A-Share Market Recap | 2026-08-03 Monday

⚠️ Risk disclaimer (up front): This recap is a post-close review of information and observations only. It does not constitute investment advice and contains no ratings, target prices, position sizing, or buy/sell recommendations. Phrases such as "delivered / did not deliver" and "continued / faded" are purely after-the-fact descriptions of price action that has already happened, plus a ranking of research attention. They do not represent any judgment about future returns.

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

1. Time window: 2026-08-03 09:30—15:00 (today's full trading session). The reconciliation baseline is this morning's pre-market list, reports/a-share/2026-08-03.md.
2. ⚠️ The single most important data-scope warning of this issue (it directly affects credibility, so it must be stated first): Sina's full-market quote feed stock_zh_a_spot returned 7/31 closing data today, not today's data. The endpoint's timestamp reads 15:30:02, yet its values are identical line by line to the 7/31 numbers cited in this morning's pre-market report (market-wide 4690 up / 728 down / 879 names up more than 5%, Demingli −1.06% with turnover of RMB 17.355 bn, GigaDevice +2.02% with turnover of RMB 34.043 bn, Kunlun Tech turnover of RMB 6.066 bn, Dongfang Electric +0.76% with turnover of RMB 1.283 bn), and Ailijiaju (603221), which is suspended from trading today, still shows +10.00% — triple cross-validation confirms the source is stale. This report has dropped that source entirely. Also broken: stock_zh_a_hist / stock_zh_a_hist_tx / stock_zh_a_daily / stock_individual_fund_flow / stock_zh_index_daily (daily bars only updated through 7/31), and stock_zh_a_spot_em / stock_board_industry_name_em (RemoteDisconnected / 502).
3. The data sources this report actually uses, each verified as today's (8/3): ① East Money limit-up pool stock_zt_pool_em / broken-limit pool / limit-down pool / previous-day limit-up pool stock_zt_pool_previous_em (cross-validated by the progression of consecutive limit-up counts: Chuanzhi Education 5 → 6 boards, Yiming Foods 4 → 5 boards, Gaozheng Explosive 3 → 4 boards, confirming 8/3); ② Tencent real-time quotes qt.gtimg.cn (timestamp 20260803; every stock-level move and turnover figure in this report comes from this source); ③ Sina index quotes stock_zh_index_spot_sina; ④ THS industry sectors stock_board_industry_summary_ths (semiconductors −3.46%, different from 7/31's +4.23%, confirming today); ⑤ East Money industry/concept fund flows stock_fund_flow_industry / stock_fund_flow_concept.
4. ⚠️ Two items could not be obtained; both are written as "no reliable data" with no speculation:
   • Today's Dragon-Tiger List: the East Money datacenter endpoint returned {"success":false,"message":"返回数据为空"}, and as of writing (after 15:30) the exchange had not yet published the Dragon-Tiger List (usually disclosed 17:00—18:00). This report names no Dragon-Tiger stocks or seats for today. Sina's stock_lhb_ggtj_sina returns a trailing 5-day cumulative scope that cannot be attributed to a single session, so it is likewise not used.
   • Northbound single-day net flow: stock_hsgt_fund_flow_summary_em returned 0.00 with trading status code 3 for both the Shanghai and Shenzhen Connect "net buy turnover" today, i.e. not disclosed (the exchanges stopped disclosing single-day net buy figures from August 2024, consistent with the note in this morning's pre-market report). This report states no direction for northbound flows.
5. Turnover scope: the two exchanges combined traded RMB 1,997.39 bn = Shanghai 952.26 bn + Shenzhen 1,045.13 bn (Sina index scope, excluding the STAR Market and the Beijing Stock Exchange), on the same comparable basis as the 7/31 figures cited this morning (1,187.7 + 1,354.3 = 2,542 bn).
6. No hindsight rewriting in this report: the wrong calls (Leo Group, Tianyu Digital Science, Meili Cloud, Fenghua Advanced Technology) and the miss (the entire grid-equipment chain) are all named in Sections 2 and 3. Today's outcome is not used to retro-fit a "we saw it coming" narrative.

