S. ethane export prices, freight rates and domestic ethylene demand. This brief was unable to obtain July–August 2026 ethane/ethylene spread data, and therefore cannot judge whether the Q3 spread is widening or narrowing — this is the largest unknown in this item and is flagged as "to be verified."
  • It is entirely unrelated to today's AI mainline and lacks thematic attributes and concentrated money, so a "good earnings but no incremental money" outcome is possible.
• Technical sentiment: Shenzhen main board ±10%, market cap RMB 85.09 billion, limited elasticity at that size. Turnover on 8/14 was just 1.05% — almost nobody trading.
• Final judgment: Measured on this brief's core criterion of "evidence strength × degree of not being priced," it is the highest-scoring name in the whole window — the largest ex-non-recurring profit, the cleanest basis, still accelerating on a single-quarter basis, no pre-announcement, zero front-running, low valuation percentile. Its weakness is not in the statements but in its attributes: it is a cyclical, not a growth stock; α, not a theme. It is given "priority deep-dive," but readers must be clear that this is buying a cheap cyclical income statement, not buying AI.

📌 Why this stock is easily missed entirely: it belongs to "chemical raw materials" and is not in any AI supply-chain stock pool. Any screening process that "picks the industry narrative first, then picks stocks within the narrative" will systematically fail to see it — even though its ex-non-recurring profit is 2.1× the largest name inside the narrative. The correct order is to run the full-market ex-non-recurring ranking first, then cut into narrative branches. Section 6.1 of this brief publishes the complete screening methodology and the exclusion list, so readers can judge the coverage.

2生益科技600183Shanghai main board ±10% | Total 74 | Priority deep-dive · Shengyi Technology

• Related news: 2026-08-15 interim report. Revenue RMB 19.026 billion (+50.05%), attributable net profit RMB 3.287 billion (+130.42%), ex-non-recurring RMB 2.908 billion (+110.99%), gross margin 30.69%.
• ⚠️ Expectation-gap correction: an earnings pre-announcement was already issued 7/14 (attributable +117%~131%), actual +130.4%, landing at the 94th percentile of the range — the upper end, but the magnitude was long since public. What is being priced today is not the size of the profit but the structure of the statements.
• Single-quarter basis (new information, positive): Q2 revenue RMB 10.884 billion (QoQ +33.7%), Q2 ex-non-recurring RMB 1.825 billion (QoQ +68.5%), Q2 single-quarter gross margin 32.63% (Q1 28.10%, +4.53pct). Ex-non-recurring YoY growth rose from +93.5% in Q1 to +122.9% in Q2 — all three metrics accelerating in the same direction.
• The core business is stronger once the subsidiary is stripped out (key): 生益科技 owns 62.50% of 生益电子. Excluding the portion attributable to 生益电子, the CCL core business's Q2 attributable net profit was RMB 1.733 billion, +94.0% QoQ, faster than the +83.9% on a consolidated basis. The engine of this round of profit elasticity is upstream CCL, not midstream PCB.
• Primary evidence for price increases (price/volume breakdown): the interim report discloses CCL sales volume of 87.1401 million square meters, +14.24%, while copper-clad laminate and prepreg revenue was RMB 12.357 billion, +47.74% → a price/mix effect of about +30%, with about 2/3 of revenue growth coming from price increases. By site, Shaanxi Shengyi volume −3.54% and Suzhou Shengyi +1.86%: mature capacity volumes have stalled, and virtually all the increment came from the new Jiangxi capacity (+95.15%). Management's own wording characterizes it as "in the face of continued sharp price increases in upstream raw materials… driving an adjustment in product pricing strategy" — i.e. cost-push pass-through, not pure demand pull.
• Profit composition that must be disclosed: of the gap between attributable net profit of RMB 3.287 billion and ex-non-recurring RMB 2.908 billion, RMB 409.9 million comes from fair-value changes in 联瑞新材's convertible bonds (12.5% of attributable profit, RMB 304 million in Q2 alone). And 联瑞新材 (688300) has fallen from RMB 244.40 on 6/30 to RMB 164.65 on 8/14 (−32.6%), so a substantial portion will very likely reverse in Q3. The ex-non-recurring basis is unaffected, but the headline attributable +130.42% is watered down.

  An intriguing closed loop: the source of that mark-to-market gain is precisely 联瑞新材, the name excluded to this brief's Pass list for weak earnings.

• Cash flow and leverage (the only healthy one on this list): H1 operating cash flow RMB 2.983 billion (+53.4%), Q2 alone RMB 2.412 billion, OCF/attributable = 1.13, with collections keeping pace with profit. Receivables +50.2% and inventory +30.5%, both below revenue growth of +50.05%, with turnover days improving on both. Against 锐捷's −RMB 3.018 billion and 星网's −RMB 3.566 billion, this carries more weight than usual. But gearing rose from 42.29% to 50.33%, finance costs +121.05%, and payables turnover days lengthened by about 23 days — part of Q2's OCF comes from occupying upstream funds.
• Valuation: market cap RMB 347.86 billion. PE-TTM 66.97 (annualized 52.91). PB 18.80, at the 98.8 percentile of all history, 4.60× the historical median (4.09). The current PB/PE implies an ROE of about 28%, while H1 annualized ROE has already reached 35.5% — the current valuation is treating peak ROE as perpetual.
• There is a genuine gap in consensus: consensus 2026E EPS is RMB 2.41 (13 houses, formed before the interim report), and H1 EPS is already 1.36 → implying H2 −22.8% QoQ. If H2 merely repeats Q2 twice, full-year EPS would be RMB 3.12, 29% above consensus. This is where today's real expectation gap lies — not in the profit itself, but in the divergence between sell-side H2 assumptions and the actual Q2 trajectory. (Adjustment needed: H1 EPS includes about RMB 0.169/share of 联瑞 convertible-bond gains.)
• Technical sentiment: Shanghai main board ±10%. On 8/14 only +1.04%, turnover 2.79%, main-force net outflow of RMB 414 million. The share price has already retraced 23.5% from its 6/25 high of 187.20, and the main advance (80.36 on 5/6 → 187.20 on 6/25, +133%) happened before the pre-announcement.
• Final judgment: The fundamentals are the most solid on the list and it is the only one with healthy cash flow. The risk is not in earnings but in the valuation position (PB at the 98.8 percentile) and the cyclical nature — profit depends heavily on price increases rather than volume, mature sites' volumes have stalled, and the whole industry is expanding in unison. It is given "priority deep-dive" because the consensus gap genuinely exists, not because of "zero front-running."

