A-Share Pre-Market Brief | 2026-08-05 Wednesday

AMEC pre-announced net profit attributable to parent of RMB 2.7–2.9 billion, +282.48%~+310.81% YoY — an extremely eye-catching headline. But taken apart:

• Ex-non-recurring net profit attributable to parent is only RMB 1.0–1.2 billion (+85.61%~+122.73%);
• the gap comes from fair-value changes and investment income on outside equity investments totalling about RMB 1.982 billion, up roughly RMB 1.815 billion YoY — this is non-recurring, not earned by the equipment core business;
• at the operating level: revenue about RMB 6.691 billion (+34.89%), gross profit up about RMB 682 million — solid core growth, but nowhere near as dramatic as +310%.

The market has already seen through this: on 8/4 the stock rose only 2.55%, with net main-force outflow of RMB 241 million, and it is still −24.35% over the past 20 sessions. "+310%" is the single most misleading number in this cycle, and this report explicitly does not use it as a basis for recommendation. Sources: Shanghai Securities News¹², Securities Times¹³

• Related news: at the 2026-08-04 US close, the Nasdaq was +2.59% at 26584.99 and Nvidia +3.25%, Meta +6.71%, Alphabet A +5.34%, Microsoft +4.69%, Amazon +4.7% (Sina Finance¹). On the industry side: TrendForce says 2026 capex of the nine major cloud providers will grow 90% and raised AI server shipment growth to 31%.
• Catalyst logic: direct order positive. North American cloud capex is the direct upstream variable for optical-module revenue; what it affects is revenue, not sentiment. Media report that its orders already cover all of 2026 and part of 2027 (21jingji¹⁴) — this order-coverage claim is a media paraphrase with no original company announcement found, so this report does not treat it as an established fact.
• Stage of the industry branch: fermentation (not ignition, not climax). On 8/4 the optical-module "易中天" trio rose together, the sector had a limit-up ladder (光迅科技 (Accelink), 剑桥科技 (C&Q Technology), 永鼎股份 (Yongding) all limit-up), and the core leader traded heavily — a healthy fermentation structure; but it is still −8.91% over the past 20 sessions, meaning the trend is still repairing, not in a climax phase.
• Fundamental verification: the core business is optical modules, a perfect match with the catalyst (not a cross-industry association). P/E (TTM) 79.97, total market cap RMB 1.20 trillion. The cheapest of the three optical-module giants (Eoptolink at 58.16 is lower, T&S Communications at 106.48 the highest). Down 8.91% over 20 sessions, the most resilient among peers. Note: this report did not break down its interim report item by item; the latest revenue/gross margin/cash flow figures cannot be confirmed for now, so the assessment is based only on valuation and fund flow.
• Industry position: global optical-module leader and the absolute core leader of this cycle.
• Technicals & sentiment: +13.24% to close at RMB 1021.99, with net main-force inflow of RMB 5.068 billion, first in the whole market, and a turnover rate of only 4.56% (low turnover in a high-priced stock, stable chips). Mind the ±20% band: +13.24% still leaves room to the limit, and must not be read as "near limit-up" using ±10% instincts.
• Final judgment: priority deep-dive. It defines the quality of today's main line — fund size, fundamental fit and trend integrity are all the best in the branch. The risk is gap-up disagreement after a large single-day gain.

• Related news: on 8/4 the "components" industry was +7.07%, 63 up / 0 down, net inflow RMB 6.507 billion, with 15 limit-ups, first among all industries (THS industry scope).
• Catalyst logic: direct order positive. High-layer boards for AI servers and switches are rising in both layer count and unit price, directly lifting revenue and gross profit.
• Stage of the industry branch: mid-fermentation. The sector already has 15 limit-ups (生益科技 (Shengyi), 深南电路 (Shennan Circuits), 鹏鼎控股 (Avary), 红板科技 (Redboard), 华正新材 (Huazheng), 奥士康 (Aoshikang), 博敏电子 (Bomin) and others) — a thick ladder with core leaders all present, but the limit-up count is already elevated and the probability of disagreement the next day is rising.
• Fundamental verification: the core business is PCB, directly tied to AI hardware demand. P/E 50.00, the lowest among PCB leaders (Shennan Circuits 62.32, Avary Holding 55.56, Redboard Technology 118.94). Market cap RMB 215.1 billion. Past 20 sessions −13.84%. Detailed financials were not checked line by line; marked as not yet verified.
• Industry position: a core supplier of high-layer PCB and one of the core leaders of this PCB cycle.
• Technicals & sentiment: limit-up +10.00% to close at RMB 111.77, net main-force inflow RMB 1.006 billion, turnover rate only 3.49% (limit-up with low turnover = holders reluctant to sell).
• Final judgment: priority deep-dive. Lowest valuation + limit-up + large net inflow + low turnover makes it the best value core leader in the PCB branch.