---

0. One-Line Recap

Today's sentiment temperature: a structural warm-up, but the indices are ice cold — this was a split session where "the index fell yet the money-making effect was not bad." The SSE Composite closed −0.59% at 3809.66, the SZSE Component −0.96%, the ChiNext Index −1.24%, while the STAR 50 at −5.08% was the deepest hole on the board. Combined turnover was RMB 1.997 trillion, a 21.4% contraction from 7/31's 2.542 trillion. Yet over the same session there were 75 limit-ups against only 16 broken limit-ups (an 82.4% seal rate; 7/31 was 99 limit-ups / 107 broken), and across the two exchanges 2435 names rose and 1041 fell (Shanghai 1078 up / 543 down, Shenzhen 1357 up / 498 down, Stock Connect scope) — the indices were dragged down by semiconductors and the heavyweights, while small- and mid-cap themes were actually getting stronger.

Today's strongest main line was not either of the top two pre-market branches as originally worded, but their "parent concept": power equipment and electricity. On THS industry data: wind power equipment +4.06%, grid equipment +3.52% (8 limit-ups, the highest count of any industry market-wide; 130 names up vs only 9 down), environmental protection equipment +2.99%, solar equipment +2.91%, electricity +2.28% (net inflow of RMB 3.211 bn, first among all industries). On East Money concept fund-flow data: the nuclear power concept rose +2.68% with a net inflow of RMB 3.343 bn, the largest net inflow of any concept sector market-wide; pumped-storage +1.760 bn, ultra-supercritical generation +1.445 bn, flexible HVDC +1.171 bn.

Pre-market list scorecard: half the branch calls were right, half the within-branch stock structure was backwards, and the avoid list was the most successful part of the whole issue.

• ✅ What we got right: the pre-market report ranked nuclear power as branch No. 1 — today the nuclear power concept's net inflow of RMB 3.343 bn led all concepts market-wide, with Jiangsu Shentong Valve (002438) limit-up +10.00%, Rongfa Nuclear Power (002366) +9.93%, China Nuclear Engineering (601611) limit-up +10.05%, CNNC Sufa (000777) +7.04%, Dongfang Electric (600875) +6.69%. The No. 1 name on the pre-market recommendation list, Dongfang Electric (600875), saw turnover expand from RMB 1.283 bn to 2.548 bn (+98.6%).
• ✅ The best call of all was an avoid: item 1 of pre-market Section 9.3, "resolutely avoid storage / AI hardware" — today GigaDevice (603986) hit a locked limit-down at −10.00% on RMB 23.033 bn of turnover, Demingli (001309) −9.56%, Shengyi Technology (600183) −3.42%, Shennan Circuits (002916) −1.85%, with semiconductors −3.46% and a net outflow of RMB 9.316 bn (the largest outflow market-wide) and the STAR 50 −5.08%.
• ❌ What we got backwards: the pre-market report placed China National Nuclear Power (601985) at No. 2 on the recommendation list as the "named, hardest" name, and today it closed −0.33%, the only one of the five to close in the red; meanwhile Leo Group (002131), judged a Pass (highest governance risk) pre-market, rose +10.09% today for a second consecutive limit-up on RMB 7.001 bn of turnover, the highest of any limit-up stock market-wide.
• ❌ What we missed: grid equipment, the genuinely strongest branch of the day, had no branch of its own pre-market and no representative stock at all.

Tone for tomorrow: the rotation of the main line from "AI applications" to "power equipment / nuclear power" is complete, and tomorrow is day two of this new main line — watch whether the equipment end can put up a second volume expansion, and whether the operator end (China National Nuclear Power / CGN Power) plays catch-up; meanwhile the semiconductor sell-off has shown no sign of a bottom and is not a place to bottom-fish.

