A-Share Pre-Market Brief | 2026-08-27, Thursday

fundamentals-analyst verified against the original interim report PDF (157 pages), the 7/13 pre-announcement filing (No. 2026-035), the 7/22–24 investor-meeting minutes (No. 2026-17) and stock-by-stock K-line comparisons with peers; risk-auditor then re-checked the actual upload dates of the filings' art_codes. The checks overturned three of my premises, all written out below.

• Related news (8/26 18:41; today is the day after the interim report = the reaction day): revenue RMB 15.296 billion (+46.33%), attributable net profit RMB 2.251 billion (+65.55%), ex-non-recurring RMB 2.239 billion (+76.94%); gross margin 31.01% (+4.74pct), weighted ROE (half-year) 12.43%, net operating cash flow RMB 1.502 billion (+0.42%). The same evening it also released an "Announcement on Provision for Asset Impairment" of RMB 363.58 million (receivables RMB 137 million + inventory RMB 207 million + fixed assets RMB 20 million), already charged against current-period profit.
• ⚠️ Correction one (timing): the draft wrote that "the RMB 4.882 billion private placement was accepted by the SZSE" happened "the same evening" as a new dilution risk — wrong. That filing was uploaded on the evening of 8/25, and the market traded all of 8/26 on this information, delivering +2.06% and the highest net-inflow ratio in the chain; moreover the placement plan was made public as early as 6/13, received SASAC approval on 7/9 and shareholder-meeting approval on 8/14, so the 8/26 "acceptance" is merely a procedural node containing neither an issue price nor a share count. By the same standard this issue applies to 生益科技 Shengyi Technology and 剑桥科技 Cambridge Industries, it is old news and does not constitute a new variable today. (The dilution itself remains undetermined and is listed as an ongoing risk.)
• ⚠️ Correction two (wording): the full 157-page interim report contains the word "HBM" 0 times. The company only writes "BT-type IC substrates, memory products benefited from the sharp rise in memory demand driven by AI applications, and related orders grew significantly." Writing it up as "HBM-related IC substrates" is extrapolation. The full text likewise contains no customer names — no NVIDIA, no North American customer wording.
• ⚠️ Correction three (relative position): I had assumed "peers fell more over the same period" — that does not hold. Computed by me over the same interval (Tencent, forward-adjusted): 60-day (6/2→8/26) 景旺电子 Kinwong +19.28%, 生益科技 Shengyi Technology −7.31%, 东山精密 Dongshan Precision −7.86%, 沪电 WUS −8.99%, 深南 Shennan −12.73%, 广合科技 Guanghe Technology −15.91%, 生益电子 Shengyi Electronics −18.71%, 兴森科技 Fastprint −21.13%, 胜宏 Victory Giant −28.54% — 深南 Shennan is the median of this group, with no idiosyncratic negative. The rebound leg matters more: 20-day (7/29→8/26) 胜宏 Victory Giant +23.72%, 生益科技 Shengyi Technology +20.18%, 景旺 Kinwong +19.69%, 生益电子 Shengyi Electronics +14.57%, 沪电 WUS +13.27%, 深南 Shennan +10.13% (second from bottom in the group). 深南 Shennan is a laggard in this AI-PCB rebound, not a defensive outperformer.
• Catalyst logic (direct · affects revenue): its main businesses are PCB, IC substrates and electronic assembly, with downstream demand from AI servers and high-speed switches. NVIDIA explicitly states "inventory rose from $25.8B to $31.6B, building for the Vera Rubin ramp this quarter" — a platform generation change is an across-the-board step up in layer count, material grade and unit price for PCB.
• Industry branch stage: low position, no ladder. Among the 52 names in the official limit-up pool on 8/26, not a single one was from the PCB sector (only 同宇新材 Tongyu New Materials, an upstream electronic chemicals name, hit a 20cm limit-up). If this line moves today, it is day one, with no overhang from high-positioned names.
• The expectation gap is not in the aggregate, it is in the structure (the core of this item):

  • Aggregate = in line. The 7/13 pre-announcement gave a dual basis: attributable net profit RMB 2.10–2.30 billion, ex-non-recurring RMB 2.08–2.28 billion. Actual attributable net profit of RMB 2.251 billion lands at the 75.7% percentile and ex-non-recurring of RMB 2.239 billion at the 79.6% percentile — both in the top quartile, but neither broke the upper bound. This is "high-quality delivery on the pre-announcement," not an "expectation gap." Attributable minus ex-non-recurring is only RMB 12 million, so non-recurring gains are near zero — a very clean set of books.

