A-Share Pre-Market Brief | 2026-07-30, Thursday

• Related news: company filing on the evening of 2026-07-29. Its controlled subsidiary Suzhou Dingxin Photoelectric Technology Co., Ltd. received purchase orders for high-power laser chips — Customer A RMB 542 million (VAT-inclusive), Customer B USD 87 million (about RMB 591 million, VAT-inclusive), totalling about RMB 1.133 billion. Delivery deadlines: Customer A by 2027-12-10, Customer B by 2028-01-03. The company's wording: "if the orders are implemented smoothly, they are expected to have a positive impact on the company's operating results for 2026 and 2027." Sources: Securities Times¹, Sina Finance².
• Positive logic: a direct order positive that affects revenue, but it must be looked through to the parent basis. The company's 2025 revenue was RMB 5.287 billion (+28.60%) and net profit attributable to parent RMB 234 million (+280.43%) (sources: C114²³, Sina²⁴, cross-verified by the fundamentals check via the East Money datacenter interface). Ratio calculations (computed by this report, presented in tiers):

The single most telling comparison: the annual report discloses that the third assessment-period target of Suzhou Dingxin's equity incentive plan is "Suzhou Dingxin's 2026 revenue of no less than RMB 100.00 million (RMB 100 million)" — the company's own 2026 revenue threshold for this subsidiary is RMB 100 million, and this order is about 11x that target. For lateral reference: 源杰科技 (Yuanjie Semiconductor) FY2025 revenue RMB 601 million with a 58.11% gross margin, 长光华芯 (Everbright Photonics) FY2025 revenue RMB 477 million with a 34.54% gross margin — this order is equal to about 1.9x the full-year revenue of China's leading listed CW laser chip companies. Gross margins in the field are very high (34%–58% at comparable companies), so if it can be delivered, the profit elasticity is far greater than the revenue elasticity; but by the same token the capacity barrier is also very high.

• Stage of the industry branch: starting (and against the sector). The branch itself is in a global de-rating phase (overnight SOX −5.33% measured close, media basis −3.56%, discrepancy to be verified), but this stock's positive is company-specific and unrelated to sector beta. There is currently no limit-up ladder and no lead name — this branch has not yet formed a ladder in the A-share market.
• Fundamentals verification (completed by fundamentals-analyst using the East Money datacenter interface + original Juchao filings; all figures below are post-verification):