---

1. Market Overview

1.1 Indices and Turnover

• Combined turnover across the two exchanges was RMB 1,997.39 bn, a contraction of 544.6 bn (−21.4%) from 7/31's 2,542 bn. The volume behind Friday's repair candle did not carry through.
• ⚠️ The pre-market directional call was falsified right here: this morning's pre-market Section 9.4 wrote that "overnight US equities rose moderately (Nasdaq +1.00%, Amazon +15%) … a gap-up open is the high-probability case", whereas in reality the SSE Composite gapped down −0.51%, the SZSE Component gapped down −0.60%, and all four major indices opened lower. The positive spillover from US equities did not translate into the A-share open at all; instead the negative spillover from Apple −7% and the intraday plunge in US storage names (pre-market news item 11) drove today's semiconductor sell-off. Note this one: pre-market treated the "index gain" of overnight US equities as the direction of spillover and ignored the "internal structure", and so got the direction backwards.

1.2 Sentiment Indicators

Sentiment read: a warm-up, but a "shrinking-volume, structural" warm-up, not euphoria. Three pieces of evidence sit side by side and do not all point the same way, so they must be read together:

1. Positive: broken limit-ups collapsed from 107 to 16, and the seal rate jumped from 48.1% to 82.4% — this is the strongest positive signal of the day. On 7/31 limit-ups kept getting cracked open; today, once sealed they stayed sealed. Follow-through quality was far better than Friday's, which is exactly the opposite of the checkpoint set in Section 8.2 of this morning's pre-market report, "if the number of broken limits again exceeds the number of limit-ups before 10:00 today, follow-through is still insufficient" — follow-through clearly improved.
2. Negative: limit-downs went from 0 to 8, and 5 of them were heavyweight-grade limit-downs on expanded volume: GigaDevice (603986) RMB 23.033 bn of turnover, China Jushi (600176) 10.734 bn, Honghe Technology (603256) 2.951 bn, Xiandao Jidian (600641) 1.949 bn, Hongqiao Holdings (002379) 1.391 bn. These are not small-cap limit-downs; this is a stampede in mainstream names.
3. Negative: yesterday's 99 limit-up stocks averaged a gain of only 1.31% today (median +0.91%), 42.9% closed outright in the red, 13 fell more than 5%, and the promotion rate was just 17.3% (17/98, excluding the suspended Ailijiaju). The vast majority of Friday's limit-up wave had no second day.

Overall conclusion: this was a rotation day in which the old ladder broke down en masse while a new ladder rebuilt itself on a high seal rate — not a continuation of a broad rally. The money-making effect concentrated in the newly launched power equipment chain, and the money-losing effect concentrated in the old semiconductor / AI hardware chain.

---

2. Pre-Market List Reconciliation

Reconciliation rules (standards set before scoring, so the goalposts cannot be moved after the fact):

• Long recommendation list (priority deep-dive / watch closely): "today's gain > +3%" counts as delivered; +0%~+3% counts as "weakly delivered"; closing in the red counts as not delivered.
• Negative list (watch only / Pass / not worth chasing): the pre-market expectation was "weak", so "today's gain < +3%" counts as a correct call and a gain ≥ +3% counts as a wrong call.
• All price moves and turnover come from Tencent real-time quotes (8/3); limit-up / broken-limit / consecutive-board status comes from the East Money limit-up pool (8/3).

2.1 Long Recommendation List Reconciliation (pre-market Section 9.1, "the 5 main-board names most worth watching today")

Long recommendation list hit rate:

• Strict scope (gain > +3% counts as delivered): 3/5 = 60% (Dongfang Electric, Weaver Network, Jiangsu Shentong Valve)
• Loose scope (beating the SSE Composite's −0.59%): 5/5 = 100%
• Limit-up hits: 1/5 = 20% (Jiangsu Shentong Valve)

2.2 Negative List Reconciliation (pre-market "watch only", "Pass", "do not chase", 21 names in total)

This is the best-performing part of this issue, and the part that best demonstrates that "fundamental verification works".

Negative list hit rate: 16/21 = 76.2% (5 wrong: Leo Group, Meili Cloud, Tianyu Digital Science, Fenghua Advanced Technology, Gravity Media)

23 ⚠️ The Biggest Misjudgment of This Issue: 利欧股份002131Leo Group

This needs its own section, because it was the only name pre-market handled at "single-veto" severity, and today it had the highest turnover of any limit-up stock market-wide.