• Related news: same as 5.2, the components sector surge.
• Catalyst logic: direct order positive. High-frequency/high-speed copper-clad laminate is the upstream material for AI PCB, with both volume and price rising.
• Stage of the industry branch: fermentation.
• Fundamental verification: copper-clad laminate core business, P/E 69.39, market cap RMB 272.6 billion. Past 20 sessions −25.56%, the largest expectation gap among the Top 10 here — i.e. the positive is not yet fully reflected in the price. Financial details not verified line by line.
• Industry position: global copper-clad laminate leader.
• Technicals & sentiment: limit-up +10.00% to close at RMB 112.23, net main-force inflow RMB 1.157 billion, turnover 2.53% (limit-up on extremely low turnover).
• Final judgment: priority deep-dive. But this must be flagged: past 20 sessions −25.56% means the 8/4 limit-up is essentially the first bullish candle of an oversold bounce, and a trend reversal is not yet confirmed — a different animal from a "trend stock".

• Related news: same as 5.1, the across-the-board overnight surge in US tech.
• Catalyst logic: direct order positive, same logic source as Innolight.
• Stage of the industry branch: fermentation.
• Fundamental verification: P/E 58.16, the lowest of the three optical-module giants; market cap RMB 624.6 billion; past 20 sessions −12.17%. Financial details not verified line by line.
• Industry position: first tier of optical modules.
• Technicals & sentiment: +13.68% to close at RMB 448.00, net main-force inflow RMB 4.549 billion (second in the market), turnover 5.52%.
• Final judgment: priority deep-dive. Clear valuation advantage; risks are the same as Innolight's (gap-up disagreement, large single-day gain).

• Related news: released its interim report on the evening of 2026-08-03 — revenue RMB 28.897 billion (+38.93%), net profit attributable to parent RMB 11.08 billion (+29.43%), and an across-the-board raise to full-year guidance: revenue from RMB 51.3–53.0 billion to RMB 58.5–60.5 billion, continuing-operations revenue growth from 18–22% to 35–39%, backlog RMB 66.4 billion (21jingji¹⁵, The Paper¹⁶). Note: this is 8/3 news and falls outside today's time window.
• Catalyst logic: direct order positive, and the size of the guidance raise exceeded prior institutional expectations.
• Stage of the industry branch: already delivered, now at a relay position. Its 8/4 limit-up dragged 凯莱英 (Asymchem) to a limit-up, 康龙化成 (Pharmaron) +10.67% and 博腾股份 (Porton Pharma) +13.58% — a complete ladder.
• Fundamental verification: the CRDMO core business matches the catalyst exactly. P/E (TTM) of only 19.46, the lowest of all 24 names here; market cap RMB 421.8 billion. Net margin about 38.3% (11.08/28.897), far above most names in this report. The RMB 66.4 billion backlog is a disclosed fact and provides revenue visibility.
• Industry position: global CRDMO leader.
• Technicals & sentiment: limit-up +10.00% to close at RMB 141.35, net main-force inflow RMB 1.833 billion, past 20 sessions +19.48% — unlike the oversold bounces in AI hardware, this is acceleration within an uptrend.
• Final judgment: watch closely (not a top pick). Its fundamentals-plus-valuation combination is the most solid in the ranking, but the catalyst was already delivered on 8/4, so buying today means buying the expectation of a second board rather than new news. If you want a "new catalyst", it does not qualify; if you want "fundamental certainty", it is first in the ranking.

• Related news: same as 5.2.
• Catalyst logic: direct order positive, AI server/switch PCB.
• Stage of the industry branch: fermentation.
• Fundamental verification: P/E 62.32, market cap RMB 226.5 billion, past 20 sessions −19.31%. Financial details not verified line by line.
• Industry position: PCB leader (dual core businesses of package substrates and high-layer boards).
• Technicals & sentiment: limit-up +10.00% to close at RMB 332.54, net main-force inflow RMB 563 million, turnover only 1.58% (among the lowest in the market, extremely stable chips).
• Final judgment: watch closely. A low-turnover limit-up is high quality, but with 15 limit-ups already in the sector, guard against a broad disagreement.