  • The genuinely incremental information is the product-line structure, of which the pre-announcement said not a word:

    Business	H1 revenue (RMB bn)	Share	YoY	Gross margin	Gross margin YoY
    Printed circuit boards	8.552	55.91%	+36.32%	36.85%	+2.43pct
    IC substrates	3.667	23.97%	+110.76%	31.35%	+16.20pct
    Electronic assembly	1.976	12.92%	+33.70%	17.30%	+2.32pct

    Table scrolls horizontally

    IC substrate revenue more than doubled and its gross margin rose 16.20 percentage points in one year (15.15%→31.35%) — direct evidence that the Guangzhou plant has moved from "loss-making ramp" to "profit contributor."

  • The AI basis is written as the company wrote it: the only quantified statement in the interim report is "in the data center field… the company's related order revenue surpassed RMB 2.0 billion, growing sharply year over year and becoming the key driver of PCB business growth" (P13). ≈ RMB 2.0 billion / RMB 15.296 billion = 13.1% of total revenue, and 23.4% of the PCB business. Whether this means "order revenue" or "order value," and whether it is limited to the PCB segment, are both readable from the Chinese original, so this issue writes only "data center related order revenue surpassed RMB 2.0 billion, about 13% of total revenue," and does not write it as an "AI revenue share."

  • Single-quarter split (interim report − Q1 report, not a media estimate): Q2 revenue RMB 8.701 billion (+53.44% YoY / +31.92% QoQ), attributable net profit RMB 1.401 billion (+61.33% / +64.81%), ex-non-recurring RMB 1.390 billion (+78.13% / +63.60%), gross margin 32.41% (+3.24pct QoQ). Q2 is the strongest quarter of the year.

  • One basis trap: Q2 gross margin was +4.82pct, yet the attributable net margin was only +0.78pct. The difference was eaten by four items — share-based payment RMB 263 million (zero a year ago), finance expense RMB 151 million (RMB 21 million a year ago, attributed to FX), asset + credit impairment RMB 363 million (RMB 203 million a year ago), and a RMB 138 million decline in government grants. Stripping out only the first-time share-based payment, H1 comparable attributable net profit would be about RMB 2.51 billion, roughly +85% YoY (an estimate, not company-disclosed).

