Credit Brief — Midea Group Co., Ltd. (美的集团股份有限公司)

000333.SZ · Shenzhen (A-share; H-shares: 00300.HK, listed 2024-09-17) · Filing window: FY2023–FY2025 annual reports + 2026 interim · Sources: official periodic reports filed via cninfo (巨潮资讯网)

⚠️ EXTRACTION DEFECT NOTICE — THE FINANCIAL SECTIONS (§4–§6) AND §9 SAMPLE TABLE OF THIS SAMPLE ARE NOT USABLE. This is the fourth cold-start run of the name → public-fetch → mechanical-extraction → English-brief chain, and the first with the QCC phase live (§1–§2 are populated and usable). The fetch and pipeline mechanics ran end-to-end at zero LLM spend, and the Phase 4.5 validator fired for the first time and failed the run correctly — but Midea's financial-statement layout breaks the mechanical extractor in four compounding ways: (1) the primary statements are combined consolidated-and-parent tables (「合并及公司资产负债表」: consolidated + parent columns for two dates side by side) with values line-wrapped across physical rows — the row parser took not a single candidate from the primary statements (「资产总计」 was never matched, so total assets / total liabilities / total equity are absent altogether); (2) rows carry a note-reference column (「附注」: 四(1), 四(2)…) — the BYD layout defect family, still unfixed upstream; (3) the unit declaration 「(除特别注明外，金额单位为人民币千元)」 is a 「单位为：X」-family phrasing the detector does not recognize, so the run normalized values as if RMB yuan; (4) starved of primary-statement candidates, the item-name matcher fell through to the notes section — every provenance cell points at note tables (pp. 222–254 of each annual report: credit-loss rows, disposal proceeds, and even bare year tokens leaking in as "values"). The 2026 interim report's statement rows are single-line, but its split year headers defeated period detection (years=[]), so the interim was excluded from the merge (no MinerU token configured → no OCR fallback). The tables in §4–§5 are retained as extracted, verbatim, solely to document the defect and validate the rendering format — do not cite any number in them. Ground-truth anchors quoted directly from the filing text are at the top of §4; §9 documents the mechanism. Remediation belongs to the extraction engine (upstream repo); this brief must be rebuilt after the fix.

All figures are extracted mechanically from official filings and are traceable to source page (see §9). Narrative sections are analyst commentary. Not investment advice. Not a credit rating.

1. Identity & legitimacy

• Company name: 美的集团股份有限公司
• USCC (unified social credit code): 91440606722473344C
• Registration status: 存续（在营、开业、在册） — active / in existence
• Founded: 2000-04-07
• Legal representative: 方洪波
• Registered capital: CNY 7,606.4 mn
• Headcount band: 900-999
• Company type: 股份有限公司（上市、自然人投资或控股） — joint-stock company
• Actual controllers: 何享健 (total 26.936%)

Source: QCC (企查查) corporate-registry screen, retrieved 2026-10-04.

• Auditor: PwC Zhong Tian (普华永道中天会计师事务所（特殊普通合伙）； Hong Kong auditor: PwC (罗兵咸永道会计师事务所)) — standard unqualified (clean) opinions for FY2023, FY2024 and FY2025 (FY2025 report signed 2026-03-30, report no. 普华永道中天审字(2026)第10014号; FY2024 signed 2025-03-28, report no. 普华永道中天审字(2025)第10014号; FY2023 signed 2024-03-26, report no. 普华永道中天审字(2024)第10017号 — per each annual report's audit-report section, read directly from the filing text)
• Headcount caveat: the QCC band "900-999" reflects the parent entity only (927 employees per the FY2025 annual report's employee section); group headcount incl. major subsidiaries is 198,570. Read the QCC band as a registry artifact for large multi-subsidiary issuers, not as group size.

2. Compliance & legal

• No records found: Dishonest judgment debtor (失信被执行人)
• No records found: Person subject to court enforcement (被执行人)
• No records found: Administrative penalties (行政处罚)
• No records found: Business-exception listings (经营异常名录)

Source: QCC (企查查) risk-screen, retrieved 2026-10-04. “No records found” is a clean screen against the cited registries — a positive signal, not a legal opinion.