3中微半导688380STAR Market ±20% | Total 74 | Watch closely · SinoMicro Semiconductor

• Related news: interim report disclosed 2026-08-17. Revenue RMB 726 million (+44.01%), attributable net profit RMB 172 million (+98.48%), ex-non-recurring RMB 159 million (+109.74%), gross margin 38.75% (up from 34.28% in 2025).
• Expectation gap: the 7/23 pre-announcement guided attributable net profit of "about RMB 165 million" (around +90.82%); actual RMB 172 million, a 4.2% beat. Among all companies in this window that had pre-announced, it is the only one whose actual figure clearly exceeded the pre-announcement. Moreover, ex-non-recurring growth (+109.74%) exceeds attributable growth (+98.48%), so core-business improvement is stronger than the headline.
• Single quarter: Q2 ex-non-recurring RMB 90 million, QoQ +30.4%, with gross margin rising.
• Stage of the industry branch: unrelated to today's mainline — 中微半导 is MCU/signal chain, with end-markets in home appliances and automotive electronics, not the AI chain. It is an earnings-α name, not a thematic name; do not map it to AI simply because it is a "semiconductor."
• Technical sentiment: STAR Market ±20%, market cap of only RMB 16.13 billion, PE 46.98. 8/14 +2.39%, main-force +RMB 1 million.
• Final judgment: the only name that beat its pre-announcement + small cap + ±20% elasticity + rising gross margin. The risk is the lack of thematic attributes, which could mean "good earnings but no buyers."

4锐捷网络301165ChiNext ±20% | Total 66 | Watch closely · Ruijie Networks

This item contains two important pieces of new information, one positive and one negative, pointing in opposite directions; both must be read together.

• Related news: 2026-08-15 interim report. Revenue RMB 8.832 billion (+32.83%), ex-non-recurring RMB 677 million (+56.56%), H1 gross margin 31.61%.

• Small expectation gap: already pre-announced 7/02 with attributable net profit of RMB 600–750 million (+32.71%~65.88%); actual RMB 694 million, landing at the 63rd percentile of the range — no surprise.

• ✅ Positive new information: the gross margin mystery is solved; it is a mix shift, not a price war. First a correction of basis — I previously compared "Q1's 35.70%" directly with "H1's 31.61%," which is a quarter against a half-year and understated the drop. Backing out the single quarter: Q2 single-quarter gross margin was 29.51%, so Q1→Q2 was actually −6.19pct (worse than the surface figure), yet Q2 single-quarter net margin rose +5.69pct to 9.79%. The interim report's channel breakdown gives the answer:

  By sales model	26H1 revenue	YoY	Gross margin	Gross margin YoY
  Direct sales (carriers + large internet firms)	RMB 5.915 billion	+51.27%	25.72%	−2.41pct
  Distribution (government/enterprise, education, healthcare)	RMB 2.917 billion	+6.49%	43.58%	+3.17pct
  Overseas	RMB 2.118 billion	+84.95%	31.24%	−10.07pct

  Table scrolls horizontally

  Attribution: gross margin 33.19%→31.61% (−1.58pct) = channel mix change −1.00pct + within-channel changes −0.58pct. Direct sales rose from 58.8% to 67.0% of revenue, while the two channels' gross margins differ by 17.9pct. Three independent pieces of evidence point to "mix" rather than "price cuts": ① distribution (the base business) gross margin rose +3.17pct against the trend — impossible in an industry-wide price war; ② domestic gross margin +0.23pct, essentially flat; ③ the drag is concentrated in direct sales and overseas. One common assumption also needs clearing up: the words "white box" and "ODM" do not appear anywhere in the interim report. What the company disclosed is own-brand products plus some outsourced manufacturing, with direct-sales customers named as China Mobile, Alibaba, ByteDance, Baidu and Tencent. The low gross margin comes from the business form itself: "large direct-sales customers + high-speed switches with a high BOM share."

• ❌ Negative new information: cash flow is the most jarring line in this report (entirely absent from the pre-announcement).

  Net cash flow from operating activities −RMB 3.018 billion (prior-year period −RMB 314 million, −861.51%); operating cash flow per share −RMB 2.71 (prior-year period −RMB 0.39); attributable net profit RMB 694 million → profit-to-cash conversion −435% (Same basis: 星网锐捷's consolidated OCF is −RMB 3.566 billion, −RMB 4.69 per share, versus −RMB 0.81 in the prior-year period.)

  Line item	End of 2025	End of 2026H1	Change
  Accounts receivable	RMB 1.856 billion	RMB 3.677 billion	+RMB 1.821 billion
  Inventory	RMB 2.960 billion	RMB 5.614 billion	+RMB 2.654 billion
  Prepayments	RMB 44 million	RMB 1.433 billion	+RMB 1.389 billion (+3,125%)
  Short-term borrowings	RMB 1.301 billion	RMB 4.728 billion	+RMB 3.427 billion
  Gearing ratio	51.20%	62.92%	+11.7pct

  Table scrolls horizontally

  The three working-capital items absorbed a net +RMB 5.86 billion, funded by +RMB 3.43 billion of short-term debt plus RMB 1.59 billion of cash burn. Two mutually exclusive explanations, which the statements themselves cannot distinguish: ① stocking chips/optical modules ahead of already-locked orders (active expansion after demand is confirmed); ② receivables growth of +98% far outpacing revenue's +33%, DSO rising from 46.3 days to 56.4 days, with RMB 47.78 million of inventory write-downs and RMB 30.57 million of bad-debt provisions already booked. The interim report does not disclose the order backlog or customer concentration, so it cannot be confirmed that the stocking is covered by orders. There is one certain burden: the RMB 4.728 billion of short-term debt only built up during the reporting period, and H1 finance costs were only RMB 30 million (+607.55%); from H2 interest will accrue on the full balance.

• Profit quality: non-recurring items are only RMB 17.14 million (2.5% of attributable profit), a clean basis. But the software VAT rebate of RMB 122 million is 18% of the RMB 677 million ex-non-recurring profit, and is essentially a policy item.

• Valuation: market cap RMB 160.252 billion. PE-TTM 171x (the previously mislabeled 115 was the annualized basis), PS 9.7x, PB 30.03. Already +114% from 6/25 to 8/14.

• Final judgment: the gross margin doubt has been cleared up (a positive), but it has been traded for a bigger doubt — cash flow. It is given "watch closely" rather than anything higher, with the risk deduction raised from −10 to −13.