• The balance sheet is the least attractive part of this report: the gearing ratio rose from 43.82% to 55.67% (+11.85pct) in half a year; interest-bearing debt went from RMB 3.510 billion to RMB 7.928 billion (+126%); quick ratio 0.71; free cash flow = −RMB 3.177 billion (−RMB 44 million in the prior-year period); contracted capital commitments swelled from RMB 2.116 billion to RMB 6.936 billion (3.28x), plus construction in progress of RMB 3.248 billion, for a combined roughly RMB 10.18 billion awaiting capitalization = 68% of current net fixed assets. Essentially none of this period's profit converted into freely disposable cash; expansion is funded by debt and equity issuance. Within inventory, raw materials rose from RMB 1.322 billion to RMB 2.603 billion (+96.9%) — if copper clad laminate / copper foil prices peak and fall back, that RMB 2.603 billion is direct impairment exposure.
• Valuation basis correction: the previously cited "PE 52.65" was the dynamic P/E (H1 annualized). Recalculated including the interim report, PE (TTM) = 56.89x; static (2025A) 72.37x; Q2 single-quarter annualized 42.30x. PB (MRQ) 13.15x sits at the 95.6th percentile of 5 years (5-year range 1.92 / median 4.10 / high 17.30), and PE (TTM) at the 87.0th percentile of 5 years. Note that the PB percentile is above the PE percentile — this is not the "low PE + high PB" cycle-top shape, but earnings and valuation rising together. Also: in the East Money research database the latest sell-side report on this stock is dated 2026-03-26, with 0 reports in the past month, so figures like "2026E PE 43x" are not reliable today; "Q2 single-quarter annualized 42.3x" is the more honest number.
• Technical sentiment (SZ main board ±10%): closed 8/26 at RMB 348.03 (+2.06%), net main-force inflow RMB 264 million (third by absolute amount in the sample, behind 沪电 WUS's RMB 622 million and 胜宏 Victory Giant's RMB 368 million); turnover rate 1.11%, volume 73,982 lots, the lowest in nearly 3 months; 20-day +10.13%, 60-day −12.73%, dist. 60d high −28.18%. No recent limit-ups and no consecutive limit-ups, so there is no one-day-wonder structure; but it is a high-beta name — 7/28 −10.00%, 7/30 −9.97%, 8/4 +10.00%, 8/19 −9.93%, repeatedly hitting the daily limits over the past month.
• Final judgment: among the 26 names in this issue, it is the only one where "money was already in before the results became public" — 沪电 WUS's RMB 622 million is a day-after-interim-report reaction, 胜宏 Victory Giant's RMB 368 million corresponds to an interim report not yet released, and only 深南 Shennan's RMB 264 million occurred before 8/26 18:41. Its bull case is not "results beat" (the aggregate landed at the 75.7%/79.6% percentiles, i.e. in line), but three numbers that the pre-announcement did not contain and that were given for the first time last night: IC substrate +110.76% with gross margin up 16.20pct, data center order revenue surpassing RMB 2.0 billion, and capital commitments swelling to RMB 6.936 billion pointing to 2027 capacity — and these three numbers were disclosed on the lowest turnover volume in nearly three months, so the information has not yet been absorbed by trading. The bear case is equally clear: the gearing ratio jumped 11.85pct in half a year, free cash flow is −RMB 3.177 billion, PB is at the 95.6th percentile of 5 years, and the 20-day rebound is second from bottom in its group. Earnings and leverage are accelerating together on the same balance sheet.

This name was added to the list after QC. The memory branch in the draft included only module makers and niche memory, missing the layer on the chain closest to "how much memory NVIDIA bought."

• Related news: ① the NVIDIA CFO document, $119B → $279B, "primarily related to memory purchases"; ② on 2026-07-17 the company released a "Voluntary Announcement on the Pre-Announcement of Increased 2026 Interim Results" (No. 2026-038): expected H1 attributable net profit of RMB 1.900–2.100 billion (+63.90%~+81.20%) and expected revenue of about RMB 3.335 billion; excluding share-based payment, attributable net profit of RMB 2.080–2.280 billion (+56.1%~+71.1%). The interconnect chip product line generated about RMB 3.111 billion in H1 (+26.4%), of which Q2 was about RMB 1.694 billion (+28.2% YoY, +19.5% QoQ). The company attributes this to rising DDR5 penetration with continued sub-generation iteration lifting RCD chip shipments, and to rising revenue from new products MRCD/MDB, PCIe Retimer, CKD and CXL MXC. (Shanghai Securities News filing page⁶)
• Catalyst logic (the most direct memory mapping on this list): the more memory NVIDIA locks up and the more DIMMs plugged into each server, the higher the usage of DDR5 RCD / MRCD / MDB / CKD. This is not the "memory price increase" price chain but the "per-box memory capacity" volume chain — its correspondence with the $279B purchase commitment is more direct than that of module makers or niche memory. There is no second name of scale at this layer in A-shares.
• Industry branch stage: not started. It rose +3.38% on 8/26 (among the top gainers in the compute chain that day), but is still −3.13% over 20 days and −39.55% below its 60-day high; there was no memory-interface-related name in the official limit-up pool on 8/26.
• Fundamentals: H1 pre-announcement basis attributable net profit RMB 1.900–2.100 billion; PE (TTM) 95.92, PB 12.17, total market cap RMB 245.3 billion; 8/26 turnover value RMB 5.973 billion, turnover rate 2.64%. The interim report is not yet disclosed and the statutory deadline is August 31 — it must land between today and next Monday.
• The negative comparison that must be written down: on the day it released this pre-announcement on 7/17, the stock fell more than 8%, while the company also announced a planned RMB 300–600 million buyback. This shows the market has already reacted negatively once to its "good news"; the point of disagreement then was whether the growth rate was already fully expected. What NVIDIA provides today is a new and larger demand anchor, but this stock has proven with past facts: its good news is not necessarily bought.
• Final judgment: on the logic chain it is the closest of the 26 names in this issue to the "$279B memory purchases," it was already moving on 8/26 (+3.38%), it is still 39.55% below its 60-day high, and the interim report is a near-term variable yet to land. The risks are valuation (PE 95.92) and the prior record of "an 8% drop on pre-announcement day." Watch closely.

