Zhongke Sanhuan (000970.SZ) Initiation Coverage: Cyclical Tailwinds Can't Offset the Question of Earnings Quality

Rating: Neutral (Watch/Avoid) | Target Price Range: RMB 8.8–9.9 | Current Price: RMB 10.77 (close, 2026-10-20)

I. Summary of Key Conclusions

Zhongke Sanhuan is a veteran leader in the first tier of China's sintered NdFeB industry, with the highest export share among peers at ~55%; its Bosch/TDK-grade overseas clients and high-grade technology are genuine assets. However, RMB 6.6 billion of net assets attributable to shareholders generated only RMB 91 million of net profit attributable to the parent in 2025 (weighted ROE 1.41%), with profitability ranking last among the four comparable companies. The gross margin recovery in 2025–2026H1 (10.8%→12.5%) is essentially a benefit of low-cost inventory and lagged price pass-through during a praseodymium-neodymium (PrNd) price rally, not an improvement in pricing power. The current price of RMB 10.77 corresponds to a P/B of ~2.0x and a TTM P/E of ~136x; sell-side consensus implies 2026H2 net profit attributable to the parent of nearly 3x the first half — an aggressive assumption. Measured against the base-case fair value of RMB 8.8–9.9 (peer median P/B discounted for the ROE gap), the stock is mildly overvalued with a negative margin of safety. Over the next six months, two key variables — the expiration of the rare earth export control suspension on 2026-11-10 and earnings delivery in the Q3 report/annual report — will determine the direction of the odds. Until then, we maintain Watch/Avoid: a break below RMB 550,000/tonne for PrNd or the signing of the acquisition at a high price would be bearish signals, while gross margin validating above 14% or the share price falling back below the base-case fair value range would be re-rating signals.

II. Investment Opportunity Assessment

Opportunity evidence grade B, market environment neutral. The stock does not yet constitute an allocatable opportunity: the current price is ~15.8% above the midpoint of base-case fair value. Upside depends on the bull scenario (tighter export controls driving a second PrNd price surge, plus a robotics theme revival; fair value RMB 13.4–15.7), while downside in the bear scenario (PrNd breakdown, further inventory impairments; fair value RMB 7.0–7.8) is of comparable magnitude, with expected return of roughly −6% from the current price.

What changes over the next six months: ① The export control suspension expires on 2026-11-10 — if extended, exports stay on the +30%–40% trajectory (2026H1 overseas revenue RMB 1.998 billion, +44%); if controls resume, per the 2025 precedent of −13.3%, a 1–2 quarter delivery gap would emerge; ② Q4–2027Q1 is the window for magnet product prices to pass through raw material cost increases (order pricing lags raw materials by 1–2 months); ③ whether 2026 net profit attributable to the parent can reach the stock option incentive threshold of RMB 97.71 million (requires ≥RMB 48.49 million in H2, only +2.5% YoY).

Primary share price driver: The PrNd price center and price pass-through (inventory of RMB 3.115 billion, 31% of total assets, gives gross margin and impairment two-way high elasticity); secondarily, the export control milestone.

Verifiable expectation gap (negative): Back-solving from the current price implies normalized net profit attributable to the parent of ~RMB 420 million (market cap ÷ peer median P/E of 31.4x), roughly 3x the company's actual earnings (TTM net profit attributable to the parent of RMB 97 million; 2026H1 annualized non-GAAP profit of RMB 65 million) and ~50% above sell-side consensus 2027E (RMB 280 million); meanwhile, the company has announced no robotics design wins or volume supply contracts (on the investor platform it acknowledged related collaboration has "relatively small order volumes"), so the theme premium lacks evidence.

Verification catalysts: Q3 report (scheduled for 2026-10-24), export control suspension expiry (2026-11-10), magnet price pass-through (2026Q4), Zhongdian Cisheng acquisition pricing (exclusivity period until 2027-03-31). Falsification conditions: PrNd oxide falling below RMB 550,000/tonne for two consecutive months, renewed export declines upon control resumption, or the acquisition closing at a high price.