• Pre-market conclusion: Pass, and specifically "the highest governance risk on the entire list", triggering a single-veto item (the Shenzhen Stock Exchange's public censure of the company, its chairman and its CFO on 2026-07-03). Pre-market Section 8.3 also stated specifically that "the first draft here originally wrote 'funds will rotate into the lower-positioned Leo Group'; that judgment has been voided".
• What actually happened today: +10.09%, a 2nd consecutive limit-up, closing at 5.13, turnover of RMB 7.001 bn, a 24.00% turnover rate, and the limit broken 22 times intraday before being resealed at 13:42.
• Review — what was wrong, and what was not:
  • What was not wrong: the fundamental and governance judgments have still not been falsified. A three-year ex-non-recurring net margin of 0.41%, the first ex-non-recurring loss in an interim report, public censure of the chairman and CFO — not one of those facts changed today, and today's rally supplied no fundamental evidence whatsoever.
  • What was wrong: pre-market equated "fundamentals/governance fail" directly with "the price will not rise", and that is a label overreaching its remit. The pre-market report itself wrote in Section 5 that "share-price elasticity and profit delivery are almost uncorrelated", yet forgot that line when assigning labels — a name with RMB 30 bn of free float, a 24% turnover rate, and the ability to reseal after 22 broken limits is a pure capital-flow battleground, and using a financial model to negate its price action means the tool itself is the wrong one.
  • An even more damaging internal contradiction: pre-market listed the lower position as one of Leo Group's characteristics (−22.1% over the past 45 sessions), while simultaneously using "a low position is not a reason to participate" as a supporting rationale for the Pass. Yet all three names that posted 2nd boards today (Leo Group +10.09%, Tianyu Digital Science +9.98%, Meili Cloud +10.01%) were exactly low-position catch-up names — a low position was the single most effective positive factor today, and pre-market treated it as neutral or even negative.

2.4 One-Line Review: What Was Right, What Was Wrong, and What to Change Tomorrow Morning

What was right was "what will fall"; what was wrong was "what will rise". What was right was the branch; what was wrong was the stock structure inside the branch.

• ✅ The category we got most right: "avoids" backed by hard evidence. GigaDevice limit-down, Demingli −9.56%, Taiji Industry −5.51%, Yonyou Network stalling on expanded volume, Servyou spiking and fading — almost everything pre-market rejected on financial or fund-flow data (semiconductor net outflow of RMB 20.5 bn, PE 130x, three straight years of losses) played out today. Fundamental verification hit 16/21 in the "mine-clearing" direction; as a defensive tool it works.
• ❌ The category we got most wrong: treating "fundamentals fail" as a forecast of "it will not rise". Leo Group, Meili Cloud and Tianyu Digital Science all posted 2nd boards, all three came out of the pre-market "watch only / Pass" bucket, and all three were low-position catch-up capital-flow names. Fundamental verification is ineffective — indeed inverted — as a tool for forecasting short-term price gains.
• ❌ Second type of error: the branch was right, the stock structure was backwards. The nuclear branch was right (net inflow of RMB 3.343 bn, first on the board), but pre-market treated the operator end (China National Nuclear Power, CGN Power) as the two hardest names, whereas in reality the equipment and construction ends (Dongfang Electric +6.69%, Jiangsu Shentong Valve +10.00%, China Nuclear Engineering +10.05%, CNNC Sufa +7.04%, Rongfa Nuclear Power +9.93%) comprehensively beat the operator end (−0.33%, +0.73%). The reason is clear: the State Council executive meeting approvals mean installed capacity 5—6 years out for operators but tenders 1—2 years out for equipment makers — the market picked the shorter lag, and pre-market had actually computed both lags in Section 7.2 but never applied them to the ranking.
• ❌ Third type of error: the strongest branch of all was missed entirely. See Section 3.

