A-Share Market Recap | 2026-08-04 Tuesday

⚠️ Risk disclaimer (up front): This recap is post-close information review and observation only. It does not constitute investment advice, and contains no ratings, target prices, position sizing or buy/sell recommendations. The terms "delivered / did not deliver", "hit rate" and "stage" used below are research-oriented classifications of the day's public market action, and do not represent any judgement about future returns.

Data-basis disclosure first (please read this section before anything else)

1. Market-data basis: all figures are 2026-08-04 close data. Index moves and turnover are taken from Sina Finance index quotes; individual-stock price change, turnover rate, traded value and net main-force inflow are taken from Tencent Finance stock quotes; limit-up / failed-limit-up (broken board) / previous-day limit-up pools are taken from East Money; industry moves and industry net inflow are taken from the THS (Tonghuashun) industry-sector basis. Industry net inflow and single-stock net main-force inflow belong to two different bases and cannot be added together directly.
2. Limit-up thresholds by board: main board ±10%, ChiNext / STAR Market ±20%, Beijing Stock Exchange ±30%. Every stock in this report is tagged with its board.
3. Reconciliation baseline (the most important methodological statement in this report): today was a broad-based rally day — of 5204 A-shares, 3406 rose and 1664 fell, with a median gain of +1.00%. Therefore "the stock went up" does not equal "the pre-market call was right". This report uses three tiers throughout: delivered = limit-up or gain ≥5%; flat = +1.00% to +5%; did not deliver = <+1.00% (failed to beat the broad-rally baseline). Only with this yardstick can "leading the main theme" be separated from "riding the tide".
4. Items not obtained, marked "no reliable data":
   • Northbound single-day net flow: the exchanges stopped disclosing single-day net buy figures from August 2024; the interface returns net buy = 0, so this report makes no statement about northbound direction.
   • Today's Dragon-Tiger List (LHB): cut-off time was 15:42 and the Dragon-Tiger List had not yet been published for the day, so the Dragon-Tiger List part of Section 5 is left blank — yesterday's data is not substituted in.
   • Two-market turnover sequential change: combined turnover of the two markets today was RMB 2.2136 trillion (Shanghai RMB 1.0084 trillion + Shenzhen RMB 1.2052 trillion), measured directly; but the turnover-value data on the previous trading day's basis was not obtained this time, so only the share-volume sequential change is given (Shanghai +3.0%, Shenzhen +7.4%), and no turnover-value sequential change is given.
5. One data-source defect that must be flagged: 高争民爆 (Gaozheng Explosive, 002827, main board) closed today at RMB 41.87, up +10.01%, with the close equal to the limit-up price — in fact its 5th consecutive limit-up; but East Money's limit-up pool does not include this stock (it was classified into the broken-board pool, having broken the board 26 times intraday before sealing again into the close). The limit-up total and the highest consecutive-limit-up height in this report have both been manually corrected to include this stock, so results will differ by 1 stock from a direct API call.
6. One naming-basis note: 601567 is disclosed by the exchange as "三星医疗" (Sanxing Medical), while the quote source shows "三星电气" (Sanxing Electric); this report follows the exchange's naming.

⚠️ Data-fetch failures and implementation details for this run (internal only):