• ⚠️ Earnings quality: this is the most important negative evidence in this case, and it is more precise than the first draft stated. The 2025 income statement shows investment income of RMB 286.8 million, of which RMB 285.4 million came from associates/joint ventures, while operating profit for the year was only RMB 223 million — i.e. stripping out equity-method investment income, 2025 core-business operating profit was about −RMB 64 million (negative). The source is stated in the annual report: a subsidiary of Shanghai Dongchang Investment, a 50%-held associate, transferred 82.34% of Shanghai Dongjiao Real Estate for RMB 511.26 million. This is a one-off property disposal gain, yet under the equity method it is counted within "ex-non-recurring" — so the "ex-non-recurring +398.61%" in 2025 cannot be read as a core-business figure. On the same logic, 2025Q4 net profit attributable to parent was −RMB 95.5 million.
• ✅ But 2026Q1 is a genuine inflection (an important upgrade to the first draft): 2026Q1 investment income was only RMB 15.4 million (RMB 304.7 million a year earlier), so the RMB 158.8 million of net profit attributable to parent came essentially from the core business, with gross margin going from 14.65% to 26.20%. And on 2026-07-06 the company issued H1 guidance: net profit attributable to parent of RMB 500–700 million (+57%~120%), ex-non-recurring of RMB 490–690 million (+55%~119%), attributed to "the optical communication segment benefiting from the explosion in AI compute demand, with both volumes and prices rising in the optical fibre market", and explicitly stating that no associate-disposal-type gains occurred in the period. → The main engine of 2026 earnings is optical fibre/optical communication price increases, not laser chips. This matters enormously for how the order is positioned: the order is "the start of a new story", not "the main driver of this year's earnings".
• Dingxin Photoelectric's basis (the first draft wrote "no reliable data available"; now established): it is a controlled, consolidated subsidiary; registered capital is only RMB 23.633875 million; the ownership percentage is disclosed inconsistently — the annual report shows direct 16.0921% + indirect 28.7646% = 44.8567%, and explicitly records that this was "diluted from 52.2656% to 44.8567%", whereas the 2025-12-22 capital-increase announcement used 52.4914% (without looking through the 75% holding chain of Wuhan Yongding Optoelectronics Group). The annual report's parent-share figure of about 44.86% is authoritative, and the trend is continued dilution. Dingxin's 2025 standalone revenue/net profit was not separately disclosed in the annual report (it did not reach the 10% materiality threshold; no reliable data available); historical anchor: audited net assets of only RMB 64.0559 million as of 2023-12-31. There is also an implicit repurchase obligation: the December 2024 capital increase brought in investors with repurchase rights, and the company has recognised RMB 50.4 million of equity repurchase payable.
• Actual capacity/customers/certifications/patents: the order itself is filed and hard, but the supporting conditions are generally unverifiable. The filing itself says the company "is advancing capacity expansion in an orderly manner" — i.e. capacity is not yet in place. Annual-report-level evidence (company's own statements): it has "a rare domestic IDM laser FAB", and its 70mW CW-DFB and 100G EML series have completed development and reliability validation, passed customer optical-module adaptation testing, and "are now being supplied", with an optical chip R&D team of 80 people. The following are all subject to no reliable data available: laser chip capacity figures (not disclosed in the annual report; the 13 million / 15 million / 600 million CPO module figures circulating in the market conflict with each other and none comes from a filing), customer certifications (Customers A and B have obtained confidentiality exemptions and their identities are not disclosed; self-media claims that it "has passed AWS and Coherent certification" appear nowhere in the filings or the annual report), optical chip patent counts, market share and industry ranking.
• Industry conditions are cross-verifiable: the annual report and third-party reporting agree — "global optics leader Lumentum's 2027 capacity is already fully locked up by NVIDIA and cloud providers, and its 2028 capacity is expected to be fully booked in the second half of 2026 (this statement is quoted from the annual report and third-party reporting; this report did not obtain a first-hand source link, to be verified)". This provides industry background for "such a large order plausibly existing", but it is not independent evidence that Dingxin won these orders.
• ✅ One important piece of positive evidence (not found in the first draft): a full cninfo search from 2020-01-01 to 2026-07-30 finds only 1 company announcement with "order" in the title — this one; and 0 with "tender win". In other words, the company is not a serial order-announcer; this is its first order announcement in more than six years, which materially raises the information content of this filing.
• Financial risk (updated to the latest disclosures, replacing the 2025-10 media data cited in the first draft): interest-bearing debt of RMB 3.980 billion (short-term borrowings RMB 2.579 billion + non-current liabilities due within one year RMB 441 million + long-term borrowings RMB 960 million) against cash of RMB 1.805 billion → net interest-bearing debt of RMB 2.175 billion, or 66% of the RMB 3.318 billion of net assets attributable to parent; current ratio 1.00, quick ratio 0.74; finance costs already reached RMB 45.3 million in 2026Q1 (annualising to about RMB 181 million, a marked increase). The controlling shareholder and its subsidiaries have provided guarantees to the company with an outstanding balance of RMB 3.303 billion, equal to 103.95% of audited net assets at end-2025 (company filing basis) — directionally this is shareholder credit support, but it also shows the company's own credit is not sufficient to support nearly RMB 4 billion of interest-bearing debt on its own. Accounts receivable are RMB 1.410 billion with 103.7 days of turnover; the top five customers account for 48.88% of annual sales; deductible losses under deferred tax are RMB 596 million. Net goodwill is only RMB 95.0 million (cumulative impairment of RMB 246.3 million, with the goodwill on Digital Communication and Jinting Wiring Harness 100% written off). Historical blow-up: net profit attributable to parent of −RMB 571 million in 2020; the associate Yongding Optical Communication has been confirmed as insolvent with its equity having no auction value, and enforcement proceedings were terminated in April 2026 (a full loss is highly likely); on 2024-07-04 it received a letter of enquiry on its 2023 annual report.
• ⚠️ Reverse signals at the governance level (this report regards these as the group most in need of caution): the controlling shareholder Yongding Group holds 24.87%, of which 216 million shares are pledged (59.4% of its holding); it sold 18.97 million shares during 2025; it disclosed a disposal plan on 2026-03-03 and on 2026-06-02 disclosed an "equity change crossing the 1% mark and results of the disposal" — i.e. it sold during the main up-leg; it published 8 "abnormal share price movement announcements" within one year (2025-09-24, 10-14, 11-27, 12-13, 12-27, and 2026-04-11, 06-05, 06-18), 3 of which attached replies to exchange enquiry letters; this order announcement was released on the same day (2026-07-30) as a "release of pledge and re-pledge" announcement.
• Industry position: a high-elasticity name (not the leader, not the lead stock). The largest domestic share in high-power laser chips belongs to 长光华芯 (Everbright Photonics, 688048, STAR Market) (media basis). The logic is "a small subsidiary winning a large order", with elasticity coming from a small denominator rather than a high position.
• Technical sentiment (the first draft wrote "no reliable data available"; now established, and the conclusion is the opposite of the first draft's): close RMB 30.71 (7/29), share capital 1.46198 billion shares, total market cap RMB 44.90 billion, PB 13.5x, PE (2025 static) 192x, PE(TTM) roughly 437x.

There were 4 limit-downs or near-limit-downs during July 2026: 7/2 −10.01%, 7/17 −9.99%, 7/20 −10.00%, 7/28 −10.00%. Daily turnover has persistently run 6%–12%, and 7/29 turnover was RMB 3.908 billion. "Is it already at a high" must be answered on two scales: relative to the start of 2025 it is absolutely at a high (up 15x, PB 13.5x); but relative to the sentiment peak at the end of June it has been halved (−58.5%) — the order announcement lands inside a sharp downward leg, not on a chase.

• Final judgment: priority deep-dive (the highest research priority), but the rank drops from No. 1 to No. 6, and "the news is very hard" must be stated separately from "the stock is very dangerous".
  • Four grounds on which it is "hard": the ratio is large enough (21.4% VAT-inclusive, ≈ one full year of optical communication segment revenue); it is the first order announcement in more than six years and specifies performance deadlines and phased delivery; the industry background is cross-verifiable (global high-power CW/EML chips really are tight); and the core business has independently verified an inflection (2026Q1 gross margin 26.20%, H1 ex-non-recurring guidance of RMB 490–690 million with last year's one-off property gain already excluded).
  • Four grounds on which it must be discounted: the parent only takes about 44.86%; capacity is not in place (the company's own 2026 revenue target for Dingxin is only RMB 100 million, and the order is 11x that); the performance period spans about 17 months, so the amount recognised within 2026 is very likely the smaller share, and the filing gives no amount guidance; and verifiability is weak — customer identities are undisclosed, there is no disclosure of prepayments, penalties or minimum purchase volumes, and the phrase "effective upon receipt and confirmation of the customer's purchase order" is close to circular.
  • And what finally holds it down is the pricing, not the announcement: a RMB 44.9 billion market cap, PB 13.5x and a TTM PE of roughly 437x already embedded expectations far beyond this order before it was announced. In other words, this order looks more like "partial evidence of delivery" against already-high expectations than a new, unpriced increment. Combined with a set of reverse governance signals (selling during the main up-leg, 59.4% pledged, 8 abnormal-movement announcements, the order announcement and the release-and-re-pledge on the same day), "priority deep-dive" here refers only to research priority and explicitly does not imply any certainty at the trading level.