5生益电子688183STAR Market ±20% | Total 63 | Watch only · Shengyi Electronics

• Related news: 2026-08-15 interim report. Revenue RMB 5.784 billion (+53.46%), ex-non-recurring RMB 1.111 billion (+110.29%), ex-non-recurring/attributable 100%, gross margin 34.31%.
• Smallest expectation gap: already pre-announced 7/14 with ex-non-recurring +105%~115%; actual +110.3%, landing at the 53rd percentile of the range — precisely the midpoint, virtually zero surprise.
• Three new negative structural data points:
  1. Q2 single-quarter gross margin turned negative QoQ: 35.21% (Q1) → 33.67% (Q2), −1.54pct. The 35.21% in 2026Q1 was its highest single quarter on record, and Q2 has already pulled back. This is the exact opposite of the parent's Q2 gross margin move of +4.53pct.
  2. Collection quality dropped sharply: the Q2 cash-collection ratio fell from 100% in Q1 to 66%; Q2 operating cash flow was RMB 302 million, −50.9% QoQ, while attributable net profit rose +49.7% QoQ over the same period. Virtually all of the 39.9% QoQ revenue increment in Q2 settled into receivables (+RMB 1.063 billion).
  3. Ex-non-recurring YoY growth is decelerating: Q1 +122.8% → Q2 +102.8% (partly a base effect).
• Mechanical explanation: 生益科技 holds 62.50%, and 生益电子 purchases about RMB 585 million of copper-clad laminate from the 生益 group (16.26% of its purchases of that category), with intra-group eliminations doubling to −RMB 640 million. The parent raising prices = the subsidiary's costs rising. Group profit is migrating from midstream PCB to upstream CCL.
• The consensus bar is too high: consensus 2026E EPS is RMB 3.15 (only 3 houses covering), and H1 is already 1.35 → implying H2 must accelerate a further +33.3% QoQ, while Q2 gross margin has just turned negative QoQ. Downward-revision risk exceeds upward-revision potential.
• The company's own risk warning (verbatim): "The PCB industry is aggressively adding capacity… the medium-to-long-term expansion of industry supply may intensify competition and trigger downward pressure on product prices."
• Valuation: market cap RMB 97.16 billion. PE-TTM 47.32 (annualized 43.72). PB 14.92, at the 82.8 percentile of the past three years, 4.99× the historical median (2.99).

  ⚠️ Its PE percentile is only 10.8% (past three years), which looks cheap on the surface, but the 2023 loss makes the PE series contain negative and extreme values, so that percentile is statistically unreliable. The PB percentile is the honest reading — and it is exactly the classic shape of "low PE percentile + high PB percentile."

• Technical sentiment: STAR Market ±20%; note that +12% is not a limit-up. On 8/14 it was already +5.59% with main-force +RMB 173 million — already front-run.
• Final judgment: the prettiest headline earnings, but it is the window's combination of "smallest expectation gap + most negative new information." Given watch only.

6中天科技600522Shanghai main board ±10% | Total 62 | Watch closely · Zhongtian Technology

• Related news: policy continuation, no new company filings. 8/14 +8.05% without a limit-up, main-force +RMB 1.526 billion.
• Why it still scores relatively high: "money without a limit-up" is the healthiest shape for a relay position — there is no overhang from unlocked limit-up chips and no chasing cost of a one-line board.
• Fundamentals: PE 33.01/PB 3.27, among the lowest valuations on the entire list. The interim report comes on 8/28, among the latest in this mainline.
• Final judgment: the best position in the mainline's second tier, but earnings verification is 12 days away, and until then only mainline sentiment can hold it up.

7星网锐捷002396Shenzhen main board ±10% | Total 62 | Watch closely · Star-net Ruijie

• Related news: ① 8/15 interim report, attributable net profit RMB 373 million (+75.90%); ② same-day filing on the RMB 395.6 million acquisition of 46% of 广州芯德 (Guangzhou Xinde) → 67% ownership, appraisal premium 165.11%.
• ⚠️ Attributable and ex-non-recurring differ by 30pct, and the pre-announcement spelled it out long ago: ex-non-recurring is only RMB 274.16 million, +45.91%; non-recurring items are RMB 98.54 million, 26.4% of attributable profit, of which fair-value changes in industry funds of +RMB 81.45 million is the single largest. And the 7/03 earnings pre-announcement had already explicitly stated "non-recurring items expected at RMB 100 million to 120 million" and explained why. The actual figure is even slightly below the lower bound of that guidance. If the market looks only at the +75.9% attributable figure, it is reading a number the pre-announcement already made clear — and reading it on the wrong basis.
• Structural facts (more extreme than expected): 星网锐捷 owns 44.88% of 锐捷网络. Item-by-item reconciliation:

  锐捷 attributable RMB 694.19mn × 44.88% = RMB 311.55mn → 83.6% of 星网's attributable RMB 372.71mn
  锐捷 ex-non-recurring RMB 676.86mn × 44.88% = RMB 303.77mn → yet 星网's ex-non-recurring is only RMB 274.16mn
  ∴ 星网's ex-non-recurring "excluding 锐捷" ≈ −RMB 29.61mn (a recurring loss)
  

  Cross-checked against the product breakdown: enterprise networking equipment (≈锐捷) revenue of RMB 8.811 billion is 82.12% of the total; non-锐捷 segments total RMB 1.918 billion, −21.0% YoY. In the consolidated statements, minority interests of RMB 387 million are larger than attributable profit of RMB 373 million.

  Conclusion: 星网锐捷 is not "a parent plus a subsidiary, two logics," but "a 44.88% stake in 锐捷 + a shrinking, recurringly loss-making legacy business + about RMB 100 million of one-off mark-to-market gains."

• ⚠️ An important correction on the "holding-platform discount": 44.88% × 锐捷's market cap of RMB 160.252 billion = RMB 71.9 billion, while 星网's own market cap is only RMB 24.847 billion, an apparent discount of 65.5%. But this discount is not realizable — 锐捷's genuine free float is only about 10.52% (星网's 44.88% + the employee holding platform 厦门锐进东方's 43.12% = 88.0% locked up, with the top ten holders totaling 89.48% and actual free-float market cap only about RMB 16.9 billion). Valuing a 44.88% block stake using a price set by a 10% float is invalid in itself. Both companies' ultimate controller is the Fujian provincial SASAC.
• Materiality of the acquisition (on three scales): for the group it is not material (RMB 50.25 million / 2025 attributable net profit of RMB 409 million = 12.3%; the filing explicitly states it "does not constitute a material asset restructuring" and requires no shareholders' meeting); but for the non-锐捷 segment it is decisive — it can flip that block from "revenue −21% + recurring loss" to "growth + profit." Four easily missed details: ① the target's profit is declining (2025 net profit RMB 65.91 million < 2024's RMB 70.59 million), while the committed first-year RMB 75 million requires about 14% growth; ② the first year carries an 80% tolerance band, so actual net profit ≥ RMB 60 million triggers no compensation — that floor is below the target's actual 2025 net profit, making the effective constraint weaker than it looks; ③ the target's owners' equity fell from RMB 413.52 million to RMB 324.22 million (down RMB 89.30 million) in the two months before the valuation base date, mechanically inflating the appraisal premium; the filing does not explain why, to be verified; ④ remeasuring the originally held 21% stake at the acquisition date will produce more than RMB 40.90 million of one-off non-recurring gains, which are non-cash and non-repeatable. Also note: the performance commitment is on a "first/second year after closing completes" basis, and the transaction has not yet closed. If closing straddles the year-end, the 2026 statements will contain no committed earnings, only the remeasurement gain.
• Cash flow: consolidated OCF −RMB 3.566 billion, of which 锐捷 accounts for −RMB 3.018 billion (85%). The RMB 395.6 million consideration is paid entirely in cash from own funds, against RMB 1.745 billion of cash at end-H1. Estimated goodwill is about RMB 360 million, roughly equal to a full year of the group's attributable net profit.
• Valuation: market cap RMB 24.847 billion, PE-TTM 43.6x, PB 3.43, PS 1.18x — still the lowest on this chain.
• Final judgment: the lowest valuation and the acquisition is entirely new information, but the "discount arbitrage" logic does not hold, and ex-non-recurring growth is far below attributable growth. Given "watch closely," scored 62.