3. Business profile

Midea Group Co., Ltd. is one of the world's largest home-appliance manufacturers, tracing its origins to a 1968 workshop in Shunde, Guangdong, and listed in Shenzhen since 2013 (A-share 000333) and in Hong Kong since September 2024 (H-share 00300). The group runs a ToC/ToB dual matrix: a smart-home business spanning HVAC (air conditioning) and consumer appliances, plus industrial-technology businesses in new energy & industrial technology, smart building technology, and robotics & automation — the latter anchored by Germany's KUKA, acquired in 2017. Overseas sales carry a high share of group revenue — FY2025 MD&A reports overseas revenue of CNY 195.9bn against CNY 458.5bn total operating revenue (≈43%) — which makes FX translation and overseas demand-cycle exposure structurally larger than at domestic-focused peers. Founder He Xiangjian (何享健) remains actual controller (26.936% per the QCC screen, §1); president and legal representative is Fang Hongbo (方洪波). For scale, the filing-text anchors in §9 put consolidated total assets at CNY 608.8bn at FY2025 year-end and FY2025 revenue at CNY 458.5bn — quoted from the filings directly because the mechanical extraction of these statements is defective (see the notice at the top).

Top customer/supplier tables not generated for this issuer.

4. Financial dashboard (CNY mn)

⚠️ DEFECTIVE EXTRACTION — DO NOT CITE. Values below are rendered verbatim from the extraction output (outputs/extracted/财务数据.json) to document the defect. Visible failure modes: nearly all values are 0.0 or — (the primary statements yielded no candidates at all); the few non-zero cells are note-table rows misfiled under statement line items (e.g. "Accounts receivable 0.4" is actually a receivables credit-loss row from the notes, p.227); 「资产总计」/「负债合计」/「所有者权益合计」 never matched, so there are no Total rows. Ground-truth anchors quoted from the filing text follow; §9 documents the mechanism.

Ground-truth anchors, quoted directly from the filings' text (consolidated 「合并」 columns; statements presented in RMB thousands, 「金额单位为人民币千元」) — for verifying the defect, not produced by the extraction pipeline:

(in RMB thousands; i.e. FY2025 total assets = CNY 608,791.8 mn. Locations: FY2023 values — balance sheet p.158 / income statement p.160 of the FY2023 AR, except FY2023 cash which was read from the comparative column of the FY2024 AR balance sheet p.155; FY2024 values — FY2024 AR balance sheet from p.155 / income statement p.157, with FY2024 total assets also visible as the comparative column of the FY2025 AR balance sheet p.131; FY2025 — FY2025 AR balance sheet p.131 / income statement p.134; 2026 H1 — interim balance sheet p.95 / income statement p.97.)

Assets (defective — see notice)

Liabilities

Equity

Income statement

5. Ratios

⚠️ DEFECTIVE EXTRACTION — DO NOT CITE. Ratios are computed from the same defective inputs (§4). Rows rendered 0.00x/1.00x or — are artifacts of zero-valued numerators and denominators; the "high" interest-coverage flag at 1.00x and the industry-benchmark lines are equally meaningless for this issuer until the extraction is fixed.

Mechanical risk flags

• ⚠ EBITDA interest coverage: 1.00 (level: high, reference threshold: <2.0)

Industry benchmark (peer set: 工商通用类 · benchmark rating A- · company rank 6/9)

• Total debt / EBITDA: 0.0x — strong (peer band ≤1.0)
• EBITDA interest coverage: 1.0x — soft (peer band 1.0~3.0)
• AR turnover: 0.0 turns — 差 (peer band <0.3)

6. Cross-check flags (mechanical reconciliation)

⚠️ DEFECTIVE INPUTS — DO NOT CITE the individual values. The reconciliation layer itself worked as designed and is one of the mechanical tripwires that caught the corruption: it raised 10 flags — 9 "same item across documents" conflicts (all severity high) and 1 cash-movement flag (mid) — every one a product of the note-table misfills and zeros described in the notice at the top. Sample bullets are kept verbatim below to document the noise pattern.