III. Investment Thesis

C1 — Zhongke Sanhuan has the weakest earnings quality among comparable leading magnetic materials companies: RMB 6.6 billion of net assets attributable to shareholders generated only RMB 91 million of net profit in 2025 (weighted ROE 1.41%, estimated ROIC ~1.9%); full-cycle non-GAAP net profit swung violently between +RMB 782 million (2022) and −RMB 17 million (2024), with a center below RMB 100 million; minority interest income persistently diverts 20%–48% of consolidated net profit, structurally diluting earnings quality. (Confidence 0.85)

Key evidence:

• Zhongke Sanhuan 2025 Annual Report / Key accounting data and financial indicators: weighted average return on net assets 1.41%; net profit attributable to parent RMB 91,318,606.81, non-GAAP ~RMB 61 million (non-recurring items were 33.6% of net profit attributable to parent, including government grants of RMB 46.40 million and FX gains of RMB 40.72 million).
• Zhongke Sanhuan 2024 Annual Report / Key accounting data (non-GAAP net profit line): 2024 non-GAAP net profit −RMB 17,265,329.59 (first negative for the full cycle); 2023 was RMB 219,128,838.23; the 2022 non-GAAP peak was RMB 782 million (2025 annual report series).
• Zhongke Sanhuan 2025 Annual Report / Significant non-wholly-owned subsidiaries: Ningbo Sanhuan Cisheng Industry & Trade (company holds 56%) posted net profit of RMB 128,689,732.05, exceeding consolidated net profit of RMB 114 million; minority interest income was 47.7% of consolidated net profit in 2024, 20.0% in 2025, and 30.4% in 2026H1.

It should be noted that "last place" mainly applies to net margin and ROE: Sanhuan's 2026H1 magnetic materials gross margin of 12.63% was on par with Zhenghai Magnetic (~12.5%); the gap lies in expenses, FX gains/losses, and minority interests — which underscores that the problem is in earnings structure and capital returns, not gross margin per se. This thesis fails if the 2026 annual report shows magnetic materials gross margin above 15% and non-GAAP net profit of RMB 150 million, or if minority interest diversion falls below 10% after the acquisition is restructured.

C2 — The gross margin recovery in 2025–2026H1 (consolidated 10.8%→12.5%) is essentially a low-cost inventory windfall and lagged pass-through gains during a rising PrNd price cycle, not an improvement in pricing power — the company itself stated that long-term contracts lock in prices and raw material cost pass-through lags, leaving "order quotations frequently facing unfavorable conditions." PrNd oxide has already fallen ~22% from its early-2026 peak of RMB 900,000–950,000/tonne to RMB 740,000/tonne (~59th percentile over the past decade); the export control suspension expiring on 2026-11-10 is the largest two-way variable — cyclical attributes outweigh growth attributes. (Confidence 0.68)

Key evidence:

• Zhongke Sanhuan 2024 Annual Report / Raw material price volatility risk: "Although the company prices on a cost-plus basis with reference to raw material price fluctuations... due to the lagged transmission of raw material price volatility, magnet producers' order quotations frequently face unfavorable conditions, and we are forced to build inventory, tying up working capital."
• Zhongke Sanhuan 2026 Semi-Annual Report / Operating discussion: magnetic materials gross margin 12.63%, up 3.24 percentage points YoY; inventory of RMB 3.115 billion, with the inventory write-down provision balance falling from RMB 170 million at end-2024 to RMB 87 million (reversals/write-offs releasing gains from stockpiled materials).
• SunSirs rare earth market data (2026-10-20) / PrNd oxide quotes: steady at RMB 735,000/tonne, having risen from ~RMB 400,000/tonne in September 2025 to ~RMB 735,000 currently (nearly doubled); per Baichuan Yingfu, mid-October at RMB 743,000–746,000/tonne, at the ~59th percentile of the 2016–2026 range and ~22% below the early-2026 peak.

PrNd surged to RMB 900,000–950,000/tonne in January–February 2026 and has since retraced for 8 months; the benefit pool of the gross margin recovery (write-down provision write-offs) is thinning. However, the possibility of a "new center" must be flagged — quota system constraints combined with an overseas official floor price (~RMB 890,000/tonne including tax) providing support may leave the PrNd center systematically above the 2023–2024 range of RMB 350,000–550,000; this is the main uncertainty in this thesis and is reflected in the bull scenario. This "cyclical windfall" judgment fails if magnetic materials gross margin stays above 14% even as PrNd falls below RMB 650,000/tonne, or if the company shifts its long-term contract pricing to quarterly raw material linkage.