8亨通光电600487Shanghai main board ±10% | Total 61 | Watch closely · Hengtong Optic-Electric

• Related news: limit-up on 8/14, turnover of RMB 19.197 billion (first in the market), main-force +RMB 3.121 billion, seal order RMB 844 million. The driver was directional policy wording.

• Key reason for the downgrade: the earnings elasticity was already announced on 2026-07-14 — H1 attributable net profit of RMB 3.016–3.568 billion (+86.94%~+121.20%), ex-non-recurring +94.14%~+129.32%, with the filing explicitly stating "benefiting from the rapid development of artificial intelligence and continued increases in data-center capex, optical communications market demand has grown and prices have risen." That is, the earnings were known a month ago, and the 8/14 limit-up was policy-driven, not earnings-driven. The formal interim report is not due until 8/27.

• One key item that cannot be verified: 亨通光电 did not appear on the Dragon-Tiger List on 8/14 (it was not in the top 5 by price-move deviation that day), so the institutional-versus-hot-money split within the RMB 19.197 billion of turnover cannot be determined. Last week's recap listed this as a must-check item for Monday — this brief states it candidly: the data does not exist; it is not that we failed to find it, but that the listing criteria were not met and the exchange never disclosed it.

• ⚠️ Seal quality: a widely misread number: the claim "seal order of RMB 844 million, the thickest in the market" fails on both counts —

  8/14 limit-up stocks	Seal order	Turnover	Seal/turnover	First seal	Final seal	Seal broken
  中石科技 (Zhongshi Technology, 300684)	RMB 1.043 billion	RMB 286 million	3.641	09:25:00	——	0
  华正新材 (Huazheng New Material, 603186)	RMB 382 million	RMB 1.539 billion	0.248	09:37:30	09:37:30	0
  阿莱德 (Allied, 301419)	RMB 166 million	RMB 438 million	0.380	10:31:42	10:31:42	0
  亨通光电 (Hengtong Optic-Electric, 600487)	RMB 844 million	RMB 19.197 billion	0.044	13:59:06	14:49:09	1
  中瓷电子 (CETC Chip Technology, 003031)	RMB 162 million	RMB 1.632 billion	0.099	14:26:30	14:26:30	0

  Table scrolls horizontally

  ① The largest absolute seal order was 中石科技's RMB 1.043 billion, not 亨通's; ② 亨通's seal/turnover ratio of just 0.044 is near the bottom among that day's limit-up stocks; ③ 亨通 did not seal for the first time until 13:59, broke the seal once mid-session, and only finally locked at 14:49 — squarely in the "afternoon seal, more prone to gap up and fade the next day" category named in risk item 6 of Section 8.4.

  Absolute seal-order size is distorted as turnover grows; only the ratio is comparable. This brief uses the ratio plus first-seal time for every limit-up stock, and 亨通 scores 61 on that basis.

• Technical sentiment: Shanghai main board ±10%. PE 35.13/PB 4.89, the lowest on the entire list, and that does not change; but both seal quality and first-seal time are below average, and whether the first board can turn into consecutive limit-ups is unverified.

• Final judgment: valuation and industry position are still the best in the mainline, but the popular claim of "best seal quality" does not stand up. Its role today is "the indicator stock for whether the mainline holds," and its own seal pattern has already given a somewhat negative reading.

9中瓷电子003031Shenzhen main board ±10% | Total 49 | Watch only · CETC Chip Technology

• Related news: none. Limit-up on 8/14, but active verification shows the company's most recent filing is a cash-management progress announcement dated 2026-07-29; there were no company filings whatsoever within the window, and no earnings pre-announcement.

  Per last month's lesson, "no news found" ≠ "no news"; limit-up stocks must have their filings actively checked. This item is confirmed to have no filings after active verification, not merely a search miss.

• ⚠️ Fund-flow structure: another easily misread basis: the common claim is "3 institutions on the Dragon-Tiger List bought, net RMB 242 million, low turnover + institution-led." But RMB 242 million is the Dragon-Tiger List total net buy, not institutional net buying. Broken out by seat:

  • Kaiyuan Securities Xi'an Taihua Road branch bought RMB 98 million / sold 0 → net +RMB 98 million (a textbook hot-money "top-seat-dominated" structure, and the largest single buyer)
  • Shenzhen Stock Connect seat net +RMB 76 million (northbound, not institutional)
  • Institution-only seats total net about +RMB 74 million, only about 31% of total net buying

  So "institution-led" does not hold. If "institutions vs hot money" is the distinguishing standard, then on the same basis 中国稀土 (China Rare Earth, 000831) saw institution-only net buying of about +RMB 253 million, far higher than 中瓷电子; and the very same Kaiyuan Securities Xi'an Taihua Road seat bought both stocks that day, a linkage worth noting. (Also note: 中瓷电子's Dragon-Tiger List success rate is 42.33%, easily confused with the 41.97% on the 中国稀土 row.)

• The part that objectively still holds: it sealed the limit-up on turnover of just 3.84%, so chip lock-up really is high. But it did not seal for the first time until 14:26:30, also an afternoon seal.

• Fundamentals: electronic ceramic packages/gallium nitride, part of the CETC system. 2026Q1 revenue RMB 1.099 billion (+79.05%), attributable net profit RMB 193 million (+57.32%). The interim report is scheduled for 8/25 and has not yet been disclosed.

• Final judgment: the technical pattern and fund-flow structure are among the healthiest on the list, but the catalyst is entirely unidentified. Under the rule "do not chase an unconfirmed catalyst," it is given watch only. If a filing or research-visit note emerges today explaining the move, it can be upgraded immediately.