• Related news: H1 pre-announced net profit attributable to parent of RMB 1.055–1.143 billion, +34%~+45% YoY; plus sector resonance from WuXi AppTec's guidance raise (East Money¹⁷).
• Catalyst logic: direct earnings positive plus industry trend positive, layered together.
• Stage of the industry branch: the only branch in a confirmed uptrend in this report.
• Fundamental verification: CRDMO core business matches. P/E 43.30, market cap RMB 73.3 billion. Past 20 sessions +30.59%, the strongest trend of all 24 names.
• Industry position: second-tier CXO leader.
• Technicals & sentiment: +10.67% to close at RMB 39.92 (on ChiNext with ±20%, +10.67% is not a limit-up), net main-force inflow RMB 237 million, turnover 7.46%.
• Final judgment: priority deep-dive. It is the only name here that simultaneously has a new earnings pre-announcement, an upward trend and a non-extreme valuation. Risks: it has already risen for several sessions, and part of the catalyst comes from WuXi (external).

• Related news: same as 5.1.
• Catalyst logic: direct order positive, optical devices supporting CPO/optical modules.
• Stage of the industry branch: fermentation, but this stock had the largest gain in the branch.
• Fundamental verification: P/E 106.48, the highest of the three optical-module giants; market cap RMB 231.3 billion; past 20 sessions −13.15%.
• Industry position: leader in the optical-device niche.
• Technicals & sentiment: +17.45%, the strongest optical-module name on 8/4, net main-force inflow RMB 2.986 billion, turnover 7.20%.
• Final judgment: watch closely. Largest gain + highest valuation = the worst risk/reward in the branch; if the sector gaps up into disagreement today, it is most likely the one that falls the most.

• Related news: new on 8/4 — PC majors HP, ASUS and Acer have begun using small quantities of CXMT DRAM chips, with qualification completed around mid-year (ITHome⁸, Ifeng Tech¹⁸). Layered on top, the semiconductor industry saw single-day net inflow of RMB 23.368 billion on 8/4, first in the whole market.
• Catalyst logic: industry trend positive (not a direct order). The grading must be explicit: CXMT is not listed, and GigaDevice and CXMT are on the same side of the domestic memory ecosystem — this is not "GigaDevice won an HP order". The original report also states that "the quantities adopted are very limited and the models are mainly sold outside the US". Do not over-extrapolate.
• Stage of the industry branch: the ignition phase of an oversold bounce. The memory sector is broadly −40%−57% over the past 20 sessions (德明利 (Demingli) −57.21%, 佰维存储 (Biwin Storage) −48.82%, 江波龙 (Longsys) −46.48%), and rebounded collectively +4%+9% on 8/4.
• Fundamental verification: core business is NOR Flash / MCU / DRAM, matching the domestic memory theme. P/E 86.91, market cap RMB 249.8 billion. Past 20 sessions −42.58%, expectations extremely low. Financial details not verified line by line.
• Industry position: domestic NOR Flash / MCU leader.
• Technicals & sentiment: +4.48% to close at RMB 356.02, net main-force inflow RMB 1.595 billion, turnover 9.46%. A modest gain but large inflow — the shape of "money ahead of price".
• Final judgment: watch closely. It is the only name here with all three of "new news + low position + large net inflow", but the catalyst is an indirect read-across and is less hard than optical modules and PCB.

• Related news: CXO sector resonance (spillover from WuXi AppTec's guidance raise); the stock itself has no new announcement.
• Catalyst logic: industry trend positive (not its own catalyst).
• Stage of the industry branch: in sync with WuXi, already delivered.
• Fundamental verification: small-molecule CDMO core business. P/E 52.51, market cap RMB 58.3 billion, past 20 sessions +5.56%.
• Industry position: small-molecule CDMO leader.
• Technicals & sentiment: limit-up +10.00% to close at RMB 161.61, net main-force inflow RMB 250 million, turnover 1.79% (low-turnover limit-up).
• Final judgment: watch closely. Limit-up quality is good, but the catalyst comes entirely from spillover from the leader and it has no news of its own — a follower, whose persistence depends on whether WuXi can post a second board.