• Related news: interim report/earnings announcement on the evening of 2026-07-29: H1 revenue RMB 39.443 billion, +25% year on year, and net profit attributable to parent RMB 4.169 billion, +123% year on year (source: East Money¹⁴).
• Positive logic: direct earnings, already landed, with no need to verify deliverability. This is the exact opposite type to Yongding: Yongding is "hard news, soft company", Western Mining is "hard company, small expectation gap".
• Stage of the industry branch: a medium-term uptrend, not a theme launch. LME copper is about USD 13,528/tonne and Shanghai copper about RMB 103550/tonne, in a historically high range (data as of the 2026-07-20 weekly report, 10 days ago; the latest copper price is unverified, to be verified. Source: BOCI non-ferrous weekly²⁵). There is no limit-up ladder — non-ferrous is an allocation-type branch, not a limit-up-type branch.
• Fundamentals verification: 2025 revenue RMB 61.687 billion (+23%), net profit attributable to parent RMB 3.643 billion (+24%); 2026Q1 net profit attributable to parent RMB 1.586 billion (+96.34%) (source: Securities Times²⁶). From this it follows: H1 RMB 4.169 billion − Q1 RMB 1.586 billion = about RMB 2.583 billion in Q2, roughly 63% growth quarter on quarter, and a single half already exceeds full-year 2025 — growth is accelerating, and that is a hard fact. Copper resource reserves are 8.4346 million tonnes, with all mines located domestically (avoiding cross-border geopolitical risk). Gross margin, operating cash flow, gearing, ROE and PE/PB are all subject to no reliable data available because the interface is down.
• Industry position: a niche leader/lead stock (one of China's integrated copper-lead-zinc miners).
• Technical sentiment: no reliable data available. All that is known is that non-ferrous metals had RMB 973 million of main-force net inflow on 7/29, and 国城矿业 (Guocheng Mining, 000688) +4.33% and 海亮股份 (Hailiang, 002203) +5.78% (7/29 measured) — money has already entered the branch, which means the expectation gap is being consumed.
• Final judgment: priority deep-dive, but positioned as "certainty" rather than "elasticity". It has the hardest earnings and the shortest evidence chain on the whole list. Its problem is not whether it will deliver, but whether the market already knows — the copper price is high, Q1 was already +96%, and RMB 4.169 billion is very likely already in sell-side models. If it opens sharply higher today, the expectation gap is consumed and its risk/reward deteriorates quickly.

• Related news: ① Samsung Electro-Mechanics announced a uniform 30% MLCC price increase from 1 August, and Murata also issued a price letter (source: Sina³); ② on 7/29 the passive component industry had RMB 2.594 billion of main-force net inflow, the largest in the whole market, with a net share of +8.32% (7/29 measured).
• Positive logic: a price positive that directly affects ASP and gross margin. Supply rigidity is verifiable: Murata/Samsung Electro-Mechanics high-end line utilisation is 85–90%, and new capacity (the Izumo plant/the Philippines) will not arrive until end-2026 to 2027; Murata plans to invest JPY 40 billion in expansion this year, with phase-two capacity expected in 2029 (industry media basis, to be verified). The demand side is verifiable: a GB200/Rubin architecture AI training server uses about 20,000 MLCCs per unit, 8–10x a traditional server (industry media basis, not cross-verified, to be verified).
• Stage of the industry branch: fermenting (days 1–2), with a lead stock and sealed boards already. Fenghua Advanced hit limit-up on 7/29 with RMB 415 million of sealing orders, the thickest in the whole market, turnover of 14.92% and main-force net inflow of RMB 1.838 billion (second among single stocks); 三环集团 (CCTC, 300408) +7.99% with RMB 801 million of main-force net inflow; 江海股份 (Jianghai, 002484) and 国瓷材料 (Sinocera, 300285) followed at +4.79% (all 7/29 measured). The ladder has formed and the lead stock is clear.
• Fundamentals verification: capacity of about 30 billion units, with a plan to expand to 60 billion units (media basis, not cross-verified against company filings). Revenue, net profit, gross margin, operating cash flow, gearing and PE/PB are all subject to no reliable data available because the interface is down. The company itself has not issued any price-increase announcement or letter — this round of increases is an action by Japanese and Korean original manufacturers, so at the company level this is an "industry trend positive", not a "direct order positive".
• Industry position: one of China's MLCC leaders / the branch's lead stock (judging by the 7/29 sealing orders and the tier of money involved, the market already treats it as the lead).
• Technical sentiment: already at limit-up, turnover 14.92%, thickest sealing orders. Whether it is a first board or how many consecutive boards, and its 20-day performance, are subject to no reliable data available. The chase risk therefore cannot be quantified, and this report applies its "blanket downweight for anything already at limit-up", deducting −9.
• Final judgment: watch closely — neither dismissed because it hit limit-up nor chased because its sealing orders are the thickest. This is the clearest "date anchor" on the whole list (effective 8/1). But it is also the first verification point for today designated by yesterday's recap, and the overnight answer leans negative: SOX fell another 5.33% (measured close) and SK Hynix's Q2 results missed with the shares down more than 19% intraday, meaning the contradiction of "money selling AI-hardware downstream while buying AI-hardware upstream" is very likely to play out a second time today. Microsoft's Azure +43% (demand is real) and Meta's −9.64% after hours (valuations are going to be cut) both holding at once is the macro version of that same contradiction. See Section 8 for the decision criteria.