→ Three concrete improvements for tomorrow morning's pre-market list:

1. Split "research attention" and "short-term price elasticity" into two fully separate columns; stop making one label carry both jobs. Proposal from tomorrow morning: fundamental verification outputs only "fundamentals pass / fail", position and chip structure output only "low position / high position", two independent parallel columns, no longer merged into the single hybrid label of "priority deep-dive / watch only / Pass" — Leo Group today is direct evidence that the hybrid label fails.
2. Within a branch, rank stocks by "the lag from catalyst to revenue", not by "whether the positive news named them". Today already proved it: an equipment end with a 1—2 year lag > an operator end with a 5—6 year lag, even when the latter is named in the announcement.
3. Give "low-position catch-up" an explicit positive weight. All three 2nd boards today were among the biggest 45-day decliners, while Zhidu (000676), at a 45-day high, rose only 1.34%. The pre-market scoring model is right to cap "expectation gap" at ≤2 pts for names at new highs, but it has no corresponding bonus for deep drawdowns.

---

3. Main-Line Validation

3.1 Did Pre-Market Call the Strongest Main Line? — It Got the "Name" Right and the "Scope" Wrong

What was right: pre-market ranked nuclear power as branch No. 1, and today the nuclear power concept led all concept sectors market-wide with a net inflow of RMB 3.343 bn — the direction was entirely correct. The pre-market read that "it was not reflected before the 7/31 close, so today is the first pricing day" also holds — the nuclear complex expanded volume collectively today (Dongfang Electric turnover +98.6%, China National Nuclear Power +82.7%, Jiangsu Shentong Valve +116%, CGN Power +68%).

What was wrong: nuclear power is only a subset of today's main line; the real main line is the larger set of "power equipment and electricity". The evidence: 4 of the top 5 THS industry gainers belong to the power equipment family (wind power equipment +4.06%, grid equipment +3.52%, environmental protection equipment +2.99%, solar equipment +2.91%), the electricity industry ranked first among all industries with a net inflow of RMB 3.211 bn, and grid equipment led all industries market-wide with 8 limit-ups.

3.2 ⚠️ The Unexpected Main Line Pre-Market Missed: Grid Equipment

This is the omission most worth recording this issue, because the clue was already in the pre-market report — it just was not put to use.

• Item 6 of the pre-market news overview already stated: "State Council executive meeting: push forward solidly with the 'Six Networks' plan and construction; the meeting heard reports on new-type power grid and logistics network construction and called for 'building out new-energy infrastructure networks and advancing intelligent retrofits of grid infrastructure plus smart microgrid construction'; grid fixed-asset investment during the '15th Five-Year Plan' period is projected to exceed RMB 5 trillion", with an impact level of A.
• But in the branch table in Section 2, this item was folded into the "compute network / 'Six Networks'" branch (strength B+), and the representative stocks given were compute/IDC names — 云赛智联 (600602), 润建股份 (002929), 深桑达Ａ (000032), 美利云 (000815) — plus two power-supply names, 卧龙电驱 (Wolong Electric, 600580) and 麦格米特 (002851). Not one of them is a grid-equipment company.
• Today's outcome: grid equipment +3.52% with 8 limit-ups (Jiusheng Electric +20%, Shunna, Baili Electric, Changlan Technology, Hualing Cable, Huijintong, Fengfan, Great Wall Electrical), while the compute names pre-market actually gave ranged from −5.51% (Taiji Industry) to +2.36% (Runjian), and the two power-supply names split to the extremes, Wolong Electric +8.90% versus Megmeet −7.13%.
• Where the error lies: pre-market mapped a policy headline about "RMB 5 trillion of grid investment" onto "compute centers", the downstream demand side, instead of "grid equipment", the direct supply side. The State Council text said "intelligent retrofits of grid infrastructure and smart microgrid construction" — the direct beneficiaries of that sentence are transmission and distribution equipment manufacturers, not IDC operators. This was an error in the direction of the value-chain mapping, not a gap in information.