• Same item across different source documents — 9 flags (high×9); first 3 shown:
  • 2023 · Same item across different source documents: 10.7 vs 0.0 CNY mn (deviation 100.0%, severity high)
    Sources: 2025年年度报告.pdf vs 2024年年度报告.pdf. Two source documents differ by 10.7 CNY mn — typical causes are restatement or differing scopes; flagged for analyst review.
  • 2023 · Same item across different source documents: 0.0 vs 1,253.0 CNY mn (deviation 100.0%, severity high)
    Sources: 2025年年度报告.pdf vs 2023年年度报告.pdf. Two source documents differ by 1,253.0 CNY mn — typical causes are restatement or differing scopes; flagged for analyst review.
  • 2022 · Same item across different source documents: 0.0 vs 0.0 CNY mn (deviation 100.0%, severity high)
    Sources: 2024年年度报告.pdf vs 2023年年度报告.pdf. Two source documents differ by 0.0 CNY mn — typical causes are restatement or differing scopes; flagged for analyst review.
• 2023 · Cash-flow statement net increase vs balance-sheet cash movement: 8.8 vs 0.0 CNY mn (deviation 100.0%, severity mid)
  Sources: 2023年年度报告.pdf p.252 「现金及现金等价物净增加额」 vs 2023年年度报告.pdf p.252 「货币资金」. CF-statement figure above the balance-sheet movement by 8.8 CNY mn — typical causes are restricted-cash release or cash-equivalent scope changes; flagged for analyst review.

7 check(s) not run for missing inputs (5 distinct after dedup): 前五大应收 vs BS 应收账款; 前五大其他应收 vs BS 其他应收款; 前五大预付 vs BS 预付款项; 现金及现金等价物净增加额 vs 货币资金变动 (×3, one per filing); 授信表内总合计 vs Σ机构行.

7. Red flags

WITHHELD IN THIS SAMPLE. The product discipline requires every red flag to anchor mechanically to §4–§6 data (value + period). That data is defective for this issuer (see the extraction defect notice at the top), and fabricating anchors from hand-corrected numbers would violate the traceability rule the product stands on. Note that the two screens that did run clean are not red-flag material in themselves: the QCC compliance screens (§2) returned no records, and the auditor opinions are unqualified for FY2023–FY2025 (§1). Red flags for Midea will be written when the extraction engine handles the combined consolidated-and-parent statement layout and this brief is rebuilt from a clean extraction. For format reference, see the completed sample catl-300750.md.

8. Industry context

Home appliances (HVAC and consumer appliances) with a growing industrial arm — building technology, industrial technology, and robotics & automation (KUKA). The Chinese appliance industry is a mature consumer-durables sector: incremental demand is tied to the property cycle and replacement demand, and the knowledge base's industry page (家电企业, wiki/entities/家电企业.md) characterizes the core tension as market saturation with leaders squeezing the tail on scale, brand and channel — "white goods strong, black goods weak" — while OEM and e-commerce-dependent players see margins compressed. The sector's supply-chain position is strong — component suppliers have less pricing power than the finished-goods makers — so sector leaders typically run large supplier-settlement floats and strong operating cash conversion; that float is an industry-structure fact, and its size on any issuer's book must be read from clean financials.

Risk anchors for a creditor, per the knowledge-base cards (家电企业, wiki/entities/家电企业.md; L0 风险卡-装备制造, wiki/entities/风险卡-装备制造.md):

1. Property-cycle beta, with a known lag. Per the 家电企业 card, growth in residential property sales leads white-goods domestic sales by roughly 2–3 quarters; since 2022's property downturn the demand base has shifted to replacement upgrades, making subsidy calendars and replacement cycles the demand swing factor.
2. The float cuts both ways. The appliance leaders' scale gives them supplier-financing float and strong operating cash generation — but that float is also the fastest-moving liability in a demand downturn, and headline debt-to-asset ratios understate how much of the liability block is operating float rather than funded debt (家电企业 card).
3. Export share is double-edged. In the rating-method anchor table cited by the 家电企业 card, overseas-sales share scores best in a 25–30% band and worse above 30% — a formalization of the FX/logistics/geopolitical exposure; the card also flags tariff friction and subsidy withdrawal as open uncertainties for exporters, and the 装备制造 risk card likewise warns of large FX-swing distortions in earnings for export-heavy manufacturers.
4. Robotics/automation inherits equipment-maker mechanics. For the KUKA-anchored ToB business, the 装备制造 card's anchors apply: staged payments (advance → delivery → acceptance → warranty) misalign revenue and cash; receivables plus inventories swell fastest exactly when orders are strong; acceptance timing is buyer-controlled; and specialized production lines have near-zero residual value once orders cancel.
5. Demand-side policy pulses. Domestic demand has been recovering under trade-in (「以旧换新」) and green/energy-efficiency incentive programs per the 家电企业 card — a tailwind that fades by policy calendar, not by company choice; watch for payback effects after subsidy windows close.