• Related news: ① 197 game licences approved in July (193 domestic + 4 imported), with the company's "Original Legend 2" clearing review on 7/23 (sources: Cailianshe²⁷, GameRes²⁸); ② Caixin Securities research notes that Century Huatong, Giant Network, Kaiying Network and G-bits are all expected to post year-on-year profit growth for H1; ③ on 7/29 the gaming sector was +4.49% with RMB 1.663 billion of main-force net inflow and a net share of +10.01%, the highest in the whole market, and the company hit limit-up with RMB 106 million of sealing orders (7/29 measured).
• Positive logic: licence → launch → gross billings → earnings; the path is complete but the cycle is long. Licences are a routine monthly event (171 in June, 197 in July) and do not in themselves constitute incremental good news — the real increment is the fund-flow fact that "the main-force net share of +10.01% was the highest in the whole market".
• Stage of the industry branch: starting (day 1). Three limit-ups, with Giant Network at RMB 140 million of sealing orders and Kaiying Network at RMB 106 million (7/29 measured). A ladder is taking shape, but there are only first boards and no consecutive limit-ups.
• Fundamentals verification: interim growth pre-announced (Caixin Securities basis; this report could not find the company's original pre-announcement, so the specific magnitude is subject to no reliable data available). Revenue, net profit, gross margin, cash flow, gearing and PE/PB are all subject to no reliable data available. The launch date and billings expectations for "Original Legend 2" are subject to no reliable data available.
• Industry position: one of the branch's lead stocks (on a par with Giant Network; judging by sealing-order size the two are comparable).
• Technical sentiment: 7/29 limit-up with RMB 106 million of sealing orders. Whether it is a first or consecutive board, turnover, and its 20-day performance are subject to no reliable data available.
• Final judgment: watch closely. Its fund-flow evidence is the strongest in the whole market (the highest net share at +10.01%, real money rather than a low-volume mark-up), but the fundamentals evidence is almost entirely absent, and the nature of this branch is "the mirror image of technology bleeding out" — the fundamental reason it rose on 7/29 is that money came out of AI hardware, not that anything new happened in the gaming industry. Another SOX decline overnight is actually favourable for it in the short run; but that also means its strength is not independent. See Section 8 for the decision criteria.

• Related news: a filing on the evening of 2026-07-29 stating plans to invest RMB 1 billion to build a robot joint reducer production base with annual output of 5 million units (source: East Money¹⁴). Supporting evidence: the Ministry of Commerce disclosed H1 industrial robot exports +18.6% and 3D printer exports +109.3% (source: Sina¹⁹). The stock was also named in Sina's [20260730 morning comment] (source: Sina²⁹).
• Positive logic: an industry trend positive, not an order positive, and the nature of the filing is one notch weaker than the first draft judged. After reading the original filing, the fundamentals check confirms: this is an "Investment Promotion and Cooperation Agreement" signed with the Jiangyang District People's Government of Luzhou — the government is responsible for assisting with project approval, environmental and safety assessments, construction permits, utility supply and implementation of business-support policies; the company is responsible for compliant operation. The agreement stipulates no investment-progress obligation for the company, no construction-start or production-start milestones, no capacity performance undertakings or penalty clauses, and discloses no specific land, tax or subsidy concessions. The filing also states explicitly: "the investment amount involved in this project is a planned investment scale, which may subsequently be adjusted in light of customer orders, fundraising and other specific circumstances, and there is a risk that the actual investment amount differs from the investment plan." Governance level: approved by the 27th (extraordinary) meeting of the sixth board of directors on 2026-07-29, with no shareholder-meeting approval required.
• Stage of the industry branch: starting. On 7/29 the robotics concept had 明新旭腾 (Mingxin Xuteng, 605068) at 3 consecutive limit-ups (media basis). The lead stock is unclear.
• Fundamentals verification (completed by fundamentals-analyst using the East Money datacenter interface + the original filing; the first draft's "no reliable data available" has now been fully filled in):