• East Money's full-market quote endpoint stock_zh_a_spot_em returned RemoteDisconnected for the third consecutive trading day (already logged on the 8/3 pre-market and 8/3 recap runs); all 6 retries failed, so we switched to Tencent's stock_zh_a_spot_tx (5536 records, includes the zljlr net-main-force-inflow field, verified usable). Recommend promoting the Tencent source to the default primary and demoting East Money to backup — three straight days of breakage is not an isolated incident.
• East Money's sector endpoints all failed: stock_board_industry_name_em = RemoteDisconnected, stock_board_concept_name_em = HTTP 502. Switched to THS stock_board_industry_summary_ths (90 industries, includes net inflow, usable).
• stock_board_concept_summary_ths returns a historical "concept-driving events" table (dates going back to 2024), not the day's concept gain/loss ranking; there is no price-change column in its fields at all, and the first call raised IndexError directly. This endpoint name is misleading — do not use it as a concept ranking again.
• stock_zh_index_spot_em / index_zh_a_hist also failed on the East Money source; indices were switched to Sina's stock_zh_index_spot_sina (measured total matches the aggregated Tencent single-stock turnover of RMB 2.2136 trillion exactly — cross-validation passed).
• Dragon-Tiger List: stock_lhb_detail_em(20260804) returned None (4 retries); this is in fact not yet disclosed by the exchange (cut-off 15:42, the Dragon-Tiger List is usually published after 18:00), not an endpoint failure. The backup channel stock_lhb_detail_daily_sina raised KeyError('股票代码') — its field structure has changed, so this backup path is dead and needs rewriting. This report truthfully leaves it blank.
• East Money's limit-up pool missed 高争民爆 (Gaozheng Explosive, 002827): the stock closed at 41.87 = limit-up price 41.87, up 10.01%, yet appeared only in the broken-board pool (26 board breaks). Reporting the limit-up count straight from the row count of stock_zt_pool_em would undercount by 1 stock and would drop the highest consecutive-limit-up height from 5 boards to 4. Manually corrected. This is a textbook case of "API basis ≠ fact"; limit-up-count data must in future be cross-checked against the broken-board pool.
• Turnover-value sequential change not obtained: Sina's index daily bars give only volume (shares), not amount, and the amount field in Tencent's daily bars is actually volume (lots) rather than value — after two cross-validations the unit was confirmed not to be value, so the turnover-value sequential change was abandoned and only the volume sequential change is given. We nearly wrote it up as a value figure based on the literal meaning of "amount" — a classic basis trap.
• Self-record on the main-theme call: this recap's conclusions are heavily negative on yesterday's pre-market report. What needs remembering is this — the pre-market run spent 97% of its analysis budget on word-frequency statistics of one policy PDF, while the two catalysts that actually ignited today (WuXi AppTec's interim report, NVIDIA's Vera Rubin / optical modules) were sitting right there in rows 7 and 2 of the pre-market Section 1 news table, only to be graded A and "multi-branch" and never unpacked further. This is not missing information; it is misallocated analytical resources. Tomorrow morning's pre-market run must first do a "line-by-line breakdown of evening announcements", and only then the policy deep-dive.

---

0. One-Sentence Recap

Today's sentiment temperature: euphoric (139 limit-ups, 0 limit-downs, broken-board rate of only 9.2%, 848 stocks up more than 5%); the strongest main theme is 【 AI compute hardware 】 — optical modules/CPO + PCB/copper-clad laminate + semiconductors all exploded on three legs at once, compounded by 【 CRO / innovative drugs 】 ignited by the better-than-expected interim report from 药明康德 (WuXi AppTec, 603259, main board); meanwhile 【 distribution grid / pumped-storage hydro / UHV 】, the bets on yesterday's pre-market list, went collectively silent on the first trading day after the policy document was made public — the grid equipment sector rose only 2.03% with net capital outflow of RMB 144 million. None of the pre-market Top 5 recommendations delivered (0/5); the core large-caps 国电南瑞 (NARI Technology, 600406, main board) closed −0.40% and 中国电建 (PowerChina, 601669, main board) closed −1.01%. Tomorrow's tone: the AI hardware main theme is in its start-up acceleration phase but the single-day gain is too large, so watch for a gap-up split; the grid line is downgraded to "policy already priced in, needs a pullback for confirmation".

One-sentence self-critique: the pre-market report's biggest problem was not wrong data but treating "what the document says" as "what the market will trade" — it even listed WuXi AppTec's interim report and the three big cloud vendors all rising 5% in its own news table, yet its "S-grade policy is the only one" framework flattened both into supporting roles.

---

1. Market Overview

1.1 Indices and Turnover

• Combined turnover of the two markets was RMB 2.2136 trillion. On a volume basis it expanded sequentially: Shanghai +3.0%, Shenzhen +7.4% (turnover-value sequential change not obtained, see item 4 of the data-basis note).
• Today's most important structural signal is "SSE 50 −0.29% vs ChiNext Index +5.65%", a gap of 5.94 percentage points. This is a textbook high-to-low rotation: capital rotating wholesale out of large-cap value (banks, insurance, baijiu, utilities) into growth (AI hardware, semiconductors, pharma). The Shanghai Composite rose only 0.33%, which does not mean today was a weak day — it is because the heavyweights were being sold.