10太辰光300570ChiNext ±20% | Total 43 | Watch only · T&S Communications

• Related news: on 8/15 it released two pieces of information pointing in opposite directions, and since there was no prior earnings pre-announcement, both are new information —
  • Positive: plans to spend RMB 720 million to buy land and build a Pingshan production base (17,200 m² of land, 68,700 m² of floor area).
  • Negative: interim revenue RMB 1.001 billion (+20.81%) but attributable net profit RMB 176 million, up only +1.71%, ex-non-recurring RMB 164 million −2.55%, operating cash flow −RMB 73 million (prior-year period +RMB 128 million, −156.88% YoY).
• A textbook "revenue up, profit flat + cash flow turning negative": revenue +20.81% while ex-non-recurring profit went backwards; operating cash flow swung from +RMB 128 million to −RMB 73 million, a swing of RMB 201 million, exceeding its entire half-year net profit.
• Three layers of uncertainty around the expansion (per the filing itself): the land must be obtained through public auction, and whether it can ultimately be won remains uncertain; the project may be delayed, changed or suspended; the amount is an estimate. The "positive" itself has not yet landed.
• Technical sentiment: ChiNext ±20%. On 8/14 it was already +7.87% with main-force +RMB 320 million — front-run before the news came out, and the interim report released afterwards was negative. PB 21.99.
• Final judgment: the market today will most likely read only "optical communications + RMB 720 million expansion," while the interim report's ex-non-recurring profit and cash flow point the other way. This is the name this brief considers most likely to be misread.

11盛科通信-U688702STAR Market ±20% | Total 41 | Watch only · Centec Communications-U

⚠️ This contract is extremely easy to read as "the only positive catalyst in the whole window that has not been priced" — this brief reaches the opposite conclusion (41 pts, watch only). Six reasons follow.

• Weekend filing: "Announcement on Signing a Major Ordinary-Course Business Contract Constituting a Related-Party Transaction." The subject is Ethernet switch chips; the amount "exceeds 50% of the audited revenue of the most recent fiscal year and exceeds RMB 100 million" (implying a floor of about RMB 576 million based on 2025 revenue of RMB 1.151 billion); the counterparty is 深圳中电港 (Shenzhen CECport, 001287), an entity controlled by CEC, a shareholder holding more than 5%.
• ❌ Downgrade reason one: this is not new information. The contract filing itself states that the company's board approved the "Proposal on Increasing the Estimated Ordinary Related-Party Transaction Quota for 2026" on July 14, 2026, and that it was approved at the extraordinary shareholders' meeting on August 6, 2026: "this contract… falls within the estimated scope of the company's annual ordinary related-party transactions, and its signing does not require submission to the board or the shareholders' meeting for approval." Tracing back to the quota filing of July 16 (no. 2026-021): the estimated 2026 quota for selling goods to CEC and its subsidiaries was raised from RMB 600 million by RMB 1.6 billion to RMB 2.2 billion, equal to 191.22% of comparable business. In other words, the magnitude information was public 5 weeks ago.
• ❌ Downgrade reason two: the market already reacted at the time — up first, then a crash. On 2026-07-16 (the quota filing date) it closed at 373.50 (+5.96%), and the next day, 7/17, it closed at 312.00 (−16.5%). From 389.99 on 6/30 to 398.60 on 8/14, just +2.2% over six weeks; over the same period 锐捷 rose +79% and 星网 +70%. "The share price didn't move last Friday" does not equal "the information wasn't priced" — it was priced a month ago.
• ❌ Downgrade reason three: the company itself discounted it in the quota filing (verbatim). "The increase in the estimated ordinary related-party transaction quota is an upper-bound forecast on the basis of potential contracts and orders, not on a revenue-forecast basis; the revenue actually recognized from sales of goods to related parties in 2026 will be lower than this quota." The contract filing likewise reiterates that "the impact on the company's 2026 operating results is relatively limited."
• ❌ Downgrade reason four (the most counterintuitive number): that channel is actually flat to weakening, not accelerating. The quota filing discloses that actual transactions with related parties in January–May 2026 were only RMB 181.6 million, annualizing to about RMB 436 million, below the RMB 495.1 million of full-year 2025. That is, in the five months before the quota was sharply raised, actual volume through that channel was flat to weaker year on year.
• ❌ Downgrade reason five: the counterparty is a distributor, as its own financials make clear. 中电港's 2025 revenue was RMB 65.525 billion with net profit of RMB 284 million → a net margin of 0.43%. A 0.43% net margin is the textbook shape of component distribution. So this contract is selling chips into the channel, not to end customers; who 中电港 ultimately sells to and how much is not disclosed and remains unknown. Moreover, 中电港 has accounted for 33.15%/26.88%/32.15% of the company's volume in this business over the past three years — it was already the largest channel, not a new customer.
• ❌ Downgrade reason six: the Big Fund is selling down and is still within its disposal window. The National IC Industry Investment Fund sold 4.1 million shares via centralized bidding between 2026/7/21 and 8/6, cutting its stake from 11.00% to 10.00% (filing 2026-024). The filing states verbatim: "following this change in equity interest, the disclosure obligor remains within the implementation period of its disposal plan."
• Extreme valuation: market cap RMB 163.426 billion, PS (on 2025 revenue) 142.0x, PB 73.2x, with negative PE (TTM attributable −RMB 152 million). To compress PS to 20x, revenue would have to grow from RMB 1.15 billion to about RMB 8.2 billion (7.1×); and even at RMB 8.2 billion of revenue with a 15% net margin, a PE of 133x would still be needed to justify the current market cap. The current price requires both a step-change in revenue magnitude and a swing from loss to a high net margin to be delivered simultaneously.
• The interim report has not yet been disclosed and is scheduled for 8/27. For Monday's pricing, the most recent confirmed financial facts remain 2026Q1: revenue RMB 248 million (+11.40%), ex-non-recurring −RMB 29 million.
• Also note: on 2026-07-01 the company voluntarily issued a share-price risk warning (a 38.53% gain in June versus 21.72% for the STAR composite index) and reiterated that "market resilience is insufficient."
• Final judgment: looking only at that weekend contract filing, it is easy to treat it as "the only unpriced positive catalyst in the whole window"; but the moment you open the 7/16 quota filing it cites, the conclusion reverses completely. In sum: the information is not new, the market has already reacted, the company itself has cooled it down, actual channel volume is weakening, the Big Fund is selling down, and valuation is PS 142x. Given "watch only."

---

6. Pass List

A note on the nature of the rare-earth Pass: being on the Pass list does not mean a judgment that it will fall.

• A common misjudgment is to say its "strength is real but the reason is empty" — that is inaccurate. Its 8/08 interim report shows gross margin +5.76pct and ex-non-recurring +55.49%, so a fundamental reason does exist; it simply falls outside the window that this brief starts at 15:00 on 8/14.
• And on the corrected seat basis, 中国稀土's institution-only net buying of about +RMB 253 million is higher than that of 中瓷电子 (about +RMB 74 million), which this brief once characterized as "low turnover + institution-led." The two should be handled to the same standard; this brief unifies them as "new information falls outside the window → neither enters the recommendation list, both are listed for continued observation."

6.1 Full Screening Methodology for the Weekend's 136 Interim Reports, and Exclusion Notes

Any stock-selection process that "cuts to an industry narrative first, then picks stocks within the narrative" will systematically miss names outside the narrative — 卫星化学 (002648), the largest ex-non-recurring profit in this window, is exactly such a case. The complete screening methodology and known defects are published below so readers can judge the coverage.