---

4. Limit-Up Ladder / Height Boards

4.1 Full Consecutive-Board Ladder (17 names with ≥2 boards)

4.2 Key Readings

1. The ceiling fell from 9 boards to 6, and the highest name is a "shrinking-volume locked board", not a "strong high-volume board". Chuanzhi Education (003032) locked the limit at 09:25 today with all-day turnover of only RMB 79 mn, a 2.88% turnover rate and zero amplitude — this is a limit sealed on no volume, meaning the chips are locked but no incremental money is participating; it looks more like "nobody is selling" than "many people are buying". This morning's pre-market positioned it "as a sentiment indicator, not as a name to own", and that positioning looks accurate today: it advanced, but it supplies no information about capital strength.
2. ⚠️ Today's "main force" is not in the consecutive-board ladder, it is among the first boards. The largest turnover in the consecutive-board ladder is Leo Group's RMB 7.001 bn, but it belongs to the fading AI application chain; the high-turnover names that genuinely represent the new main line are all first boards: 华电辽能 (600396) +10.00% on turnover of RMB 3.410 bn with a 15.62% turnover rate (electricity), 通威股份 (600438) +10.02% on turnover of RMB 1.549 bn (solar), 顺钠股份 (000533) +9.99% on turnover of RMB 1.261 bn with a 16.88% turnover rate (grid equipment), 中国核建 (601611) +10.05% on turnover of RMB 867 mn (nuclear construction). The new main line has breadth today but no height yet.
3. Board breakdowns: 42.9% of yesterday's limit-up stocks closed red today, and 13 fell more than 5%. The worst breakdowns cluster in yesterday's strongest names: 普联软件 (300996) −8.22% (2nd board broken, 30.48% turnover rate), 麦格米特 (002851) −7.13%, 宏景科技 (301396) −5.73%, 太极实业 (600667) −5.51%, 泰坦股份 (Titan, 003036) −9.19%, 海星股份 (Haixing, 603115) −7.41%, 宝鼎科技 (Baoding Technology, 002552) −7.25%. Yesterday's limit-up names in the compute / AI hardware chain were today's disaster zone.
4. Seal quality was extremely polarized. On one side, clean boards with zero breaks (Chuanzhi Education, Gaozheng Explosive, Infore Environment, Zhongyan Dadi, Delong Huineng, Hengyin Technology, Haoersai and Jiangsu Shentong Valve all had 0 breaks); on the other, repeated cracking: Leo Group broken 22 times, Shenwu Energy Saving 8 times, Meili Cloud 6 times, 百合花 (Lily Group, 603823) 10 times. Leo Group, which resealed after more than 20 breaks, is the single most contested name of the day.
5. All 8 limit-downs cluster in two chains: the semiconductor chain (GigaDevice −10.00% / 23.033 bn, Xiandao Jidian −10.00% / 1.949 bn) and the fiberglass / materials chain (China Jushi −9.99% / 10.734 bn, Honghe Technology −10.00% / 2.951 bn); plus 宏桥控股 (Hongqiao Holdings, 002379) −10.01%, 麦迪科技 (Maidi Technology, 603990) −9.98%, 东晶电子 (Dongjing Electronic, 002199) −10.02%, 明新旭腾 (Mingxin Xuteng, 605068) −10.00%. ⚠️ The limit-downs in GigaDevice and China Jushi alone, on combined turnover of RMB 33.767 bn, were one of the main contributors to today's index decline.

---

5. Fund Flows

⚠️ Two key data items in this section could not be obtained, stated up front (see item 4 of the scope disclosure at the top):

• Today's Dragon-Tiger List: no reliable data — the exchange had not published it as of writing (the East Money endpoint returned "the returned data is empty"), so this report names no Dragon-Tiger stocks, net buy amounts or seat preferences for today. Tomorrow's pre-market should go back and read today's Dragon-Tiger List, focusing on the nature of the buy/sell seats in 利欧股份 (002131), 华电辽能 (600396) and 久盛电气 (301082).
• Northbound single-day net flow: no reliable data — the Shanghai/Shenzhen Connect "net buy turnover" returned 0.00 with trading status code 3, i.e. not disclosed (the exchanges stopped disclosing single-day net figures from August 2024). This report states no direction for northbound inflows/outflows. The only thing available is market breadth on the Stock Connect scope: Shanghai 1078 up / 543 down, Shenzhen 1357 up / 498 down.