The 8/4 post-close announcement of "plans to use up to RMB 7 billion of its own funds for entrusted wealth management" was widely read by the market as a positive showing an abundance of cash. After pulling the financial statements, that statement is in direct conflict with the company's balance sheet:

Three contradictions along the timeline (all publicly disclosed facts):

1. 2026-04-21: the company accepted a shareholder loan of up to RMB 2 billion from its second-largest shareholder (annualised rate ≤4.5%) (National Business Daily¹⁹).
2. 2026-06-26: announced a plan to raise RMB 8 billion via a private placement, of which RMB 1.5 billion was explicitly for "replenishing working capital and repaying bank loans", precisely because the gearing ratio had reached 85.89% at the end of 2026Q1 (Securities Times²⁰).
3. 2026-08-04: announced that it has "RMB 7 billion of its own idle funds" available for entrusted wealth management.

In June it said "gearing is 85.89%, we need to raise RMB 1.5 billion from the market to replenish working capital and repay loans"; 39 days later it says "we have RMB 7 billion of idle own funds to invest". Moreover, placement proceeds are legally subject to dedicated-account management and cannot be counted as "own funds", so "the placement money has arrived" cannot make this self-consistent.

Other key facts:

• The announcement allocates RMB 6 billion to "agreement deposits" — an agreement deposit is a cash-management tool for corporate settlement accounts and is not a wealth-management investment in substance; against a backdrop of RMB 25.89 billion of interest-bearing debt and receivables already pledged to secure RMB 4.248 billion of finance leasing, it is more likely a deposit-balance arrangement required by the banks. This is analytical inference, not company disclosure.
• Free cash flow: −RMB 12.63 billion in 2025 and −RMB 7.31 billion in 2026Q1; the company has never generated positive free cash flow since listing, with all growth funded by debt and equity issuance.
• The depreciation cliff has not yet landed: 2026Q1 fixed assets RMB 11.248 billion + construction in progress RMB 6.039 billion; actual depreciation charged in 2025 was only RMB 598 million. On a five-year straight-line basis, steady-state annual depreciation would be about RMB 2.1–3.3 billion (this report's calculation, not company disclosure). The company has still not publicly stated the depreciation life of its compute servers, and has not replied even after an investor formally asked in June 2026.
• The company has never disclosed by name the customers, contract amounts or terms of its compute business, which already accounted for 43.6% of 2025 revenue.
• P/B 22.7x, versus only 2.1–5.1x for IDC peers with the same business model (宝信软件 (Baosight Software) 5.14×, 光环新网 (Sinnet) 2.14×); the 30% ROE is mainly levered out of 5.0x gearing, and that leverage is already at the ceiling of an 85.89% gearing ratio.

Positive facts that deserve a fair record: the growth is real (2026Q1 revenue +192.9%, net profit attributable to parent +343.4%; 2026H1 pre-announced RMB 1.5–1.9 billion, +247%~+340%); ex-non-recurring quality is high (2026Q1 ex-non-recurring RMB 751 million ≥ headline RMB 750 million, so it does not rely on non-recurring items); goodwill is zero; there is no lock-up expiry overhang; and the pledge ratio is only 7.27%. If it delivers RMB 4 billion of net profit attributable to parent in 2026, the forward P/E would be about 29.5x, comparable to 工业富联 (Foxconn Industrial Internet) at 29.3×.

Conclusion: Pass. Not because the company is fabricating or the growth is fake, but because — the announcement being read as a positive today cannot have its "RMB 7 billion of own funds" wording supported by the balance sheet, and the company stays silent on the three pieces of information that determine its value (depreciation life, compute customers, contract amounts). On a stock already up 329.75% over the past year at a 22.7x P/B, using an announcement that conflicts with the books as a reason to buy does not produce a workable risk/reward. (Recorded as a factual note only, with no inference drawn: in April 2026 the second-largest shareholder priced an inquiry-based transfer at RMB 241.37 per share, almost exactly matching the 8/4 close of RMB 241.00.)

This is the only heavyweight company announcement inside today's news window, and also the one most likely to lead people astray via the number "RMB 7.670 billion". After reading the original announcement (No. 2026-040) and checking the financials:

① The company is not a compute company. 2025 revenue mix: digital goods integrated operations RMB 1.880 billion, 89.97% of revenue, 16.83% gross margin (buying, distributing and marking up phone credit/data/video membership vouchers); IoT data operations RMB 137 million (6.56%, and down 3.85% YoY in 2025); cloud terminal technology and cloud services only RMB 56 million (2.69%).