• Related news: no company-level news; it depends on the NVIDIA guidance mapping.
• Catalyst logic: direct. As the main global AI server ODM, NVIDIA's Q3 guidance of $108B and the Rubin ramp this quarter correspond to demand for full-system assembly and thermal modules.
• Fundamentals: PE (TTM) 25.62 — the lowest on the whole list, PB 6.84, total market cap RMB 1,202.0 billion.
• Technical sentiment: closed 8/26 at RMB 60.57 (+0.70%), net main-force inflow RMB 9 million (essentially zero), turnover rate only 0.37%, turnover value RMB 4.413 billion; 20-day +6.08%, 60-day −23.68%, dist. 60d high −28.15%.
• Final judgment: a turnover rate of 0.37% is the lowest on the list, meaning the market is simply not trading this stock on the NVIDIA results. Its merits are a low valuation, a low position, and the fact that low ODM elasticity also means smaller drawdowns; its drawback is that a trillion-RMB market cap needs enormous money to move, so it is not an elasticity name. Its main use is as a criterion for "is today an index-level compute rally," not as a source of elasticity. Watch closely.

• Related news: no company-level news. The interim report is not yet disclosed (the latest filing stops at 8/12) and the statutory deadline is August 31 — it must fall between 8/27 and 8/31, making it a near-term two-way variable.
• Technical sentiment: closed 8/26 at RMB 249.44 (+1.82%), net main-force inflow RMB 368 million (+5.96%), turnover rate 2.86%; 20-day +23.72% (No. 1 in the AI-PCB rebound), 60-day −28.54%, dist. 60d high −33.62%.
• Final judgment: it has leading fund backing within the PCB branch, but it is the biggest gainer of this rebound (20-day +23.72%) and its interim report has not landed. The contrast with 深南电路 Shennan Circuits is clear: 深南 Shennan is "results delivered, gained the least," 胜宏 Victory Giant is "results not delivered, gained the most." Watch closely, and the discount for an event that has not landed must be counted in.

There are two reasons for removal, both found by QC.

• ⚠️ Correction (timing): the interim report was disclosed on the evening of 8/25, not the evening of 8/26. That one-day difference changes the entire argument — the +4.31% and net main-force inflow of RMB 622 million (first by amount in the sample) on 8/26 should be explained mainly as a day-after-interim-report reaction, not as "money pre-positioning into PCB on the eve of NVIDIA's results." The draft treated it as the core evidence for "money is choosing PCB"; that evidence does not hold.
• ⚠️ Correction (data): the draft wrote "actual figures not obtained, left blank" — they are in fact publicly available. H1 revenue RMB 13.689 billion (+61.17%), attributable net profit RMB 2.923 billion (+73.72%), ex-non-recurring RMB 2.795 billion (+70.00%), weighted ROE 17.38%, operating cash flow RMB 1.151 billion (−45.14% YoY).
• Expectation gap (turns neutral-to-weak after recomputation): attributable net profit landed at roughly the 54.7% percentile of the 7/13 pre-announcement (RMB 2.83–3.0 billion), ex-non-recurring at roughly the 43% percentile (RMB 2.73–2.88 billion). The ex-non-recurring percentile is below the attributable percentile — the opposite direction from 深南电路 Shennan Circuits (79.6% > 75.7%), meaning its outperformance is not purely operational the way 深南 Shennan's is.
• Technical sentiment: 20-day +13.27%, 60-day −8.99%, dist. 60d high −24.64%. On 8/26 it spiked to +6.65% intraday before closing +4.31%.
• Final judgment: the results themselves are very good (revenue +61.17%, ex-non-recurring +70.00%), but for today it is "yesterday's story": the interim report is out, the reaction is complete, the expectation gap sits below mid-range, and operating cash flow is −45.14% YoY. Watch only.