5.1 Industry Main-Force Net Inflow/Outflow (THS scope, unit: RMB bn)

⚠️ This table is the clearest picture of the day: money moved wholesale out of "semiconductors/electronics" and into "electricity/grid/solar/wind". The semiconductor single-day net outflow of RMB 9.316 bn is 4.6x the runner-up (electronic chemicals at 2.033 bn), while electricity plus grid equipment plus solar plus wind took in a combined net RMB 7.554 bn. This is not sector rotation inside a broad rally; this is a migration of capital with a clear direction.

A parallel comparison with this morning's pre-market: item 6 of the pre-market scope disclosure wrote that "although semiconductors closed +4.23% on 7/31, the sector saw a net outflow of RMB 20.502 bn, the largest net outflow of any sector market-wide", and concluded on that basis that "rising while being sold and rising while being bought are two different things". Today semiconductors caught down in price at −3.46% while money kept leaving at 9.316 bn — that pre-market judgment that "the money is already heading for the exit" was fully validated today. It is the single highest-value piece of fundamental/fund-flow analysis in this issue.

5.2 Concept Sector Main-Force Net Inflow (East Money scope, unit: RMB bn)

Reading: the nuclear power concept led all concept sectors market-wide with a net inflow of RMB 3.343 bn, the strongest single validation of this morning's pre-market call to rank nuclear power as branch No. 1. But note at the same time that pumped-storage, ultra-supercritical generation, flexible HVDC, solar thermal power and controlled nuclear fusion, ranked just above and below nuclear power, all belong to the larger theme of "power equipment / generation" — what the market bought today is the broad direction of "electricity", and nuclear power is merely its most elastic sub-line. One counter-example must be set alongside it: virtual power plants rose 2.64% yet saw a net outflow of RMB 948 mn, i.e. "up but sold", which makes it an unsuitable way to participate in the power main line tomorrow.

5.3 Stock-Level Capital Strength (proxied by turnover and turnover rate, as the Dragon-Tiger List is unpublished)

• The 6 largest limit-up stocks by turnover today: 利欧股份 (002131) 7.001 bn (24.00% turnover rate), 华电辽能 (600396) 3.410 bn (15.62% turnover rate), 通威股份 (600438) 1.549 bn, 百合花 (603823) 1.543 bn (broken 10 times), 美利云 (000815) 1.526 bn, 顺钠股份 (000533) 1.261 bn (16.88% turnover rate).
• The 3 largest limit-down stocks by turnover today: 兆易创新 (603986) 23.033 bn, 中国巨石 (China Jushi, 600176) 10.734 bn, 宏和科技 (603256) 2.951 bn.
• Volume changes in the new main line's core names (vs 7/31): 东方电气 (600875) 1.283 bn → 2.548 bn (+98.6%), 中国核电 (601985) 1.002 bn → 1.831 bn (+82.7%), 江苏神通 (002438) 85 mn → 184 mn (+116%), 中国广核 (003816) 425 mn → 714 mn (+68%). All four nuclear names expanded volume significantly, which is hard evidence that the branch is real.
• Volume/price divergence in the old main line's core names: 用友网络 (600588) traded RMB 2.756 bn yet rose only 1.08%, 蓝色光标 (300058) traded 12.015 bn (23.48% turnover rate) yet rose only 2.58%, 智度股份 (000676) traded 1.650 bn and rose 1.34%. Large turnover + small gain = a textbook churn-and-distribute pattern.

---

6. Outlook for Tomorrow

6.1 Main-Line Continuation Assessment

6.2 Focus List for Tomorrow (branch + main-board names + validation points)

⚠️ The following is a ranking of research attention and contains no buy or sell recommendations.