1.2 Advance/Decline and Sentiment

Sentiment temperature verdict: euphoric. The basis is four conditions holding at once — 139 limit-ups with 0 limit-downs, a broken-board rate of only 9.2%, 848 stocks up more than 5% versus only 7 down more than 5%, and yesterday's limit-up stocks still averaging +4.77% today. But one hedge must be recorded alongside: the euphoria sits not in yesterday's consecutive-board tier but in the institutional names that started moving today (large-cap names like Innolight and WuXi AppTec hitting limit-up / rising sharply). This is a different kind of euphoria from a "hot-money consecutive-board sentiment climax", and its ebb rhythm is different too.

---

2. Reconciliation of the Pre-Market List

Reconciliation baseline: median gain of all A-shares +1.00%. Delivered = limit-up or ≥5%; flat = +1.00% to +5%; did not deliver = <+1.00%.

2.1 Pre-market Section 9.1 "The 5 main-board stocks most worth watching today"

Top 5 hit rate: delivered 0/5 = 0%; beat the broad-rally baseline (+1.00%) 2/5 = 40%. This is a failing recommendation list, and the failures are concentrated in the two highest-ranked names (both #1 and #2 lagged the baseline) — the ranking direction is negatively correlated with actual performance, which is worse than merely "weak overall".

2.2 Remaining names on the pre-market Section 3 master list

2.3 Hit-Rate Summary

One-sentence conclusion: the pre-market list was clearly stronger at "avoiding" than at "recommending". An avoidance-correct rate of 43.3% against an avoidance-wrong rate of 23.3% is close to 2:1, while the delivery rate of the positive Top 5 recommendations was 0. The real value of the pre-market report today lay in telling people what not to buy (Tongwei, Sanxing Medical, China National Nuclear Power and China Yangtze Power all played out), not in what it suggested watching.

2.4 Self-Critique: What Was Called Right, What Was Called Wrong, and How to Fix It Tomorrow Morning

① What was called right was "disconfirming judgements"; what was called wrong was "narrative judgements". Everywhere the pre-market report used first-hand financial evidence to negate a popular narrative, it was right today: 通威股份 (Tongwei, 600438, main board) interim pre-announcement rejecting the solar anti-involution narrative (−0.08% today), 中国核电 (China National Nuclear Power, 601985, main board) "110 million kW is not a new number" (−1.11%), 涪陵电力 (Fuling Electric Power, 600452, main board) "restatement of an old target" (−0.50%), 三星医疗 (Sanxing Medical, 601567, main board) "the first move is already out" (−3.00%). Everywhere it derived "where capital should go" from a policy text, it was wrong today: the core large-caps of both the S-grade distribution-grid branch and the A+-grade pumped-storage branch lagged the baseline. Lesson: text analysis is good at excluding, not at selecting. Word-frequency statistics can tell you "UHV is not the focus of this document" (and that was not disconfirmed today either — the UHV large-caps Pinggao, XD and TBEA all failed to hit limit-up), but it cannot derive "therefore capital will go to the distribution grid".

② The biggest methodological error: using "main board first" as an exclusion filter rather than a screening filter. Today's strongest line had its leading core almost entirely on ChiNext — 中际旭创 (Innolight, 300308, ChiNext) +13.24%, 新易盛 (Eoptolink, 300502, ChiNext) +13.68%, 天孚通信 (T&S Communications, 300394, ChiNext) +17.45%. The pre-market report listed AI compute as branch 5 and marked it "not a primary pick" because it was "not main board". The point of "main board first" is to control trading-attribute risk, not to black out the pricing signal of an entire main theme. The fix: branch-strength ranking must be assessed market-wide (including ChiNext / STAR), with main-board-first applied only at the final stock-list level; and at the same time every non-main-board main theme should get a main-board mapped name added — today's main-board mappings for this line (亨通光电 Hengtong Optic-Electric 600487 +6.12%, 中天科技 Zhongtian Technology 600522 +5.46%, 生益科技 Shengyi Technology 600183 limit-up, 东山精密 Dongshan Precision 002384 limit-up) were all tradeable and all delivered, and the pre-market report did write all of them into the branch representatives, yet not one made the Top 5.