Screening universe: the 136 Shanghai/Shenzhen/Beijing A-share companies that disclosed 2026 interim reports between 2026-08-15 and 08-17 (data source: Eastmoney RPT_LICO_FN_CPD, using NOTICE_DATE ≥ 2026-08-15).

What this brief actually screened: ① the top 40 by descending YoY growth in attributable net profit; ② 52 companies filtered by technology/electronics industries (telecom equipment, components, semiconductors, optoelectronics, computer hardware, etc.), each read individually; ③ for every name that made it into the main text, supplementary checks of ex-non-recurring profit, single-quarter breakdown and earnings pre-announcements.

Known methodological defects (stated candidly):

1. Step ① looked only at the descending attributable-growth ranking, so negative-growth companies are inherently out of view. 网宿科技 (300017, attributable −6.61%) was therefore very easy to miss — and it was the market's most money-concentrated stock last Friday. Ranking is screening, and a descending list cannot show you the negatives.
2. Step ② screened by industry, excluding non-technology industries wholesale. 卫星化学 belongs to "chemical raw materials" and was therefore not among the 52 — yet its ex-non-recurring profit (RMB 6.053 billion) is the largest in this window.
3. This brief was unable to pull ex-non-recurring figures for all 136 companies (Eastmoney's batch ex-non-recurring API RPT_DMSK_FN_MAINFINADATA returned empty, so figures were pulled name by name, and time did not permit full coverage). Names of the "attributable soaring but ex-non-recurring going backwards" type may therefore still have been missed — three have been confirmed: 聚辰股份, 芯朋微 and 中利集团.

Other weekend names confirmed after supplementary checks but not entering the recommendation list:

The correct order adopted by this brief: run the full-market ex-non-recurring ranking first (three columns: ex-non-recurring growth + ex-non-recurring/attributable + Q2 single-quarter QoQ), then cut into narrative branches — not the other way round.

---

7. Within-Branch Rankings

7.1 AI Copper-Clad Laminate / PCB

⚠️ Parent and subsidiary: 生益科技 owns 62.50% of 生益电子, already consolidated (30.4% of revenue, 20.1% of attributable profit). The two are not independent logics, and they point in opposite directions — the parent raising prices means the subsidiary's costs rise.

• The most important conclusion in this branch: under the single label "AI materials/PCB," the four names' interim reports are +110.99%, +110.29% (structurally weakening), +4.42% and +6.26% respectively. A label is no substitute for verifying each name.

7.2 Data-Center Switches / Network Equipment

⚠️ 星网锐捷 owns 44.88% of 锐捷网络, and 83.6% of its attributable profit comes from that stake. The two are two prices for the same asset; one should be chosen rather than holding both. ⚠️ The "discount arbitrage" logic does not hold: 锐捷's genuine free float is only about 10.52%, so valuing a 44.88% block stake at the price set by a 10% float is invalid in itself. ⚠️ A structural observation running through this branch: 锐捷's OCF of −RMB 3.018 billion and 星网's −RMB 3.566 billion, plus 盛科 being stocked by its channel — the AI network side's current high growth has the shape of "stock up first, pre-fund first," with the supply chain pushing cash pressure from downstream to the mid- and upstream. It magnifies profit when demand materializes, and simultaneously exposes inventory write-downs, bad debts and interest when demand weakens at the margin. The interim reports cannot distinguish which one it is.

7.3 Optical Communications / Optical Fiber & Cable (prior day's mainline)

• A verification gap that must be stated: 剑桥科技's main-force fund flow is +RMB 1.254 billion, but Dragon-Tiger List net buying is only RMB 24 million. The two bases are computed differently and cannot simply be subtracted, but a divergence of this size indicates that its buying is highly dispersed rather than led by concentrated money, hence the downgrade.

---

8. Today's Opening Verification Signals

Per last week's lesson, this section writes the verification points first and lets them determine the conclusions. Two new meta-rules this issue: ① for any company that has already issued an earnings pre-announcement, a "gap up" should not be read as an earnings beat; ② "the share price didn't move" does not equal "the information wasn't disclosed" — you must trace back to the first disclosure date (盛科通信 is the price paid for this rule).

8.1 Call-Auction Signals (09:15–09:25)

The gap-up gate stays at +5%, but it must be adjusted for the board's price limit: +5% on a ±20% board is equivalent to +2.5% on a ±10% board.

8.2 Sector Signals

• Bullish confirmation: ① three or more rapid limit-ups appear in the components/PCB sector; ② telecom equipment (SW tier-2) sees net main-force inflow for a second consecutive day (8/14 was +RMB 11.873 billion); ③ 生益科技 outperforms 生益电子 (verifying "profit migrating upstream").
• Bearish confirmation: ① 网宿科技 gaps down and keeps weakening, dragging the whole IDC/compute-hosting chain; ② 锐捷网络 gaps up and fades (the cash flow has been understood); ③ Moutai's negative-growth interim report drags on the SSE 50.
• Neutral/rotation: optical communications closes down while copper-clad laminate rises — i.e. "a change of link within the mainline."

8.3 Stock-Level Signals

• First 30 minutes of follow-through: if 生益科技 and 中微半导 hold above their opening prices, the structural information has been accepted.
• Parent/subsidiary divergence: the relative strength of the two pairs 生益科技 vs 生益电子 and 星网锐捷 vs 锐捷网络 is today's cleanest verification pair — if money is really reading statement structure rather than labels, clear divergence should appear; if each pair rises and falls together, the market is still buying labels.
• 中瓷电子 (003031): if low turnover + institutional net buying appears again, its unidentified catalyst may be genuine industry information, and the label can be upgraded.
• Seal quality: for any limit-up, look at the seal-order/turnover ratio; only a ratio above 1 is a real seal (last week 蓝盾光电's seal order of RMB 31 million looked sizeable, but the ratio was only 0.01).

8.4 Risk Signals

1. The pricing process for 网宿科技 is today's most important risk thermometer. It was last Friday's most money-concentrated stock and it holds an un-pre-announced, negative-growth interim report. Its trajectory will directly reflect how fast thematic money reacts when it meets falsification.
2. The AI network side's cash flow is a newly added systemic risk item this issue. 锐捷 −RMB 3.018 billion, 星网 −RMB 3.566 billion; if concentrated discussion of "AI hardware companies' cash flow" emerges intraday today, the whole chain could come under pressure — and it happens to be one of the directions where money was most concentrated last Friday.
3. Gap up and plunge: last Friday the whole market's average open→close was −0.02%, and only 40.7% of stocks had a positive open→close; "heading down after the open" is the current norm.
4. The relay effect has deteriorated for several sessions running: last Friday the "yesterday's consecutive-limit-up" index was −2.23% and "yesterday's second board or higher" was −4.91%; among the 11 consecutive-limit-up stocks, 蓝盾光电 (5 boards), 华西股份 (3 boards) and 共进股份 (2 boards) all had poor seal quality.
5. The core name failing to follow: if 亨通光电 breaks its board, the whole optical communications chain loses its height (of the 6 telecom limit-ups last Friday, 5 were first boards).
6. Index drag: 贵州茅台's H1 attributable net profit was −1.95%, and the SSE 50 was already −0.41% last Friday.
7. Overnight risk from afternoon seals: last Friday 32% of limit-ups were only sealed after 13:00, which are more prone to gapping up and fading the next day.
8. A medium-term risk not yet priced: 生益科技's Q2 attributable profit includes RMB 304 million of 联瑞 convertible-bond gains, while 联瑞新材 has already fallen 32.6% since 6/30. A Q3 reversal will affect the Q3 report's attributable profit (not ex-non-recurring). It is not a disturbance today, but "attributable +130%" should not be used as a valuation anchor.