6.3 Avoid Tomorrow

1. The entire semiconductor/storage chain (兆易创新 603986, 德明利 001309, 生益科技 600183, 深南电路 002916, 先导基电 600641) — a single-day net outflow of RMB 9.316 bn, a leader hitting limit-down on RMB 23.0 bn of turnover, and the STAR 50 at −5.08%, all three at once. Do not bottom-fish on the day of a limit-down; wait for at least one high-volume stabilizing up-candle.
2. Yesterday's AI application chain names stalling at highs (用友网络 600588, 智度股份 000676, 税友股份 603171, 浙文互联 600986, 三维通信 002115, 云赛智联 600602) — all of them posted "large turnover + a gain of <1.5%" today, a textbook churn-and-distribute pattern; not one of the pre-market fundamental objections was overturned today.
3. The compute network concept (太极实业 600667, 深桑达Ａ 000032, 润建股份 002929, 麦格米特 002851) — boards broke and prices fell today, and the pre-market conclusion that "not one company can prove with financial data that its compute business is growing" still stands.
4. High-position consecutive boards and low-volume locked boards (传智教育 003032 at six boards, 一鸣食品 605179 at five boards) — Chuanzhi Education traded only RMB 79 mn today at a 2.88% turnover rate, a locked board on no volume, so once it opens there is no follow-through; on top of that the Shanghai Stock Exchange applied self-regulatory measures to 137 abnormal trading incidents this week, and the 9-board Ailijiaju (603221) is suspended today — the regulatory environment is clearly tightening around high-position consecutive boards.
5. Names with more than 20 broken limits and heavy disagreement (利欧股份 002131) — it took a 24% turnover rate and 22 broken limits to seal today, making it the most contested name on the board; its governance red flags (public censure of the company, its chairman and its CFO by the Shenzhen Stock Exchange) and its first ex-non-recurring interim loss did not change at all today. ⚠️ This report acknowledges it got the price direction wrong on this name (see Section 2.3), but it does not revise its fundamental and governance judgment because the price went up — those are two different things.
6. The virtual power plant concept — up 2.64% today yet with a net outflow of RMB 948 mn, the only sub-line within the broad power theme that "rose while being sold", which makes it an unsuitable way to participate in the power main line.

6.4 Input Notes for Tomorrow Morning's Pre-Market List

1. The main-line rotation is complete; branch No. 1 in tomorrow morning's pre-market should be the "power equipment (grid / wind / solar / pumped-storage / nuclear)" family rather than nuclear power listed on its own. Today's evidence: 4 of the top 5 industry gainers belong to this family, all of the top 4 industries by net inflow belong to it, and 7 of the top 8 concepts by net inflow belong to it.
2. Stock-level research on grid equipment must be added — this is the biggest miss of the day. Tomorrow morning's pre-market needs at least one round of fundamental verification (share of revenue tied to grid investment, order backlog, current earnings direction) on 长缆科技 (002879), 华菱线缆 (001208), 风范股份 (601700), 汇金通 (603577), 长城电工 (600192), 百利电气 (600468), 顺钠股份 (000533); today's pre-market preparation on this chain was zero.
3. Go back and read today's Dragon-Tiger List (not obtained for this report). Focus on verifying: whether the RMB 7.001 bn of turnover behind 利欧股份 (002131) was hot money or institutions, and the nature of the buy seats in 华电辽能 (600396) and 顺钠股份 (000533) — this directly determines whether tomorrow's power main line should be treated as "money accepting the logic" or "hot money trading a concept".
4. Change the scoring model to a two-column output (fundamentals pass/fail, position high/low) and stop merging them into a single label. All three of today's 2nd boards — Leo Group, Meili Cloud, Tianyu Digital Science — came out of the "watch only / Pass" bucket and all three were low-position catch-up names. The current label system fails systematically in the quadrant of "fundamentals fail but chip position is good".
5. Within a branch, rank by "the lag from catalyst to revenue", not by "whether the name was called out". Today's nuclear branch already supplies direct evidence: the equipment/construction end (1—2 year lag) comprehensively beat the operator end named in the announcement (5—6 year lag).
6. Spillover from overnight US equities must be read through internal structure, not the index move. All four major indices gapped down today and the STAR 50 fell −5.08%, while pre-market had used "Nasdaq +1.00%" to conclude that "a gap-up open is the high-probability case" — what actually transmitted through was Apple −7% and the plunge in US storage names, not the Nasdaq's positive return.

---

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