This is the item QC overturned most thoroughly, and also the item carrying the most information in this issue.

• ⚠️ Correction: the draft wrote "interim report not obtained, left blank," interpreted "20-day −10.53%, dist. 60d high −39.95%" as "a post-ebb low position, not yet started," and on that basis listed it in the No. 2 recommended slot, calling it "the most elastic." In fact the interim report was already disclosed on the evening of 8/21: H1 revenue RMB 41.778 billion (+182.49%), attributable net profit RMB 13.651 billion (+241.70%), ex-non-recurring RMB 13.092 billion (+229.32%), weighted ROE 37.62%, operating cash flow RMB 1.800 billion (−44.08% YoY).
• The price reaction after disclosure (the single most important piece of evidence today): 8/21 (disclosure day, not yet public during the session) +4.29% → 8/24 −7.72% → 8/25 −2.78%, −10.3% over two days; on 8/26 it was up only +0.58%, with the main force still a net outflow of RMB 79 million.
• Why this evidence is key: when a supply chain gets sold down 10% even in the face of an interim report with attributable net profit +241.70%, its "low position" is not "nobody has looked at the results" but "people looked and sold." The recommendation rationale "lowest position therefore greatest elasticity" is directly falsified by its own price behavior five days ago.
• A second corroboration in the same chain: 华工科技 HGTech's 8/26 evening interim report showed attributable net profit +30.17% but ex-non-recurring only +2.42% and operating cash flow −RMB 1.198 billion (see §5⑦).
• Technical sentiment: PE 48.66, PB 24.97, total market cap RMB 995.3 billion; 8/26 turnover rate 1.97%, turnover value RMB 18.604 billion (second on the list).
• Final judgment: it is true that overseas interconnect names led after hours and that this A-share name is indeed the lowest positioned, but the combination "low position + excellent results + still sold" does not support putting it in a recommended slot. If it gaps up today, it is more likely a mapping-driven gap-up — whether anyone actually takes the other side has to be read from post-open fund flows, not gambled on with pre-market reasoning. Watch only.

• ⚠️ Correction: the draft cited only attributable net profit of RMB 1.186 billion (+30.17%) and wrote that "attributable net profit +30.17% shows profit growth came from mix and gross margin, not scale" — that attribution is refuted by first-hand data: ex-non-recurring was only RMB 747 million (+2.42%), non-recurring gains were RMB 439 million, 37% of attributable net profit, and operating cash flow was −RMB 1.198 billion, −148.78% YoY. Profit growth came mainly from non-recurring gains, not from mix and gross margin.
• A consistency-of-criteria problem (must be booked against myself): in §5⑨ this issue praises 亨通光电 Hengtong Optic-Electric with "ex-non-recurring > attributable, not reliant on subsidies," and downgrades 亨通 Hengtong on "operating cash flow −RMB 865 million." 华工 HGTech passes neither of those same two tests — its ex-non-recurring is only 63% of attributable net profit, and its cash flow of −RMB 1.198 billion is 38% worse than 亨通 Hengtong's. Yet the draft downgraded 亨通 Hengtong and put 华工 HGTech into the No. 5 recommended slot. This is a criterion being skipped on the name I preferred.
• The half that still holds: the filing-grade statement "800G and 1.6T optical module series have achieved global volume delivery" is real, and it is the only filing-grade evidence of 1.6T delivery from an A-share company in this issue, not an interactive-platform claim. But revenue grew only +2.53% YoY — most of the company's scale is still non-compute business such as laser equipment, and "1.6T volume delivery" cannot currently be seen in total revenue.
• Technical sentiment: closed 8/26 at RMB 103.80, −1.40%, net main-force outflow RMB 159 million, turnover rate 3.09%; 20-day +3.53%, 60-day −30.34%, −44.69% below its 60-day high, the deepest drawdown on this list.
• Final judgment: "filing-grade delivery evidence" and "near-zero ex-non-recurring growth + heavy cash bleed" hold simultaneously. The hardness of the catalyst far exceeds the hardness of the results. Watch only.

Added after QC; its purpose is to provide a counterexample for the memory branch.