---

9. Final Conclusions

9.1 The 5 Stocks Most Worth Watching Today

Four structural notes (informational, not operating advice):

1. The 5 names above correspond to 5 independent logics: C2 chemicals, copper-clad laminate, MCU, switches, optical fiber & cable. Parent/subsidiary de-duplication has been completed — 生益科技 owns 62.50% of 生益电子 (the latter is listed "watch only" and is not in the table); 星网锐捷 owns 44.88% of 锐捷网络, and their profits overlap heavily, so they are counted as a single logic for dispersion purposes (星网's PE 43.6x/PB 3.43 is lower and the 芯德 acquisition is entirely new information, but ex-non-recurring growth is only +45.91% and, excluding 锐捷, there is a recurring loss).
2. ⚠️ 亨通光电 (600487) is not in this table; it ranks 8th overall (61 pts). The widely quoted "seal order of RMB 844 million, thickest in the market, best seal quality" does not hold: the largest seal order in the market was 中石科技's (RMB 1.043 billion), 亨通's seal/turnover ratio is only 0.044, and it did not seal until 13:59, broke the seal once, and only locked at 14:49, placing it in the high-risk "afternoon seal" category named in Section 8.4. See 5.8.
3. ⚠️ 盛科通信 (688702) is not in this table; it ranks 15th overall (41 pts). It is easily read as "the only unpriced positive catalyst in the whole window," but its RMB 2.2 billion related-party transaction quota was announced 7/16 and approved by the shareholders' meeting 8/6, and the market reacted at the time (7/17 −16.5%). See 5.11.
4. This table is a watch and research list; it contains no position, trading or holding instructions. All "choose one/de-duplicate" wording above refers to how statistical dispersion is calculated and constitutes no allocation advice.

9.2 Today's 3 Strongest Branches

Data-center switches drop from the 2nd to the 5th branch, not because of earnings but because of cash flow: 锐捷's OCF of −RMB 3.018 billion (−RMB 2.71 per share) and 星网's −RMB 3.566 billion (−RMB 4.69 per share), plus sharply higher gearing. This was newly disclosed over the weekend and was entirely absent from the earnings pre-announcements.

9.3 Directions with Insufficient Evidence / Concentrated Risk Today

1. 网宿科技 (300017) and pure thematic chasing of IDC/compute hosting. The strongest name in the market last Friday, but it disclosed an interim report that same evening with revenue −1.38% and attributable net profit −6.61%, with no prior earnings pre-announcement.
2. A third category of samples such as 中利集团 (Zhongli Group, 002309), with "attributable profit soaring while ex-non-recurring is still a loss." Attributable +261.37% but ex-non-recurring still a loss of about −RMB 85 million. Together with 聚辰股份 and 芯朋微, there are at least three such names in this window.
3. 盛科通信 (688702)'s "major contract." The information is not new — the RMB 2.2 billion quota was announced 7/16 and approved by the shareholders' meeting 8/6, and the market has already reacted (7/17 −16.5%); Jan–May actual related-party volume annualizes below full-year 2025; the company itself says the "impact is relatively limited"; the Big Fund is still in its disposal window; PS 142x.
4. Chasing already-pre-announced companies by treating "surging earnings" as new news. 生益科技, 生益电子, 锐捷网络, 星网锐捷 and 中微半导 were all pre-announced in early July. If anyone pitches these names today on the grounds of "surging interim results," that is old news.
5. Fake-earnings stocks where "attributable soars but ex-non-recurring goes backwards." 聚辰股份 (688123, attributable +156% / ex-non-recurring −47.30%), 芯朋微 (688508, attributable +74% / ex-non-recurring −53.30%). By the same token, 星网锐捷's +75.9% attributable corresponds to ex-non-recurring of only +45.91%.
6. 生益电子 (688183). The prettiest headline ex-non-recurring figure at +110.29%, but the smallest expectation gap (landing at the pre-announced midpoint) with the most negative new information (Q2 gross margin −1.54pct QoQ, cash-collection ratio 100%→66%), and consensus still implies H2 must accelerate a further 33%.
7. High-level consecutive-limit-up stocks as a group. 蓝盾光电 (300862, 5 boards, seal/turnover 0.01, seal broken 5 times), 华西股份 (000936, 3 boards, seal broken 7 times), 共进股份 (603118, 2 boards + the company has clarified there is no undisclosed material information + selling assets + self-disclosed PE of 157).
8. 太辰光 (300570)'s "expansion narrative." On the very same day as the RMB 720 million expansion, the interim report showed ex-non-recurring −2.55% and operating cash flow −156.88%, and the land has not yet been won at auction.
9. Money-less catch-up names within optical communications. 永鼎股份 (600105), 通鼎互联 (002491).
10. The compute-server end. 紫光股份 (Unisplendour, 000938) and 浪潮信息 (Inspur Electronic Information, 000977) have seen net main-force outflows for 2 consecutive days.

9.4 Final One-Sentence Judgment

Trace every one of the weekend's "positives" back to its first disclosure date and there is almost nothing new inside the AI narrative: the profits of the 生益 companies, 锐捷, 星网 and 中微 were pre-announced in early July, and 盛科's RMB 2.2 billion related-party quota was announced in mid-July, on which the market promptly fell 16.5%; what genuinely saw daylight for the first time was 网宿科技's decline in both revenue and profit, Moutai's negative growth, 太辰光's backward-going ex-non-recurring profit, and operating cash flows of −RMB 3.018 billion at 锐捷 and −RMB 3.566 billion at 星网. The one report that is truly large and clean, and completely un-pre-announced and un-front-run, sits outside the narrative instead — 卫星化学, ex-non-recurring RMB 6.053 billion, +109% YoY, ex-non-recurring/attributable 97.2%, turnover of just 1.05% on 8/14 — and it is a cyclical on a PE of 9.68, not AI. So today's correct posture is not "chase the earnings" but first sort out which numbers are new; and once you do, you find that the best answer is not on the line everyone is looking at.