③ The most glaring point: the catalyst was in our own news table and never unpacked. Row 7 of the pre-market Section 1 news table read "Summary of the evening announcements of August 3: AMEC pre-announces growth, WuXi AppTec interim report, Gaole large compute order…", graded A, with the branch field filled in as "multi-branch". Today 药明康德 (WuXi AppTec, 603259, main board) sealed the limit-up, with traded value of RMB 10.502 billion and net main-force inflow of RMB 1.833 billion, igniting the whole CRO / innovative-drug line single-handedly. And its interim figures (H1 revenue RMB 28.897 billion +38.93%, net profit attributable to parent RMB 11.080 billion +29.43%, full-year revenue guidance raised from RMB 51.3-53.0 billion to RMB 58.5-60.5 billion) were already public at the pre-market cut-off. The pre-market report compressed a disclosed, quantifiable, guidance-raising company-level catalyst into four characters in a summary table. The fix: tomorrow morning's pre-market run must first do a "line-by-line breakdown of evening announcements" (especially interim reports / results pre-announcements / guidance raises), grading each one separately, and only then do the policy deep-dive. With the order reversed, the budget gets misallocated.

④ 生益科技 (Shengyi Technology, 600183, main board) deserves its own entry: the classification judgement was right, the trading conclusion was wrong. The pre-market report put it on the Pass list, on the grounds that "the company's main business is copper-clad laminate / electronic materials, not memory chips, with no direct transmission from memory price increases — a textbook case of concept ≠ beneficiary". That classification judgement was entirely correct: it hit limit-up today, and indeed not as a memory stock, but as a PCB / copper-clad laminate name within the AI server hardware chain (the THS "components" sector rose +7.07%, with 63 gainers and 0 decliners). But the conclusion was wrong: the pre-market report only proved "it does not belong to narrative A" and then casually Passed on it, without asking one more question: "then which narrative does it belong to, and how strong is that narrative?" The fix: when Passing on a stock, "it does not belong to X" can only imply "do not buy it for X's reasons", not "do not watch it". After rejecting one attribution, a positive attribution must be done as a follow-up. This is the most expensive and the most reusable lesson in this report.

⑤ One mechanism that was validated and should be kept. The pre-market Section 8.3 wrote one signal: "blocked by a stronger stock: if the memory oversold bounce (GigaDevice) runs stronger than power equipment, capital is rotating the main theme." Today 兆易创新 (GigaDevice, 603986, main board) +4.48% versus 平高电气 (Pinggao Electric, 600312, main board) +1.59% — this signal triggered accurately. The problem is not that the signal wasn't designed, but that nothing happened after it triggered. Tomorrow morning's pre-market run should upgrade this class of "main-theme disconfirmation signal" from descriptive prose into entries with prescribed actions (trigger → downgrade that branch's strength that day).

---

3. Verification of Today's Main Theme

3.1 Line-by-Line Reconciliation of the Pre-Market Branches

3.2 Did the Pre-Market Report Call the Strongest Main Theme Correctly?

No. The strongest main theme given pre-market was 【 distribution grid / transmission-distribution-microgrid coordination 】 (grade S, the only one); today's actual strongest main theme was 【 AI compute hardware 】.

Quantifying the gap on the THS industry basis:

Grid equipment ranks mid-pack among the 90 industries, and is the only watch-closely branch that "rose yet saw net outflow".