• Related news: interim report on the evening of 8/25: revenue RMB 8.278 billion (+6.96%), attributable net profit RMB 543 million (+20.28%), ex-non-recurring RMB 504 million (+58.80%), operating cash flow RMB 118 million (−91.90% YoY).
• Its information value lies in that +6.96%: as the main domestic memory packaging/testing and module entity, in a global memory price-up cycle its revenue grew only +6.96% YoY. This is hard evidence that "the global memory upcycle has not yet been fully transmitted to this A-share segment," and it is one of the quantitative bases on which this issue keeps memory-branch names out of the recommended slots.
• Technical sentiment: 8/26 −1.60%, turnover rate 3.55%, PE 46.78 / PB 4.31; 20-day −9.76%, dist. 60d high −45.27%.
• Final judgment: ex-non-recurring +58.80% shows operations are improving, but neither the revenue growth rate nor the cash flow (−91.90%) supports it as an elasticity name. Watch only; its main role in this issue is to provide data support for the judgment that "the memory logic is hardest but this A-share segment is weakest."

• Related news: LME three-month copper closed 8/26 +$76 at $14,350/tonne (approx. +0.53%); the structural backdrop is the LME spot squeeze — inventory has nearly halved since mid-May and the spot premium is the steepest in five years.
• ⚠️ The read moments must be written separately: the line above is on the LME 8/26 close basis; the other is COMEX copper at $6.6135/lb, +0.21%, read at 08-27 07:19. Two exchanges, two read moments; they cannot be jointly called "how much it rose overnight." The two are mutually consistent in magnitude (LME $14,350/tonne ≈ $6.51/lb, with COMEX at a 1.6% premium).
• Industry branch stage: A-shares ran ahead of the metal. On 8/26 A-share copper names already completed a round of concentrated pricing — 江西铜业 Jiangxi Copper limit-up +10.01%, 金诚信 JCHX Mining limit-up, 铜陵有色 Tongling Nonferrous +4.98%, 洛阳钼业 CMOC +6.35%, 白银有色 Baiyin Nonferrous limit-up.
• Fundamentals: PE 14.51 / PB 1.92, the lowest tier on the whole list; on 8/26 limit-up seal orders RMB 207 million, turnover rate 4.21%, turnover value RMB 4.150 billion, dist. 60d high −11.00%.
• Final judgment: the squeeze logic is real and ongoing, but today copper names need the metal to keep rising in order to "ratify" yesterday's gains, not the other way round. This is a follow-through position, not a starting position. Watch only, do not chase a gap-up.

fundamentals-analyst verified item by item against the original filings and the interim report (308-page PDF). It overturned one of my suspicions from the search stage and also overturned the impression given by media headlines; both directions must be written.