• The growth is real, but profit quality is deteriorating: FY2025 revenue +14% while net profit was −13.9%, i.e. negative operating leverage. Three reasons (all from the annual report): ① structural dilution — low-margin differential revenue rose from RMB 470 million to RMB 686 million, and the gross margin on "other precision components" fell from 51.4% to 35.7%; ② depreciation rigidity — depreciation rose from RMB 256.3 million to RMB 327.4 million (+27.7%); ③ FY2025 asset impairment of RMB 136.2 million, including RMB 97 million of goodwill impairment (a key audit matter).
• ⚠️ The most glaring item: cumulative free cash flow over three years plus 26Q1 is about −RMB 1.35 billion. This is a company that keeps pouring all of its operating cash and external financing into capacity — not a company with spare cash for new investment.
• ✅ Can it afford the RMB 1 billion? The answer is "not out of its own money, it must raise financing", and the filing admits it. 26Q1 cash was only RMB 253.9 million (RMB 450.7 million at end-FY2024 → RMB 279.7 million at end-FY2025 → RMB 253.9 million in 26Q1, a continuous decline). Order-of-magnitude comparison (computed by this report): RMB 1 billion = 3.9x cash = 3.2 years of operating cash flow = 26.5% of net assets attributable to parent = 12.2% of total market cap. The "funding method" box in the filing ticks own funds + bank loans, and does not tick "raised funds"; the risk warning reads: "the investment scale of this project is relatively large, and the funding sources are the company's own funds and bank loans. Implementation of the project may put some pressure on the company's cash flow and fundraising."
• ⚠️ A key piece of background easily overlooked: the company only just issued the "Hao 26 Convertible Bond" of RMB 1.8 billion in June 2026 (net proceeds RMB 1.7895 billion, received 6/22 and listed 7/10, with issuer and bond ratings of AA−). But this RMB 1.8 billion cannot be used for this project, as the use of proceeds is already locked: ① the intelligent manufacturing core components project (phase two), RMB 1.30 billion; ② working capital replenishment and repayment of bank borrowings, RMB 500 million. In other words, the company's committed/in-hand capex for 2026 is about RMB 1.3 billion + RMB 1.0 billion = RMB 2.3 billion, while FY2025 operating cash flow was only RMB 309 million; and the fact that RMB 500 million of the RMB 1.8 billion convertible bond is for "working capital and loan repayment" is itself a signal of tight liquidity rather than ample ammunition. Interest-bearing debt (26Q1): short-term borrowings RMB 580.8 million + long-term borrowings RMB 1.2537 billion = RMB 1.834 billion, plus notes payable of RMB 313.8 million and construction in progress already at RMB 840.3 million. A rough pro-forma gearing estimate (estimated by this report, not disclosed by the company): starting from 48.82% and adding the RMB 1.8 billion convertible bond, of which RMB 489 million repays loans, gearing rises to about 56.5% (convertible bonds contain an equity component, so the actual accounting treatment would be slightly lower; this is an upper-bound estimate).
• Pure arithmetic reference: RMB 1 billion of fixed assets on 10-year straight-line depreciation ≈ RMB 100 million a year of new depreciation, equal to 36% of FY2025 net profit attributable to parent of RMB 277.1 million. The filing says outright that it "will add rigid expenditures such as fixed asset depreciation and finance costs, which may dilute the company's short-term profitability".
• ✅ Core business and technology extensibility: there is real evidence, this is not pure concept (an upgrade to the first draft). FY2025 revenue mix: synchroniser systems RMB 1.261 billion (46.9%, 33.10% gross margin), differential systems RMB 686.4 million (25.5%, 13.57%), other precision components RMB 411.6 million (15.3%, 35.69%), aviation components RMB 263.1 million (9.8%, 43.53%). The annual report describes the company as "China's leading synchroniser enterprise", supplying Mercedes-Benz, BMW, Audi, Volkswagen, Hongqi, Geely, Changan, FAW, BYD, AITO, Xiaomi, NIO, Li Auto and others. The annual report already lists the "robot components business" as a standalone segment, with the wording: "serialised products such as high-precision planetary reducers and high-precision cycloidal reducers have been developed"; its industry codes already include "C3453 manufacture of gears and gear reducers and gearboxes"; and within the automotive business it is already mass-producing planetary reducers (the annual report lists "stators and rotors, motor shafts and planetary reducers" as key development items). The process chain overlaps heavily: precision forging, high-precision machining, heat treatment, laser welding and assembly — exactly the core process set for gear-type reducers. R&D expense was RMB 155.1 million in FY2025 (5.77% of revenue) and 7.02% of revenue in 26Q1 (a clear step-up).
• ⚠️ But two important qualifications (where this report thinks misreading is most likely): ① the technology route is "planetary + cycloidal", not harmonic/screw — the reviewer full-text searched the FY2025 annual report (about 195,000 characters) and found "harmonic" 0 times, "screw" 0 times, "humanoid" 0 times; the filing likewise never uses the phrase "humanoid robot". It should therefore not be equated with a beneficiary of the mainstream humanoid-robot joint solution (harmonic reducers/planetary roller screws); it is closer to industrial/general robot joints and precision transmission. ② There has been no prior robot order/contract/tender-win/customer-nomination announcement of any kind — the reviewer went through the company's most recent 300 filings (back to 2024-10-29) and found not one.
• The company's own characterisation (filing text, the strongest basis for the downweight): "As of now, the company's robot joint reducer products are at an early business stage, have not yet had a substantive impact on the company's operating results, and there remains considerable uncertainty in subsequent market development, customer development and order landing, with material uncertainty in future expected benefits." The FY2025 annual report's revenue mix does not separately list robot/reducer revenue (if any exists, it should be inside the RMB 411.6 million of "other precision components"; the specific amount is subject to no reliable data available). Customer lists, order backlog and product prices — none are disclosed; no reliable data available, and no speculation is offered.
• The ramp-up timing and phasing of "5 million units a year": the filing says nothing at all. The reviewer read the full text word by word, and the only two capacity-related statements are "the total investment in this project is RMB 1 billion, and once built it can form annual capacity of 5 million robot joint reducers" and, in the risk warnings, "the construction period of this project is relatively long and it is difficult for it to make a material contribution to the company's operating results in the short term". There is no ramp-up timetable, no phasing plan, no capacity ramp cadence and no benefit estimate (revenue/profit/IRR/payback).
• ⚠️ The order-of-magnitude contradiction (the basis for this report's active downweight; it still stands): China's robot reducer market was about RMB 3.5 billion in 2025 and is forecast at about RMB 4.7 billion in 2026 (Zhiyan Consulting basis; the publication year of the forecast could not be confirmed, to be verified) (sources: Zhiyan Consulting³⁰, ifeng³¹). The output value corresponding to "5 million units a year" very likely already approaches or exceeds the size of the entire national market, and must be discounted heavily.
• Industry position: a leader in automotive synchronisers; a new entrant/high-elasticity name in robot reducers. The domestic reducer share leaders are Leaderdrive, Laifual, Shuanghuan Driveline and Zhongdali De (media basis), and Haoneng is not among them.
• ✅ Technical sentiment (the first draft wrote "no reliable data available"; now established, and it is this report's most valuable single finding): close RMB 8.91 (7/29), share capital 920.3 million shares, total market cap RMB 8.200 billion, PE(TTM) 29.6x, PE (ex-non-recurring) 32.0x, PB 2.17x (computed by this report).

In other words: it made a fresh one-year low 3 trading days (7/24) before the announcement, and has been essentially flat over 20 days — it has not been chased up at all. One trading detail worth recording objectively: on 7/29 the share price went 8.60 → 8.91 (+3.60%) on volume of 16.1 million shares versus an average of about 13.6 million over the prior 20 days (about 1.18x), with a turnover ratio of 1.75% — a mild volume increase, not an anomaly. This report makes no judgment about information transmission; it merely records the facts.