3.3 Unexpected Main Themes the Pre-Market Report Missed Entirely

① Optical modules / CPO / optical communications — today's absolute core

• 中际旭创 (Innolight, 300308, ChiNext) +13.24%, traded value RMB 49.790 billion (#1 market-wide), net main-force inflow RMB 5.068 billion (#1 market-wide)
• 新易盛 (Eoptolink, 300502, ChiNext) +13.68%, traded value RMB 29.914 billion; 天孚通信 (T&S Communications, 300394, ChiNext) +17.45%, traded value RMB 15.873 billion
• Main-board mappings: 光迅科技 (Accelink, 002281) limit-up, 剑桥科技 (CIG, 603083) limit-up, 永鼎股份 (Yongding, 600105) limit-up (net main-force inflow RMB 1.669 billion), 亨通光电 (Hengtong Optic-Electric, 600487) +6.12%, 中天科技 (Zhongtian Technology, 600522) +5.46%
• Catalyst: NVIDIA released the Vera Rubin platform and the Spectrum-6 switching system, driving silicon-photonics penetration, with 800G/1.6T optical module demand continuing; Nomura raised its 2026-2028 revenue and profit forecasts for Innolight. (Securities Times¹, East Money²)
• The pre-market report did in fact catch the overseas signal (news-table item 2: Microsoft/Amazon/Google all up 5%; item 6: Morgan Stanley raised cloud capex to USD 1.2 trillion), but graded it A, placed it in branch 5, and did not make it a primary pick because it was not main board.

② PCB / copper-clad laminate / components — the sub-sector with the most limit-ups

• Components sector +7.07%, 63 gainers and 0 decliners (the only industry with zero decliners)
• Limit-ups: 东山精密 (Dongshan Precision, 002384, main board) traded value RMB 15.329 billion, 沪电股份 (WUS Printed Circuit, 002463, main board), 深南电路 (Shennan Circuits, 002916, main board), 生益科技 (Shengyi Technology, 600183, main board), 鹏鼎控股 (Avary Holding, 002938, main board), 红板科技 (Redboard Technology, 603459, main board), 华正新材 (Huazheng New Material, 603186, main board), 江海股份 (Jianghai Capacitor, 002484, main board), 中京电子 (Zhongjing Electron, 002579, main board)
• ⚠️ Basis reminder: THS shows the components sector's leading gainer as "嘉立创 (JLC) +146.28%", but N 嘉立创 (JLC, 001232, main board) is a new listing whose first trading day is today; its gain does not reflect pricing by existing capital and should not be used as evidence of sector strength.

③ CRO / innovative drugs — ignited by a single interim report

• 药明康德 (WuXi AppTec, 603259, main board) limit-up +10.00%, traded value RMB 10.502 billion, net main-force inflow RMB 1.833 billion, sealed again after breaking the board 6 times intraday
• Limit-up alongside it: 凯莱英 (Asymchem, 002821, main board), 普洛药业 (Apeloa Pharmaceutical, 000739, main board), 创新医疗 (Innovative Medical, 002173, main board), 双成药业 (Shuangcheng Pharmaceutical, 002693, main board), 罗欣药业 (Luoxin Pharmaceutical, 002793, main board), 哈三联 (Harbin Sanlian, 002900, main board), 百花医药 (Baihua Medical, 600721, main board), 珍宝岛 (Zhenbaodao Pharmaceutical, 603567, main board), 贝瑞基因 (Berry Genomics, 000710, main board), 济民健康 (Jimin Health, 603222, main board), 义翘神州 (Sino Biological, 301047, ChiNext) +20%
• Catalyst: WuXi AppTec disclosed its 2026 interim report on the evening of August 3 — H1 revenue RMB 28.897 billion (+38.93%), net profit attributable to parent RMB 11.080 billion (+29.43%, first time above RMB 10 billion in a first half), Q2 single-quarter revenue RMB 16.462 billion (+47.71%); full-year revenue guidance raised from RMB 51.3-53.0 billion to RMB 58.5-60.5 billion. (Sina Finance³, East Money⁴)
• ⚠️ The pre-market directional call on pharma was backwards: news-table item 19 labelled pharma "bearish-leaning" (on the basis of the centralized-procurement selection announcement + AstraZeneca down 9% in London), grade C. Today the medical services sector rose +5.54%, the 4th strongest industry market-wide.