② There are almost no compute assets on the books. 2026Q1 fixed assets RMB 230.5 million, construction in progress RMB 138.9 million, intangible assets RMB 683.2 million. Of these, intangibles jumped from RMB 9.36 million in 2024 to RMB 688 million, and the company's "2024 Annual Financial Statement Report" states verbatim: "the increase in intangible assets from the beginning of the period was mainly due to the company acquiring land use rights during the period"; construction in progress is the "Fengzhushou Future Tower" project — i.e. that money bought land and an office building, not data halls and servers (Shenzhen Stock Exchange original²¹). Existing compute capacity relies on renting third-party IDC.

③ The contracting entity is a company founded five months ago. 雅安云算 (Ya'an Yunsuan) was established in March 2026 with registered capital of RMB 50 million — the contract value is 153 times its registered capital. The announcement also states that "over the past three years, the company and Ya'an Yunsuan have had no similar transactions" with them (Companies A/B), i.e. no history of cooperation.

④ The core business fundamentals keep deteriorating (verified against akshare financials):

Revenue CAGR over three years is about 33%, but gross margin and net margin are both in one-way decline — the growth was bought with concessions. Operating cash flow was negative for 4.5 of the years from 2021 to 2025. In 2026Q1, receivables of RMB 1.146 billion + prepayments of RMB 1.194 billion = RMB 2.340 billion, 63.0% of total assets and more than the entire net assets attributable to parent (RMB 1.910 billion).

⑤ Earnings quality: this "big order" does not improve the profit structure. RMB 4.608 billion ÷ five years = RMB 921.6 million of tax-inclusive revenue per year, corresponding to average annual net profit of RMB 60–72 million, an implied net margin of 6.5%–7.8% (6.9%–8.3% excluding tax) — essentially the same level as the company's existing thin-margin distribution business (5.76%–7.28%). That is: using heavy assets plus high leverage to earn a margin about as thin as the existing asset-light business.

⑥ The funding gap is a hard constraint. The announcement states explicitly that within the RMB 3.062 billion procurement, "payment from own funds is expected to account for 5% to 20%, and finance-lease funding 80% to 95%". Yet the payment clause of the procurement agreement reads: "within 5 working days after the agreement is signed and takes effect, Party A shall pay Party B a prepayment of 30% of the total amount agreed in the agreement" — taken literally that is about RMB 919 million, while the company's cash at the end of 2026Q1 was only RMB 116.7 million. The board separately says this will use "batched ordering and batched payment"; there is interpretive space between the two statements and the announcement does not clarify it (flagged here as an open item).

⑦ The company's own risk warning is unusually heavy (announcement verbatim):

"Under the finance-lease model, the lessor has the right to repossess and auction the compute servers, resulting in the loss of the company's core assets and a halt to compute services, and potentially triggering cross-default and property-preservation measures, with a material adverse effect on the company's ability to continue as a going concern."

⑧ A procedural fact widely missed: the RMB 3.062 billion procurement agreement has not yet taken effect. The announcement states that the proposal "still needs to be submitted to the company's shareholders' meeting for approval", and Dentons' legal opinion says the procurement agreement "takes effect from the date it is approved by Fengzhushou's shareholders' meeting". The services contract is already effective, but without servers no service can be provided. Until the shareholders' meeting approves it, this transaction is legally incomplete.

⑨ The terms are asymmetric. Company A's late-delivery penalty is "RMB 250,000 per day, capped at RMB 5 million" — the cap is only 0.16% of the RMB 3.062 billion purchase value. If Company A fails to deliver on time, the company's recourse against its supplier is nearly negligible, while it still bears breach liability toward downstream Company B. In addition, the announcement declares that A and B are unrelated to Fengzhushou, but it does not declare that Company A and Company B are unrelated to each other (cannot be verified, flagged as an open item).

⑩ A technical detail worth pressing on: the services contract's network configuration specifies "including China Mobile public-network bandwidth 4G, excluding dedicated lines". For large-scale distributed training clusters that configuration is clearly on the low side (training normally needs IB/RoCE high-speed interconnect and large egress bandwidth), and looks more suited to inference/rendering/offline batch workloads — if so, unit compute pricing and renewal elasticity would be lower than the market imagines for "AI training intelligent-compute leasing". This is an analytical judgment, not company disclosure.