• PEG is not applicable: FY2025 net profit was −13.9% and 26Q1 −0.45%, so the denominator is negative or near zero. Any PEG computed from "future robot business growth" is treating an assumption as a valuation, and this report does not use it.
• Final judgment: watch closely; the total score falls from 59 in the first draft to 52 (because the nature of the filing weakens further from an "investment announcement" to an "investment-promotion framework agreement", and the company itself says there is no substantive impact yet), but a lower ranking does not mean a worse risk/reward structure — quite the opposite, it is the cleanest risk/reward structure on the whole list.
  • Weak news, good position: the hardness of the positive only reaches "framework agreement", but it lands on a stock at PB 2.17x, PE(TTM) 29.6x, −37.25% year to date, −48.5% from the high, with a fresh one-year low 3 days before the announcement. It forms a perfect contrast with Yongding: Yongding is "the hardest news + the most dangerous position (PB 13.5x, +240% over 250 days)", Haoneng is "the softest news + the cleanest position".
  • The real bottleneck is cash and depreciation, not technology: the technology extension has real evidence (synchroniser leader, planetary reducers already in mass production, a standalone segment with serialised products in the annual report, R&D expense ratio up to 7.02%), but RMB 253.9 million of cash on the books, three-year FCF of −RMB 1.1 billion, interest-bearing debt of RMB 1.834 billion and another RMB 1.3 billion of convertible-bond-funded projects under construction mean a new RMB 1 billion project must extend leverage and front-load depreciation.
  • It should not be understood as a "humanoid robot" play (the annual report mentions "humanoid/harmonic/screw" 0 times each). This report therefore still does not place it in priority deep-dive, but does place it among the five key names in 9.1, on the grounds of "expectation gap + position" rather than news strength.

• Related news: on 7/29 the gaming sector had RMB 1.663 billion of main-force net inflow with a net share of +10.01%, the highest in the whole market; the company was +9.99% at limit-up, with RMB 140 million of sealing orders and RMB 612 million of main-force net inflow (all 7/29 measured). It is on the interim pre-announcement list (Caixin Securities basis). No new product approval was found for the company in the July licence batch; existing titles include "Supernatural Action Team", launched on 7/2, and a "Candle in the Tomb" IP collaboration map, "Yunnan Insect Valley" (media basis).
• Positive logic: indirect (sector flows + interim results expectations), with no new company-level filing. This is the key difference from Kaiying Network: Kaiying has a specific product that cleared review on 7/23, whereas Giant Network has no corresponding new company-level news this period.
• Stage of the industry branch: starting (day 1), and its RMB 140 million of sealing orders is the thickest in the branch, making it a lead stock of the branch.
• Fundamentals verification: interim growth pre-announced (research basis; the company's original pre-announcement was not found, and the magnitude is subject to no reliable data available). All other financial metrics are subject to no reliable data available.
• Industry position: a lead stock of the branch.
• Technical sentiment: 7/29 limit-up with RMB 140 million of sealing orders. Volatility note: the stock fell more than 7% in a single day on 7/23 (source: Sina³²), showing the branch has been choppy recently.
• Final judgment: watch closely. Same score as Kaiying Network, but its evidence structure is one notch weaker (no company-level news), resting only on flows and the sector. If one of the two must be chosen as the branch's representative, Kaiying Network's evidence chain is more complete.

• Related news: a filing on the evening of 2026-07-29 stating that a subsidiary plans to build a cable materials plant with annual capacity of 100,000 tonnes (investment RMB 450 million) and a Qingdao submarine-cable insulation materials project (investment RMB 370 million), RMB 820 million in total (source: Sina Stock Navigator²⁰).
• Positive logic: an industry trend positive. Again "an investment announcement ≠ an order", but the submarine-cable insulation materials direction itself has scarcity value (high-voltage submarine cable insulation compounds have long been dominated by foreign firms), and if localisation lands the barriers are relatively high.
• Stage of the industry branch: not started. There was no related limit-up ladder and no lead stock on 7/29.
• Fundamentals verification: all financial metrics are subject to no reliable data available. The scale of the RMB 820 million investment relative to its net assets and cash flow cannot be assessed. The certification cycle for submarine-cable insulation materials and whether it already has customers are subject to no reliable data available.
• Industry position: cannot be confirmed for now (its position in cable materials has not been verified).
• Technical sentiment: no reliable data available.
• Final judgment: watch closely; it is the "good direction, little verification" type. Localisation of submarine-cable insulation materials is a real proposition, but the filing only reaches the "plans to invest" stage, and revenue is at least two gates away — the construction period and the certification period. It should not be treated as an order positive.

• Related news: ① raw milk in the main producing regions (Inner Mongolia, Hebei) at RMB 3.05/kg (as of 7/23), +0.3% week on week and +0.3% year on year (year-on-year growth turning positive) (source: Securities Times¹⁵); ② on 7/29 the dairy concept was +6.18%, first among concept sectors, with 35 constituents rising and the company up nearly 3% (media basis); food and beverage had RMB 1.779 billion of main-force net inflow (7/29 measured).
• Positive logic: indirect, and the direction is questionable. Rising milk prices are positive for upstream farms (Youran Dairy H1 net profit RMB 739–903 million versus a loss of RMB 297 million a year earlier; AustAsia RMB 90–130 million versus a loss of RMB 378 million; China Modern Dairy pre-tax profit of no less than RMB 43 million versus a loss of RMB 972 million — all three are listed in Hong Kong), and represent rising raw material costs for downstream dairy companies. Yili is downstream. Its benefit logic can only be "easing industry price competition + leader share and pricing power", not "rising milk prices directly boosting profit" — a far weaker logic than the market narrative.
• Stage of the industry branch: day 2 of fermentation, with a thick ladder concentrated in small caps. On 7/29 beverages and dairy had 7 limit-ups plus 4 in snack foods; the limit-up list included 均瑶健康 (Junyao Health, 605300), 李子园 (Li Ziyuan, 605337), 一鸣食品 (Yiming Food, 605179), 阳光乳业 (Sunshine Dairy, 001318) (Shanghai/Shenzhen main boards), 欢乐家 (Huanlejia, 300997), 熊猫乳品 (Panda Dairy, 300898) (ChiNext), 骑士乳业 (Knight Dairy, 920786) (BSE) and others; Yiming Food is already at 2 consecutive limit-ups (media plus 7/29 measured). The lead stock is Yili, but the lead only rose 3% — the ladder and the lead are badly disconnected.
• Fundamentals verification: the absolute leader of the dairy industry; this is one of the most solid fundamental judgments on the whole list (an industry position that holds without any data). But specific revenue, net profit, gross margin, cash flow, gearing and PE/PB are all subject to no reliable data available because the interface is down.
• Industry position: leader/lead stock.
• Technical sentiment: 7/29 up nearly 3% (media basis). Whether volume expanded and its recent position are subject to no reliable data available.
• Final judgment: watch closely, and it is the only main-board name in the dairy branch worth watching. The reason is not that it rose the most (it rose the least), but that the strength of the branch's fundamental evidence (+0.3% year on year) is far weaker than the strength of its price reaction (sector +6.18%, 7 limit-ups), and in that combination chasing limit-up small caps carries the highest risk while buying the lead is relatively controllable. But it must be acknowledged — if this branch is only "the mirror image of technology bleeding out", the lead will bleed too, just more slowly.