④ Semiconductors

• Sector +6.01%, net inflow RMB 23.368 billion (#1 across all industries); 和林微纳 (Helin Micro-Nano, 688661, STAR Market) +20%, 金海通 (Jinhaitong, 603061, main board) limit-up, 长电科技 (JCET, 600584, main board) +7.23%, 北方华创 (Naura, 002371, main board) +7.13%, 澜起科技 (Montage Technology, 688008, STAR Market) +7.01%

---

4. Limit-Up Tiers / Board Heights

4.1 Consecutive-Limit-Up Tiers

4.2 Promotion Counts and Quality

• 75 limit-ups yesterday → 20 limit-ups today, promotion rate 26.7% (first-board promotion is on the low side)
• 17 names with ≥2 consecutive boards yesterday → 9 limit-ups today, consecutive-board promotion rate 52.9% (healthy inside the tier)
• Yesterday's limit-up stocks today averaged +4.77%, median +4.39%, 62 up 12 down — no broad collective board-breaking occurred, and the "yesterday's strong themes break their boards collectively" risk signal listed in pre-market Section 8.4 did not trigger

4.3 Board Breaks and the Ebbing Tide

4.4 Core Large-Caps (High Traded Value + High Recognizability)

4.5 Broken-Board Structure (Important)

Of today's 14 broken-board stocks, most still closed up sharply: 大族激光 (Han's Laser, 002008, main board) +8.67%, 兴森科技 (Fastprint Circuit, 002436, main board) +9.24%, 宏和科技 (Honghe Technology, 603256, main board) +8.87%, 中钨高新 (China Tungsten & Hightech, 000657, main board) +8.79%, 胜蓝股份 (Shenglan Technology, 300843, ChiNext) +17.88%, 欧莱新材 (Oulai New Material, 688530, STAR Market) +17.04%, 博迁新材 (Boqian New Materials, 605376, main board) +9.32%. This is "high-level turnover" rather than "capital exhaustion" — only a minority dived into the red after breaking the board. A broken-board rate of 9.2% with broken-board stocks not falling is a sign of healthy sentiment, not a topping signal.

---

5. Capital Flows

5.1 Northbound Capital

No reliable data. The exchanges stopped disclosing single-day net buy figures for the Shanghai/Shenzhen Stock Connect from August 2024; the interface returns net buy = 0, so this report makes no judgement whatsoever about northbound direction. The only confirmable figures are southbound: Stock Connect (Shenzhen) net buy of RMB 2.863 billion, Stock Connect (Shanghai) −RMB 0.439 billion, Hang Seng Index −0.66%.

5.2 Dragon-Tiger List

As of the cut-off (15:42) the exchanges had not yet published today's Dragon-Tiger List data, so this section is left blank — yesterday's data is not substituted in. The seat characteristics recorded pre-market yesterday (8/3) were "mostly hot money, low institutional participation". Whether large net buying from institution-dedicated seats appears today on the back of gains in large-cap names like Innolight and WuXi AppTec needs to be confirmed once the Dragon-Tiger List is published — and that is precisely the first thing tomorrow's pre-market run should go back and do.

5.3 Industry Net Main-Force Inflow (THS basis, RMB bn)

Top net inflows

Top net outflows

One-sentence portrait of today's capital flows: sell everything high-dividend / large-cap value (banks, insurance, baijiu, power, utilities), buy everything AI hardware (semiconductors, communication equipment, components, optics & optoelectronics). Grid equipment sits in between — up 2.03%, yet with net capital outflow, a catch-up state of "nobody is selling but nobody is really buying either".

5.4 Single-Stock Net Main-Force Inflow/Outflow (Tencent basis)

Top 10 net inflows: 中际旭创 (Innolight, 300308, ChiNext) +RMB 5.068 bn, 新易盛 (Eoptolink, 300502, ChiNext) +RMB 4.549 bn, 东山精密 (Dongshan Precision, 002384, main board) +RMB 4.364 bn, N 嘉立创 (JLC, 001232, main board, first trading day) +RMB 3.069 bn, 天孚通信 (T&S Communications, 300394, ChiNext) +RMB 2.986 bn, 药明康德 (WuXi AppTec, 603259, main board) +RMB 1.833 bn, 永鼎股份 (Yongding, 600105, main board) +RMB 1.669 bn, 兆易创新 (GigaDevice, 603986, main board) +RMB 1.595 bn, 工业富联 (Foxconn Industrial Internet, 601138, main board) +RMB 1.411 bn, 光迅科技 (Accelink, 002281, main board) +RMB 1.394 bn.