⑪ Scale comparison (the most jarring set of numbers): the 15.49% single-day gain on 8/4 added about RMB 1.57 billion of market cap, which is 4.4–5.2x the project's total expected net profit over its full five-year life (RMB 300–360 million). The implied P/E on the incremental profit is 163–195x.

⑫ On the shareholder side: the largest tradable shareholder, 海峡创新 (Haixia Innovation) (8.12% stake), has pledged 80.92% of its holding cumulatively, with 云南国际信托 (Yunnan International Trust) as pledgee; in May 2026 the pre-IPO lock-up shares were released (controlling shareholder 罗洪鹏 (Luo Hongpeng) holds 22.07%), and the −17.09% over the past 60 sessions coincides with that release date (correlation, not proof of causation). Goodwill is only RMB 813,200, so there is no goodwill risk.

Conclusion: Pass. This is indeed a real order with a contract, a value and a supporting legal opinion (not a concept read-across), but its nature is "a heavy-asset sublease business that levers the company's own balance sheet, earns the spread between upstream and downstream, and concentrates credit risk on a single anonymous customer". With cash of RMB 116.7 million, operating cash flow negative for years, six consecutive years of declining core gross margin and a procurement agreement not yet in effect, the single-day increase in market cap already equals 4–5x the project's total five-year net profit — the risk/reward does not hold.

有研新材 (Grirem Advanced Materials, 600206, SSE main board) — headline growth +40.89%, ex-non-recurring only +16.98%

• The revenue mix decides everything: in 2025, platinum-group products (auto exhaust catalyst materials) generated RMB 3.992 billion of revenue, 41.8% of the total, but at a gross margin of only 1.63%, contributing just 7.6% of gross profit; thin-film materials (i.e. semiconductor sputtering targets) generated RMB 2.288 billion, 24.0% of revenue, at an 18.63% gross margin, yet contributed 49.6% of gross profit.
• The 50% revenue growth this period came mostly from platinum-group +65% (1.63% gross margin) rather than targets — which is exactly why net profit attributable to parent was +40.89% while ex-non-recurring was only +16.98% (the gap comes from RMB 30.67 million of investment income, RMB 18.38 million of fair-value changes and so on). On a trailing ex-non-recurring basis, the P/E is about 130x.
• Cash flow has deteriorated badly: 2026H1 operating cash flow was −RMB 950.4 million (5.2x the full-year net profit attributable to parent); inventory RMB 1.886 billion (+49.24%), receivables RMB 1.489 billion (+44.04%), cash −55.91%; asset impairment (inventory write-down) −RMB 83.666 million, 4.2x the prior-year period.
• The evidence on targets themselves is solid (this deserves a fair record): the reply to the Shanghai Stock Exchange's inquiry letter lists customers verbatim as "subsidiaries of China Electronics Technology Group, BOE A, HiSilicon Optoelectronics, TSMC, 芯联集成 (Xinlian Integrated), CXMT and YMTC"; the 2026 interim report states that "in the first half, 319 sample-validation products were advanced in parallel, with 105 cumulatively completing full-process reliability qualification and entering volume supply, covering 7nm and below advanced logic, next-generation HBM memory, 3D NAND stacking and advanced packaging", and that "target product sales revenue grew more than 45% YoY".
• But "number one in domestic targets" does not hold up against public filings: 江丰电子 (KFMI, 300666, ChiNext) had 2025 ultra-high-purity target revenue of RMB 2.850 billion at a 34.24% gross margin, both better than Grirem's thin-film materials at RMB 2.288 billion / 18.63%.
• The structural change under way is the key point: the company plans to transfer a 45% stake in 有研稀土 (GRIREM Rare Earth) (priced at RMB 454.2 million) and a 51% stake in 有研国晶辉 (GRINM Guojinghui) (priced at RMB 243.2 million) to 中国有研 (China GRINM), after which neither will be consolidated, with the aim of "further focusing on the integrated-circuit core business". If completed, the company would go from "a non-ferrous new-materials general store" to something close to a pure semiconductor target platform — that is the real pricing logic behind a 96x P/E, but it still requires shareholders' meeting approval.
• Claims circulating in the market that must be flagged as unconfirmable: ① "SMIC is one of its target customers" — "SMIC" cannot be found anywhere in the full inquiry-letter reply or the full 2026 interim report; ② "passed TSMC 5nm qualification" — official filings do not specify a process node; ③ "100% domestic share in cobalt targets, 60%+ in copper targets" — not found in any official document; ④ "target gross margin around 40%" — in direct conflict with the 18.63% disclosed in the annual report. All four lack official evidence and this report does not accept them.
• Judgment: watch only. The catalyst tier is indirect supply-chain beneficiary (with named customers and evidence of volume supply, so the evidence quality is relatively good within that tier), but the high growth in the current statements is driven mainly by low-margin trading-type business, the ex-non-recurring valuation is extremely high, and cash flow is a large net outflow.