• Related news: no company-level filing. It is purely a mapping of the "Haoneng's RMB 1 billion reducer base + Ministry of Commerce robot exports +18.6%" branch.
• Positive logic: an indirect supply-chain positive / branch association. Under the iron rules, this level should not enter the main recommendations.
• Stage of the industry branch: starting, lead stock unclear.
• Fundamentals verification: media reports describe it as one of China's major RV/reducer makers (alongside Leaderdrive, Laifual and Zhongdali De). All specific financial data is subject to no reliable data available.
• Industry position: one of the niche leaders (media basis, not cross-verified).
• Technical sentiment: no reliable data available.
• Final judgment: watch only. Its industry position is higher than Haoneng's, but it has no news of its own this period — buying it means buying someone else's filing. This is exactly the "distant mapping" trap summarised in yesterday's recap. It is listed to present the branch's internal ordering completely, not to recommend it.

• Related news: the MLCC/passive component price-increase branch; +4.79% on 7/29 following Fenghua Advanced higher (7/29 measured). No company-level filing.
• Positive logic: an industry trend positive, but with a category mismatch. The category Samsung Electro-Mechanics and Murata are raising prices on is MLCC (multilayer ceramic chip capacitors), whereas Jianghai's main business is mostly aluminium electrolytic capacitors. Although both are passive components, the supply/demand structure and price pass-through differ, so this mapping must be discounted.
• Stage of the industry branch: fermenting; it occupies a follower/catch-up slot (up 4.79% on 7/29, clearly weaker than Fenghua Advanced at limit-up).
• Fundamentals verification: all subject to no reliable data available.
• Industry position: catch-up/back row (within the specific MLCC logic).
• Technical sentiment: 7/29 +4.79% (measured).
• Final judgment: watch only. The category mismatch is a hard flaw. If one is to participate in passive component price increases, the lead is Fenghua Advanced; Jianghai is the classic case of "the name matches but the category does not".

This section was written by bull-bear-debater as the strongest bull case and the strongest bear case independently, with neither side accommodating the other. This report ultimately adopts "watch only", but must acknowledge that the bull case's evidence is far stronger than the first draft assumed, and that the first draft contained one serious omission.

⚠️ First, owning the first draft's omission: the first draft read this filing as a straightforward "buyback support" story and made no mention at all of the fact that chairman Zhu Yiming had already sold 11.1106 million shares (1.58% of total share capital) between 2026-05-06 and 06-12, cashing out about RMB 4.4 billion at an average price of about RMB 396, with the disposal plan fully executed (sources: Sina Finance¹¹, East Money¹²). That fact is the heart of the matter, and the "buyback + stake increase" headline buried it.

Key facts (verified; closing price RMB 364.03, total market cap RMB 255.456 billion)

The bull case's four strongest points

1. The PE the market is using is stale: the commonly cited PE(TTM) of 88.9x uses a denominator that excludes that RMB 5.4 billion of Q2 profit. On the most conservative basis of "H1 ex-non-recurring annualised to RMB 9.7 billion", it is only 26.3x; ex-non-recurring TTM including the guidance works out at 44.2x.
2. Upstream prices have not fallen, and GigaDevice's categories are still rising: 3Q26 conventional DRAM contract prices are up 13–18% quarter on quarter and NAND up 10–15% (still rising) (TrendForce, data as of 2026-07-03, to be verified); domestic NOR makers raised their full line by 25% from 7/6 (64Mb from RMB 6.34 to RMB 7.92) (Huaqiangbei channel basis, data as of 2026-07-11, 19 days ago; the late-July trend is unverified, to be verified); niche DRAM DDR3 1Gb +30.9% and DDR4 1Gb +32.7%, with "a price but no goods" and no new capacity planned by the original manufacturers (Huaqiangbei channel basis, data as of 2026-07-11, to be verified). GigaDevice's product mix sits precisely on the two tightest segments (NOR + niche DRAM), whereas what is being cut overseas is HBM/enterprise NAND/PC memory — not the same supply/demand pool.
3. ⚠️ The internal structure of the sector on 7/29 does not actually support "the whole memory chain collapsing" (this point directly corrects the first draft's wording): on the same day four memory design and interface names closed up or flat — 江波龙 (Longsys, 301308, ChiNext) −0.11%, 佰维存储 (Biwin, 688525, STAR Market) +0.98%, 澜起科技 (Montage Technology, 688008, STAR Market) +0.71%, 紫光国微 (Unigroup Guoxin, 002049, SZ main board) +1.63%; what hit limit-down was packaging (通富微电 (TFME, 002156) −10.00%) and compute hardware (紫光股份 (Unisplendour, 000938) −10.00%, though it is still +24.72% over 20 days, i.e. a high-position catch-down). → 7/29 was not "memory de-rating", it was a catch-down in compute hardware that had not yet fallen plus a stock-specific sell-off in GigaDevice; the de-rating on the memory design side was already finished in early-to-mid July.
4. The buyback is "cancellation and capital reduction" rather than treasury shares (permanently extinguishing shares, directly boosting EPS and net assets per share); Zhu Yiming has also undertaken not to sell for 12 months from 2026-07-29; and the 12/13 start date for his buying may not have been his choice — his disposal completed on 6/12, and a shareholder above 5% may not buy within 6 months of selling (the short-swing trading restriction), so 6/12 + 6 months = 12/12, i.e. "I start buying on the first day the law allows me to" (this is an inference; the filing does not state the reason, to be verified).