Top 10 net outflows: 比亚迪 (BYD, 002594, main board) −RMB 1.486 bn, 长鑫科技 (ChangXin Memory, 688825, STAR Market) −RMB 1.233 bn, 紫金矿业 (Zijin Mining, 601899, main board) −RMB 0.811 bn, 农业银行 (Agricultural Bank of China, 601288, main board) −RMB 0.724 bn, 贵州茅台 (Kweichow Moutai, 600519, main board) −RMB 0.664 bn, 中国平安 (Ping An, 601318, main board) −RMB 0.552 bn, 华电辽能 (Huadian Liaoning Energy, 600396, main board) −RMB 0.507 bn, 中芯国际 (SMIC, 688981, STAR Market) −RMB 0.487 bn, 风华高科 (Fenghua Advanced, 000636, main board) −RMB 0.458 bn, 长江电力 (China Yangtze Power, 600900, main board) −RMB 0.450 bn.

⚠️ Two divergences that need to be flagged separately:

• 华电辽能 (Huadian Liaoning Energy, 600396, main board) rose 5.54% yet had net main-force outflow of RMB 507 million, turnover rate 20.73% — textbook high-level distribution; the pre-market qualitative call to Pass on it (PE 563.29) was confirmed on the capital-flow level.
• 长鑫科技 (ChangXin Memory, 688825, STAR Market) closed +0.02% with net main-force outflow of RMB 1.233 billion against a backdrop of RMB 23.368 billion of net inflow into the semiconductor sector, on traded value of RMB 27.763 billion. The memory leader failing to follow the rally is a signal that this line is not homogeneous internally.

---

6. Tomorrow's Outlook

6.1 Judgement on the Continuity of Tomorrow's Main Themes

6.2 Key Watch Items for Tomorrow

The following is a watch-priority ranking and confirmation-point list, not a rating and not buy/sell advice.

6.3 What to Avoid Tomorrow

1. The two names from yesterday's pre-market Top 5 that have already been disconfirmed: 国电南瑞 (NARI Technology, 600406, main board), 中国电建 (PowerChina, 601669, main board) — failing to rise on the first day of policy implementation makes rising on day two harder still; their turnover rates today were 0.93% and 0.98% respectively, capital attention is not low, it is absent.
2. High-level second- and third-tier grid equipment names: 顺钠股份 (Shunna Electric, 000533, main board, 6 boards in 9 days, turnover rate 20.00%), 长缆科技 (Changlan Technology, 002879, main board, turnover rate 31.90% with net main-force outflow of RMB 55.84 million), 风范股份 (Fengfan Power Equipment, 601700, main board, 2 boards, negative PE) — they have now hit limit-up for two straight days against a backdrop of sector net outflow, textbook end-stage behaviour of a branch in its ebbing phase.
3. 华电辽能 (Huadian Liaoning Energy, 600396, main board): up 5.54% but with net main-force outflow of RMB 507 million, turnover rate 20.73%, PE 563.29 — the most extreme sample of price up, money out.
4. Isolated high-level consecutive-board names: 传智教育 (Chuanzhi Education, 003032, main board, 7 boards), 高争民爆 (Gaozheng Explosive, 002827, main board, 5 boards, 26 intraday board breaks, turnover rate 21.20%), 神雾节能 (Shenwu Energy Saving, 000820, main board, 4 boards, turnover rate 32.30%) — none of them is connected to either of today's main themes, and there will be no bid to absorb them once sentiment cools.
5. Chasing today's 20cm limit-up names on ChiNext / the STAR Market: 光库科技 (Advanced Fiber Resources, 300620, ChiNext) +20.00%, 汉仪股份 (Hanyi Fonts, 301270, ChiNext) +20.01%, 宏景科技 (Hongjing Technology, 301396, ChiNext) +20.00%, 中青宝 (Zhongqingbao, 300052, ChiNext) +20.02%, 和林微纳 (Helin Micro-Nano, 688661, STAR Market) +20.00% — the volatility-amplification effect of 20cm names on day two of a main theme is pronounced.
6. N 嘉立创 (JLC, 001232, main board): +146.28% on its first trading day with a turnover rate of 75.23%; first-day pricing of a new listing has no reference value, it should not be used as evidence of components-sector strength, and it is not suitable for a sector-logic relay trade.