江南新材 (Jiangnan New Material, 603124, SSE main board) — the two placement projects are completely different in nature and cannot be lumped together

• The core business is essentially PCB electroplating copper consumables, not a PCB maker: in 2025, copper ball revenue was RMB 7.868 billion, 76.2% of the total, but at a gross margin of only 2.54% (copper prices pass straight through plus a wafer-thin processing fee); copper oxide powder RMB 1.871 billion, 18.1%, at an 11.01% gross margin; high-precision copper-based heat sinks RMB 308.8 million, 3.0%. Overall gross margin is only 4.10%.
• The copper oxide powder project = genuine capacity expansion. The product is used in PCB copper electroplating (HDI, IC substrates, FPC, mSAP and other high-precision boards) and is a real existing business; the IPO prospectus states that it covers 28 of the top 30 PCB makers by composite ranking (including 鹏鼎控股 (Avary Holding), 深南电路 (Shennan Circuits), 胜宏科技 (Victory Giant), 奥士康 (Aoshikang) and others). This AI → high-end PCB → electroplating consumables transmission chain is real and verifiable.
• The liquid-cooling module project = clearly more story than substance. What exists today is "high-precision copper-based heat sinks" (copper base plates), not finished modules; the announcement itself says the proceeds are to "open up back-end processes such as CNC machining, microchannel machining and surface cleaning, and automated surface inspection" and to "fill a capacity gap" — i.e. the module-level production line is not yet complete. More importantly: the heat-sink gross margin has already collapsed from 28.24% (2023) → 24.08% (2024) → 5.18% (2025) (back-solving implies only about 2.8% in 2025H2 alone), so the volume was bought with low-margin orders. And the announcement never discloses a single customer name or a single order value anywhere; searches also found no public evidence of cooperation with any server maker, CDU maker or Nvidia cold-plate supply chain (the Group A cold-plate suppliers announced by Nvidia at GTC 2026 were AVC, Cooler Master, Jentech and Delta — Jiangnan New Material is not among them).
• Profit quality is actually good: 2026H1 pre-announced net profit attributable to parent RMB 160–175 million (+51.5%~+65.7%), with ex-non-recurring RMB 144.5–159.5 million (+55.6%~+71.8%), growing faster than the headline figure.
• But cash flow is the fatal flaw: cumulative operating cash flow over the five years 2021–2025 was −RMB 3.71 billion, against cumulative net profit attributable to parent of only +RMB 790 million (2025 OCF/net profit = −4.03x). The cause is that upstream electrolytic copper smelters give almost no credit terms (2026Q1 accounts payable were just RMB 104 million) while downstream PCB makers demand them — every extra RMB 1 of revenue requires extra working capital.
• Two facts that correct market perception: ① dilution is not 30%. The 43.72 million shares / 30% is the issuance cap; based on the 8/4 close of RMB 110.18, actual dilution works out to about 9.1%, or about 11.1% at a floor price of 80% of that. ② It is raising again before the previous proceeds are spent: net IPO proceeds were RMB 331.1 million, and as of 2026-06-30 utilisation was only 61.50%, while this placement seeks RMB 1.6 billion = 4.83x the net IPO amount.
• The RMB 8 billion of revenue at full capacity must be discounted: on the current product mix, RMB 4.8 billion of copper oxide powder corresponds to about RMB 530 million of gross profit; RMB 3.2 billion of liquid-cooling modules at the current 5% heat-sink gross margin corresponds to only RMB 160 million. Big revenue ≠ big profit, and the announcement itself notes that it "does not constitute a profit forecast or an earnings commitment". R&D spending is only 0.48% of revenue, so the moat comes more from scale and customer qualification than from a technology premium.
• Judgment: watch closely (the copper oxide powder line); the liquid-cooling line awaits order verification. The catalyst tier is industry trend, and the placement announcement itself contains no direct order at all.