C3 — The proposed cash acquisition of Zhongdian Cisheng (holder of a 44% minority stake in core subsidiary Sanhuan Cisheng, plus consumer electronics magnetic assembly capabilities) and the deeply out-of-the-money stock option incentive plan (exercise price RMB 15.99; targets of 2026/2027 net profit +7%/+23% vs. 2025) constitute two types of marginal improvement options: if priced reasonably, the Cisheng stake consolidation alone could add ~RMB 20–30 million annualized to net profit attributable to the parent (equivalent to roughly 20%–30% of 2025 levels) and ease minority interest diversion; but pricing is undetermined, the exclusivity period runs to March 2027, and realization is uncertain. (Confidence 0.52)

Key evidence:

• Zhongke Sanhuan suggestive announcement on planning an equity acquisition and signing a letter of intent (2026-08-05) / exclusivity clause: "From the effective date of this letter of intent to March 31, 2027, Party B and Party C commit to exclusively advance the completion of this acquisition with Party A"; the counterparty is a minority shareholder of the company's controlled subsidiary Sanhuan Cisheng Industry & Trade; cash acquisition, with pricing to be negotiated separately on the basis of a state-asset-filed appraisal.
• Zhongke Sanhuan 2025 Annual Report / Significant non-wholly-owned subsidiaries: Sanhuan Cisheng 2025 revenue RMB 3.014 billion, net profit RMB 129 million, minority stake 44.00%; look-through estimate: acquiring 51% of Zhongdian Cisheng would add ~RMB 29 million annualized to net profit attributable to the parent.
• Zhongke Sanhuan announcement on completion of initial grant registration of the 2026 stock option incentive plan (2026-09-02): exercise price RMB 15.99/share (~48% premium to current price); targets based on 2025 net profit: 2026 growth ≥7% (≥RMB 97.71 million) and 2027 ≥23% (≥RMB 112 million).

Key consideration risk: Sanhuan Cisheng paid RMB 340 million in dividends to minority shareholders in 2025 (of which RMB 286 million in H1), far exceeding the listed company's own dividend (RMB 30.09 million) — if the acquisition is priced high, it amounts to "paying a second time" for cash already distributed, depleting the RMB 966 million net cash buffer. This thesis fails if the acquisition is terminated, or if the 44% stake is valued at more than double its book net assets.

C4 — The current price of RMB 10.77 implies a P/B of ~2.0x and a TTM P/E of ~136x, already pricing in an optimistic combination of "sustained high PrNd prices + gross margin convergence with peers + the robotics theme": sell-side consensus 2026E net profit attributable to the parent of RMB 195 million implies H2 profit of nearly 3x H1, while the company's H1 annualized non-GAAP profit was only RMB 65 million. Measured against the base-case fair value of RMB 8.8–9.9 (peer median P/B of 2.03x discounted 15%–20% for the ROE gap), the current price is ~9%–22% higher, the margin of safety is negative, and the odds skew downward (though two-way volatility is large). (Confidence 0.58)

Key evidence:

• Zhongke Sanhuan 2026 Semi-Annual Results Express (2026-07-21): book value per share RMB 5.43 (2026-06-30), implying P/B of ~1.98x at current price; TTM net profit attributable to parent of ~RMB 97 million implies a P/E of ~135.6x (our estimate).
• Sell-side consensus (median of 3 estimates over the past 6 months, market data terminal, 2026-10-20): 2026E EPS RMB 0.16 (net profit attributable to parent ~RMB 195 million, implying +110.8%); 2027E RMB 0.23 (~RMB 280 million); consensus implies 2026H2 net profit attributable to parent of RMB 146 million, 2.96x the actual H1 figure (RMB 49 million) (our estimate).
• Broker research report (2026-08-23) / target price: "Lowered target price to RMB 18.67 (from RMB 19.76, corresponding to 3.6x 2026E P/B)" — the sell-side anchor of 3.4–3.6x P/B is about 1.5x Sanhuan's own 5-year median P/B of 2.35x, a distinctly theme-driven valuation.

Valuation anchor discipline: this report's base-case P/B of 1.6–1.8x is anchored on the peer median P/B of 2.03x (median of Zhenghai 2.03 / Yunsheng 1.79 / Jinli 4.22) discounted 15%–20% for the ROE gap (Sanhuan normalized ROE ~3.3% vs. peers' 5.5%–6.8%), and also sits at the lower edge of its own 5-year median P/B of 2.35x — independent of the current multiple. The apparent "low percentile" of the stock's P/B at the 21st percentile of 5 years must be read alongside ROE — at the same P/B, its earnings content is only 1/4–1/3 of peers'; a low percentile does not mean cheap. The valuation judgment needs revisiting if 2026 net profit attributable to the parent reaches RMB 190 million or above (sell-side consensus level), or if the share price breaks below RMB 7.8 into the bear-case range.

IV. Core Financial Data and Latest Results Review

4.1 Core Financial Data (Unit: RMB 100 million, unless otherwise noted)

Note: 2026Q1 single-quarter revenue was RMB 1.677 billion (+14.8%) with net profit attributable to parent of RMB 16 million (+21.5%); net cash (cash-like assets − interest-bearing debt) was ~RMB 966 million as of 2026-06-30 (~RMB 0.79/share).