• My suspicion was falsified (to be booked): I had suspected that the "RMB 1.831 billion of bids won on 8/26" was the old RMB 1.868 billion filing of 2025-10-27 stitched into news. The suspicion does not hold. There is indeed a brand-new filing — "Announcement on Winning Bids for Marine Energy Projects," No. 2026-074, uploaded 2026-08-26 18:57 with a filing date of 2026-08-27, total winning amount RMB 1.831 billion, covering Datang Hainan Danzhou 1.2 million kW offshore wind, Shenergy Hainan CZ2, Changle Outer Sea, Three Gorges Yangjiang Qingzhou V/VII, as well as projects in Europe, the Americas and Asia-Pacific. It and the RMB 1.868 billion filing of 2025-10-28 are two independent filings.
• But the size and basis of the orders must both be stated clearly: ① the company itself says it equals 2.74% of 2025 audited revenue; ② the filing explicitly states "the above winning projects have not all had contracts signed… there is a degree of uncertainty" — it is a mixed basis of bid-winning notices and partially signed contracts, not locked-in orders; ③ the wording is "recently cumulative bids won and contracts signed of about RMB 1.831 billion," not new business on 8/26 alone; ④ the company issues 3–4 such filings a year. In sum: real, but a routine periodic disclosure at a normal cadence, not an unusual event.
• The impression given by media headlines was falsified (more important): "net profit surged 93%" sounds like a strong positive, but compared against the company's own 2026-07-14 pre-announcement — attributable net profit guided to RMB 3.0157–3.5684 billion (+86.94%~+121.2%) and ex-non-recurring to RMB 3.0496–3.6023 billion — actual attributable net profit of RMB 3.1196 billion lands at the 18.8% percentile and ex-non-recurring of RMB 3.1515 billion at the 18.4% percentile, both bases landing at the bottom edge in tight agreement, 5.24% below the midpoint of the range. The pre-announcement was issued on the 13th day after H1 ended, so even with nearly complete information it still only delivered the bottom edge. Corroboration: after the pre-announcement the stock rebounded +6.24% on 7/14, then resumed falling, closing at 58.39 on 7/17 and bottoming at 47.20 on 7/30 — the market did not price that pre-announcement as a positive back in July.
• The hardest negative fact: net cash flow from operating activities of −RMB 864.8 million (2025H1 was +RMB 4.9 million), profit-to-cash conversion −28%. Accounts receivable +RMB 4.380 billion (+25.2%), contract assets +RMB 874 million (+53.0%), inventory +RMB 1.420 billion, while accounts payable −RMB 275 million and notes payable −RMB 525 million (upstream float fell rather than rose), for an increase in net working capital absorption of about RMB 5.9 billion. Accounts receivable reached 51.8% of H1 revenue; receivables + notes + contract assets total RMB 26.642 billion = 79% of net assets attributable to parent. Short-term borrowings +40.09% in half a year.
• The positive item on profit quality (which must be given): ex-non-recurring (RMB 3.1515 billion) exceeds attributable net profit (RMB 3.1196 billion), with non-recurring items totaling −RMB 31.86 million — profits do not rely on subsidies and are in fact dragged down by asset disposal losses. Gross margin 16.77% (+3.18pct), ROE 9.43%. But the company itself attributes the gross margin improvement to "growth in optical communications market demand and rising prices" — that is price-driven and reversible, not structural operating leverage (the expense ratio went from 7.87% to 7.95%, essentially unchanged).
• The "AI compute" label must be stripped off: in the interim report, both the revenue-and-cost breakdown by industry/product and the segment financial information are ticked "not applicable", and the only quantitative statement about any segment in the entire text is "optical communications business revenue grew more than 130% YoY, with gross margin above 60%." The company does have products such as OM4/OM5 multimode fiber, high-density MPO, 4-core MCF patch cords, the full 10G–800G AOC range (already in volume application), and a CPO polarization-maintaining fiber technology breakthrough, but it disclosed no revenue or share figure for any of these products — not a single number; the 1.6T core optoelectronic conversion device is a "proposed" project, and the AI optical interconnect / thermal management / power tracks are to be "pursued through a controlling acquisition or strategic investment" — all are planning-stage wording. On the 2025 annual report basis, the main body is still smart grid (37%) + copper conductors (30%, gross margin only 1%). Equating "optical fiber and cable" directly with "compute" is an order-of-magnitude error.
• Valuation basis correction: the previously cited "PE 25.63" was the dynamic P/E (interim report annualized); the true PE (TTM) is 38.19, static 59.66. PB 4.74 sits at the 94.9th percentile of 8.6 years of history (historical median 1.76, current level 2.7x that).
• Position (only half of which was told previously): +161%~+165% year-to-date and −47.3% from the 2026-06-25 peak of RMB 123.00 both hold at once. This is a stock that has already halved and is now on the 20th trading day of a rebound after a sharp fall.
• Technical sentiment: closed 8/26 at RMB 64.83 (+0.36%), net main-force outflow RMB 422 million, turnover rate 8.60%, turnover value RMB 13.678 billion (extremely high against a RMB 159.9 billion market cap). Main-force flows have been highly unstable over the surrounding days: 8/19 −RMB 2.275 billion, 8/24 −RMB 800 million, 8/25 +RMB 1.841 billion, 8/26 −RMB 422 million. The controlling shareholder has pledged 62.1% of its holding.
• Final judgment: the orders are real and the profit growth rate is real, but the stack of four — landing at the bottom edge of its own pre-announcement + operating cash flow swinging from positive to −RMB 865 million + missing segment data making the AI share unverifiable + PB at the 94.9th percentile of 8 years — does not support placing it near the top of the watch list. Yesterday it was listed in a secondary recommended slot (+0.36% that day, not delivered). Today it is downgraded to watch only.