The bear case's seven strongest points (the side this report ultimately adopts)

1. Once you price the support, it cannot hold: the RMB 2.0 billion buyback ceiling = 0.783% of market cap; the buyback plus the stake increase, RMB 3 billion in total, = 1.174% of market cap and = 9.2% of 7/29's single-day turnover of RMB 32.44 billion, and it must be spread over the coming year. Using roughly 1%-scale money to fight a stock that is down 52.85% over 20 days with 13.5% daily turnover is symbolism, not support.
2. ⚠️ The cash-flow direction is a net outflow of RMB 3.4 billion: the major shareholder took RMB 4.4 billion out of the secondary market in May–June (average price about RMB 396, and the current price of RMB 364.03 is only 92% of that), and two months later proposes that "the company" spend RMB 2.0 billion buying and that he spend RMB 1.0 billion buying. His own hard-cash consideration is only RMB 1.0 billion (and that is a "no less than" floor, with no ceiling, no price and no cadence undertaking), while the RMB 2.0 billion buyback is all shareholders' money. Net cash-out is about RMB 3.4 billion.
3. For roughly the next month, the support on the tape is RMB 0: every public statement says the chairman "proposes"; the buyback price ceiling, term and funding source are all undisclosed; and a buyback for the purpose of reducing registered capital requires a shareholder-meeting resolution → from proposal to the first order placed, the timeline runs in months.
4. About RMB 2.05 billion of that RMB 6.9 billion H1 figure is stock-trading paper gains, and July is giving them back day by day: RMB 6.9 billion − RMB 4.850 billion ex-non-recurring = about RMB 2.05 billion of non-recurring gains, which the company itself attributes to "an increase in the period-end fair value of securities investments held". Quantifiable give-back so far: the 650,600-share stake in 联讯仪器 (Lianxun Instruments, 688808) was worth RMB 1.511 billion at end-June → at the 7/29 close of RMB 1,788.14 it is RMB 1.163 billion, a shrinkage of RMB 348 million (−23.0%); 臻宝科技 (Zhenbao Technology, 688797) is −53.8% so far in July (the number of shares held is undisclosed → the precise amount is subject to no reliable data available; only the negative direction is certain). → Q3 statements are very likely to contain a fair-value change loss that directly reduces net profit attributable to parent. The chain "H1 net profit attributable to parent RMB 6.9 billion → annualised RMB 13.8 billion → PE 18.5x" is therefore poisoned.
5. The second derivative of the cycle has turned negative, and downstream spot prices are already falling: contract price increases have narrowed from previously more than 30% to 13–18% (TrendForce, data as of 2026-07-03, about 27 days ago, to be verified), and TrendForce's own text attributes the cooling to "not an improvement in supply, but consumers reaching the limit of what they can bear"; meanwhile downstream channel spot prices have already fallen 20–30% from their highs (DDR4 16GB RMB 650 versus a prior high above 800; DDR5 16GB RMB 1,080–1,200 versus a prior high of 1,500; used DDR4 about RMB 450 versus more than 700 at the start of the year. Huaqiangbei channel prices, data as of 2026-07-21, 9 days ago; the trend since then is unverified, to be verified). The bear case's core argument: the financial statements are the last to know — the 57% gross margin in Q1/Q2 rests on the timing gap of "procurement costs locked in low while selling prices rise"; once the price increases stop, the scissors close in reverse.
6. Those RMB 59.4–63.7 billion of "implicit assets" in ChangXin cut both ways: the valuation benchmark is itself the market price of 长鑫科技 (ChangXin Memory, 688825) on day 3 of listing, with about 93% of shares locked up, daily turnover of about 20%, up about 6.1x in 3 days and a market cap of about RMB 3.54 trillion (media basis, to be verified); and it does not enter the income statement and is not cash (it only lifts other comprehensive income); more importantly — if the global memory cycle peaks, ChangXin's RMB 3.54 trillion market cap is the first thing to be re-rated, at which point GigaDevice's "implicit asset" and its core-business earnings would shrink at the same time. This is not risk diversification, it is a double bet on the same risk exposure.
7. The A/H safety margin has disappeared: the H shares (03986.HK) closed at HKD 436.60 on 7/29 ≈ RMB 377.02, a premium of only +3.6% over the A shares at RMB 364.03 (versus about 17% at the end of June) — the internationally priced half has already surrendered the premium, and the A shares are no longer "relatively cheap".

This report's conclusion and reasoning: maintain "watch only", but reclassify it from "a buyback cannot save the cycle" to "a bet in which both sides are extremely strong and which will not be settled by an announcement".

• The decisive basis for adopting the bear case is not the cycle, but the RMB 3.4 billion of net cash-out plus the fact that the buyback is only a proposal: for roughly the next month there is no actual buying on the tape, while the largest known seller has just taken RMB 4.4 billion out at an average price above the current price.
• But two places where the first draft got the bull case wrong must also be recorded: ① the first draft's claim of "a heavy fall across the whole memory chain" was inaccurate — on 7/29 the memory design and interface names (Longsys 301308, Biwin 688525 and Montage Technology 688008 are all non-main-board; Unigroup Guoxin 002049 is Shenzhen main board) closed up or flat, and what actually hit limit-down was packaging and compute hardware; ② the first draft described GigaDevice as having only a "buyback story", making no mention at all of its extremely strong operating data: H1 ex-non-recurring +791%, Q2 alone ex-non-recurring RMB 3.440 billion, and a Q1 gross margin of 57.08%.
• The core disagreement in one sentence: is that RMB 3.440 billion of quarterly ex-non-recurring profit in Q2 the start of a new normal, or the peak of this cycle? This disagreement will not be settled by an announcement; it will only be settled by two lines of numbers in Q3 — the gross margin and the fair-value change.