9. Method & provenance

Pipeline ran fully mechanical in this run: cninfo fetch (4 filings: FY2023–FY2025 annual reports + 2026 interim, HTTP, no LLM) → inventory/processing → pdftotext-based financial extraction (MinerU OCR not invoked — no token configured; the engine's "try MinerU" fallbacks bailed immediately) → reconciliation sidecar → ratio computation → validation. Zero LLM calls at any phase: qualitative extraction and material digest were skipped by flags, the material-routing planner degraded to its static table by design (no API config passed), and the QCC phase was pre-seeded (see below). 18 phases, ~7.6s wall.

Phase 4.5 (validate_extraction) — first light after being skipped alongside --skip-qcc in all prior cold-start runs — fired and FAILED the run correctly: 3 errors (BS.总资产 = None; BS.总负债 = None; BS.所有者权益合计 = None — the totals were never extracted) and 2 warnings (no interim-period data; one value-level suspicion). This is the intended behavior: the validator is the mechanical gate that catches broken extractions before ratios are computed. In this run the pipeline continued past it (non-fatal at the orchestrator level) and computed ratios from the corrupt inputs — those ratios are quarantined in §5.

The extraction is defective for this issuer's layout — mechanism, verified against the raw pdftotext dumps (outputs/extracted/raw/):

1. Combined consolidated-and-parent statements. Midea presents 「合并及公司资产负债表」/「合并及公司利润表」 — four value columns (consolidated + parent, two dates) per row. In the FY2025 annual report, rows read 「货币资金 | 四(1) | 85,247,150 140,410,308 | 23,178,076 | 18,441,820」 with values wrapped across physical lines; the row parser matched no candidates from the primary statements at all (pp. 131–134 of the FY2025 AR; every provenance cell in the merged JSON points at pp. 222–254 — the notes section).
2. Note-reference column (「附注」四(n)) between line item and values — the BYD defect family, unfixed upstream.
3. Unit declaration unrecognized. The statements declare 「(除特别注明外，金额单位为人民币千元)」 — a 「单位为：X」-family phrasing (BYD defect ① family); the run normalized values as if RMB yuan (log: 「pdf_text 元→万元 已归一」).
4. Notes-section fallthrough. Starved of primary-statement rows, the item-name matcher filled slots from note tables — e.g. "Accounts receivable" ← 「应收账款坏账损失(附注四(5))」 (p.227), "Fixed assets" ← 「固定资产减值损失(附注四(17))」 (p.227) — producing zeros, credit-loss amounts and even bare year tokens (0.2025-style) as values.
5. 2026 interim excluded. The interim's statement rows are single-line, but its split year headers (「2026 年 / 6 月 30 日 / 合并」 on separate lines) defeated period detection (years=[] → "too thin" → no MinerU token → no markdown sidecar → coverage gap), so the merged dataset covers FY2021–FY2025 only.

Ground truth for the rebuild (consolidated, RMB thousands, quoted from filing text — see the anchor table in §4):

The auto-generated sample table below is retained verbatim as defect documentation (it shows the notes-section fallthrough at raw-token level); its "values" and pages are not usable citations. The QCC facts in §1–§2 and the auditor/headcount facts read directly from filing text are unaffected by the extraction defect.

This brief is built solely on officially filed public disclosures. It is not investment advice, not a credit rating, and not an offer of any service. Figures may contain extraction or caliber effects; verify against the original filings before relying on them.