---

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

Data-Sourcing and Run Log (not sent to clients)

1. This brief's core thesis was overturned three times

First draft: "the weekend earnings chain saw zero front-running and will be priced all at once on Monday," on which 生益科技 was given 9 pts for expectation gap and 生益电子 80 pts.

First overturn (fundamentals-analyst): 生益科技/生益电子 had already issued earnings pre-announcements on 7/14. I then used stock_yjyg_em(date='20260630') to systematically check every candidate and found that 锐捷 (7/02), 星网 (7/03) and 中微半导 (7/23) had all pre-announced too.

Second overturn (a second fundamentals-analyst): after rewriting, I elevated 盛科通信 to "the only unpriced positive catalyst in the whole window," gave it a full 10 pts for expectation gap and put it in the Top 5. But its RMB 2.2 billion related-party transaction quota had already been announced on 7/16 (filing no. 2026-021) and approved by the shareholders' meeting on 8/6, and the market first rose 5.96% and then crashed 16.5% the next day.

Third overturn (the risk-auditor's "survivorship bias" item): 卫星化学 (002648) — ex-non-recurring RMB 6.053 billion, +109.01%, ex-non-recurring/attributable 97.2%, no earnings pre-announcement, turnover of just 1.05% on 8/14 — is the largest, cleanest and genuinely least front-run report in the whole window, and it appeared exactly zero times in the first draft. Worse, its code was already in the candidate pool of work/f4.py / work/f6.py and its quotes had already been pulled, and it was silently discarded solely because it did not belong to the "AI network side" narrative.

The first two are the same error: mistaking "the share price didn't move" for "the information wasn't disclosed." The third is a different error: cutting into a narrative branch before running the full ranking. Stacked together, the two errors amount to "I decided the story first, then went looking for data to support it." Both have been written up as meta-rules in Section 4 of the main text, with the screening methodology published in Section 6.1.

2. Errors the risk-auditor caught that I had not found myself (all corrected)

One item the auditor raised that I did not adopt after checking: it suggested switching PE to TTM across the board. What I changed was labeling the basis + giving both figures at key points, without a wholesale replacement — because Eastmoney's "PE (dynamic)" is the basis on which all stocks in this brief are cross-comparable, and switching everything to TTM would break within-group comparability. The auditor itself also noted that this basis has a historical error of −46%~+30% on 锐捷, so PE-TTM 171x is labeled separately for 锐捷 in the main text.

3. 网宿科技 was found by self-inspection (before the auditor came back)

While self-checking against the "survivorship bias" item in the auditor's prompt, I filtered the 136 weekend interim reports by industry down to 52 technology names and listed them all, which is how I saw its revenue −1.38% and attributable net profit −6.61%. It was last Friday's most money-concentrated stock, and its interim report was logged at 18:19 on 8/14 — 3 hours after the limit-up.

The root cause is the same as with 卫星化学: looking only at the TOP 40 by descending "attributable YoY growth" means negative-growth companies are simply not in view. The next interim-report sweep must look at three dimensions at once: ① a descending ranking by ex-non-recurring growth; ② whether the prior session's strong performers disclosed interim reports; ③ the whole market rather than a single industry.

4. Gaps in the data audit trail (raised by the auditor, and accurate)

Line 19 of work/f4.py has a syntax error (SyntaxError: unmatched '}', a dangling literal left behind when sed edited the CAND dict), so it never executed successfully, and the intended output work/snap0814b.json does not exist. The work/snap0814.json that was actually written contains only 34 names and does not include 生益电子, 锐捷网络, 中微半导, 贵州茅台 and other core names — those figures were only printed to stdout by f6.py and never written to disk. The main text's claim of "verified as not stale" has no auditable trail for that subset. The auditor independently re-verified that the values themselves are correct (生益电子 +5.59%/main-force +RMB 173 million/PE 43.73/PB 14.92; 锐捷 +0.42%/−RMB 43 million/PE 115.42/PB 30.03). But procedurally this is a genuine gap: after sed edited the script, I moved on without verifying the exit code.

5. API status

• push2.eastmoney.com intermittently returned 502 (same cause as 8/14), push2delay was normal, and f124 timestamp checks confirm a genuine 15:34–16:14 closing snapshot.
• The limit-up pool uses the official push2ex pool with validation of tc == len(pool) (63 == 63).
• RPT_DMSK_FN_MAINFINADATA returned empty for batch ex-non-recurring pulls → switched to stock_financial_abstract_ths name by name, so a full ex-non-recurring ranking of the 136 companies could not be produced (this is one of the technical triggers for missing 卫星化学, though the root cause is still process ordering).
• All 74 Dragon-Tiger List entries for 8/14 were retrieved. 亨通光电 is not among them — it did not meet the "top 5 by daily price-move deviation" threshold, so the exchange never disclosed it. Last week's recap listed "check 亨通's Dragon-Tiger List seats" as a Monday must-do; the conclusion is that the data does not exist, not that it was not found.
• Conflicting timestamp for the Centec filing: the local API returns 8/16 18:28 (the upload moment), while CNINFO labels it 8/17 and the filing is dated "August 17, 2026." The main text explicitly states that no precise timestamp is used.

6. Incomplete / to be verified

• July–August 2026 ethane-ethylene spread data was not obtained. All conclusions on 卫星化学 (currently Top 1) rest on the H1 spread widening, and the Q3 direction is unknown. This is the biggest unknown for this brief's No. 1 name and is flagged in the main text.
• CCL spot prices and July–August quotes for upstream copper foil/fiberglass cloth/resin were not obtained. The 生益科技 conclusion rests on "+30% price increases," while mature sites' volumes have already stalled. Both companies hold results briefings on 8/17, where this can be confirmed.
• 锐捷's order backlog and customer concentration are undisclosed, so it is impossible to judge whether the +RMB 5.86 billion of working-capital absorption is "order-driven" or a bet.
• The full ex-non-recurring ranking of the 136 interim reports was not completed, so omissions may remain.
• Consensus estimates were formed before the interim reports and need to be re-pulled after 8/17.

7. Methodology: the three actions that actually worked this issue

1. Trace back to the first disclosure date — for any "earnings/contract/quota" news, do not accept the surface timestamp; trace the prior document it cites. This issue that intercepted 5 pre-announced names plus 盛科通信.
2. Compute the ex-non-recurring/attributable ratio first — this intercepted 聚辰, 芯朋微 and 中利集团, and corrected 星网锐捷 from +75.9% to +45.91%.
3. The full ranking comes before narrative branches — this issue it was precisely by not doing this step that the No. 1 name was missed. Next issue this must come first in the process.

An honest self-assessment: this brief's thesis was overturned three times, and the final conclusion (the best name is outside the AI narrative) is almost the exact opposite of the first draft (earnings delivery on the AI network side). Of the three overturns, two came from sub-agents and one from a prompt in the auditor — not one of them was found by me independently. That shows this slot is still very weak at "finding counterexamples to itself," and it was precisely this weakness that nearly pushed a piece of old news to clients as the headline.