6.4 Input Notes for Tomorrow Morning's Pre-Market List (the most important section)

1. ⚠️ Change the workflow order: unpack the evening announcements first, then do the policy deep-dive. The catalysts of today's two strongest main themes (WuXi AppTec's interim report, the overseas optical-module industry news) were both already public at yesterday's pre-market cut-off and were both written into the news table, yet were never unpacked because the entire analysis budget went to word-frequency statistics of one policy PDF. Tomorrow morning we must first go line by line through the interim reports / results pre-announcements / guidance raises from the evening of 8/4, grading each one separately, before doing anything else.
2. ⚠️ Change the branch-ranking basis: strength ranking must be assessed market-wide, and "main board first" applies only at the final stock-list level. Yesterday, "not main board" pushed AI compute down to branch 5, which directly missed the name with the highest traded value and the highest net main-force inflow in the entire market. Also required: every non-main-board main theme must have a main-board mapped name placed on the formal list (today's 亨通光电 600487, 中天科技 600522, 东山精密 002384 and 生益科技 600183 all validate that this route works).
3. ⚠️ Change the Pass logic: after rejecting one attribution, a positive attribution must be done as a follow-up. The lesson from 生益科技 (Shengyi Technology, 600183, main board) — the judgement "it is not a memory stock" was right, but the very next question should have been "then what kind of stock is it, and how strong is that line?" The answer was PCB / copper-clad laminate, which was precisely the sub-sector with the most limit-ups today.
4. ✅ Keep and reinforce "disconfirmation-style analysis". Today's avoidance-correct rate of 43.3% against an avoidance-wrong rate of 23.3%, with all four disconfirmations — 通威股份 (Tongwei, 600438), 中国核电 (China National Nuclear Power, 601985), 涪陵电力 (Fuling Electric Power, 600452), 三星医疗 (Sanxing Medical, 601567) — playing out. Using first-hand financial evidence to negate a popular narrative is currently the most reliable capability of this process and deserves further investment.
5. ⚠️ Upgrade "main-theme disconfirmation signals" into entries with prescribed actions. Yesterday's Section 8.3 "blocked by a stronger stock" signal triggered accurately (GigaDevice +4.48% stronger than Pinggao Electric +1.59%), but nothing followed the trigger. Every entry on tomorrow morning's signal list must spell out "how to adjust branch strength once triggered".
6. ⚠️ The pharma direction needs re-assessment. Yesterday pharma was labelled "bearish-leaning" (on the basis of centralized procurement + AstraZeneca down 9%); today medical services rose +5.54%, ranking 4th across all industries. Tomorrow morning we need to re-examine the relative weights of these two opposing forces — the centralized-procurement negative versus the earnings delivery of innovative drugs/CRO — and must not carry yesterday's conclusion forward.
7. Go back and review the Dragon-Tiger List. Today's Dragon-Tiger List was not disclosed by the cut-off. The first thing tomorrow's pre-market run should do is look back at the 8/4 Dragon-Tiger List, focusing on confirming whether Innolight, WuXi AppTec and Dongshan Precision saw large net buying from institution-dedicated seats — this determines whether today's main theme is "hot-money sentiment" or "institutional accumulation", and the persistence of the two is completely different.
8. Technical source-switching recommendation. The primary market-data source should be switched from East Money to Tencent (East Money's full-market quote endpoint has failed for three consecutive trading days), with sector data taken from THS.

---

⚠️ Risk disclaimer: this recap is post-close information review and observation only and does not constitute investment advice. Data may differ in timeliness or basis; please refer to exchange/company disclosures, and do not use this directly as a basis for trading.

Sources (external information):

• ETF close commentary | ChiNext Index up 5.64%, compute hardware supply chain rebounds sharply — Sina Finance⁵
• Optical chips surge across the board! ChiNext Index up nearly 5% — Securities Times¹
• "Optics" is back? The "Yi-Zhong-Tian" trio soars together, optical module giant hits limit-up — East Money²
• WuXi AppTec's better-than-expected interim report ignites the market, A and H shares both surge — Sina Finance³
• RMB 400 billion giant seals limit-up! CRO concept explodes — East Money⁴
• Blowout results from the leader + guidance raise! CRO names surge collectively — East Money⁶
• New-type power system "15th Five-Year Plan" released, building a green low-carbon power landscape — Caixin⁷