4.2 Reasons for Metric Changes (Items with YoY ≥ ±20%)

• FY2024 net profit attributable to parent −95.6% / non-GAAP turning negative: The company attributed this to the lagged pass-through of falling rare earth raw material prices, pressured selling prices, and RMB 140 million of inventory write-downs (2024 annual report: asset impairment losses of RMB 140 million).
• FY2025 net profit attributable to parent +660.5%: The company attributed this to volume growth (+24.99%), cost reduction and efficiency gains, higher FX gains (RMB 40.72 million), and a sharp reduction in asset impairment losses (provision of RMB 49 million, write-offs of RMB 102 million) — roughly one-third of profit came from non-recurring items; non-GAAP profit was only RMB 61 million.
• FY2025 operating cash flow −105.9% (+RMB 1.326 billion → −RMB 78 million): The company attributed this to increased cash outlays for procurement (cash paid for goods RMB 5.976 billion vs. RMB 5.114 billion the prior year) — cash tied up in high-priced stockpiled materials; the profit recovery lacked cash backing.
• 2026H1 financial expenses +232.6% (+RMB 83.15 million): Net FX loss of RMB 78.76 million (vs. a gain of RMB 67.29 million in the prior-year period), equivalent to 1.6x net profit attributable to parent for the period — the main cause of "revenue growth without profit growth."
• 2026H1 overseas revenue +44.1%: Exports re-accelerated after the control suspension, with the share recovering from 47.5% to 55.3%.

4.3 Latest Results Review (2026H1)

2026H1 revenue was RMB 3.614 billion (+23.7%), net profit attributable to parent RMB 49.21 million (+11.9%), non-GAAP net profit RMB 32.30 million (only +2.2%) — revenue rose on both volume and price (strong volume growth + PrNd average prices up ~35%+ YoY), but profit was squeezed by three factors: ① FX loss of RMB 78.76 million (RMB appreciation; forward hedging coverage only ~3.9%); ② Q2 single-quarter gross margin fell 0.70pp QoQ to 12.14% (high-priced raw materials beginning to bite, low-cost inventory gains fading); ③ minority interest diversion of 30.4%. By quarter, Q1/Q2 net profit attributable to parent was RMB 16 million / RMB 33 million respectively; Q2 single-quarter non-GAAP profit of RMB 24.03 million rose sharply QoQ — the marginal direction is improving but absolute levels remain thin. Against sell-side consensus: 2026E net profit attributable to parent of RMB 195 million implies RMB 146 million in H2, nearly 3x H1, requiring gross margin to jump above 15% or large inventory gains — a run rate of only RMB 65 million annualized non-GAAP in H1 does not support it. Against the option incentive threshold of RMB 97.71 million: H2 needs ≥RMB 48.49 million (only +2.5% YoY), achievable provided the 2025Q4 pattern of "year-end concentrated provisions" (single-quarter net profit attributable of just RMB 0.99 million) does not repeat. Management gave no quantitative guidance, leaving no reliable market anchor for judging "beats or misses"; only the two internal benchmarks above can serve as references.

V. Business Model and Earnings Quality

Business model: Asset-heavy manufacturing (sintered NdFeB capacity of 25,000 tons plus bonded 1,500 tons), build-to-order production, predominantly direct sales, cost-plus pricing with long-term contract price locks; raw materials account for ~79% of costs. Exports make up roughly 50–55% of revenue (the highest among the four comparables), settled in USD/EUR. No recurring revenue base and weak pricing power — price pass-through lags by 1–2 quarters; during price upcycles the company books inventory gains first, and during downturns it absorbs impairments first.

Cash content of earnings: OCF/net profit attributable to parent for 2023–2026H1 was 5.2 → 110.4 → −0.86 → 0.64, swinging violently in sync with stockpiling/de-stocking cycles; FCF/net profit over the same period was 3.65 → 92.5 → −2.57 → −0.68. In FY2025 profit recovered but OCF turned negative and FCF was −RMB 235 million — book profits during the price upcycle turned into RMB 3.1 billion of inventory; "profit turning into inventory" is the norm for this business. On a cross-cycle view (combined 2023+2024 FCF of RMB 2.1 billion) cash generation is acceptable, but single-year volatility is extreme. Recurring earnings test: Of FY2025 net profit attributable to parent of RMB 91 million, non-recurring items were RMB 30.71 million (33.6%), mainly government grants of RMB 46.4 million and FX gains of RMB 40.72 million — stripping these out, core earnings were only RMB 61 million; moreover, in 2026H1 FX already flipped to a loss. The "two-way swing" of non-recurring items is itself a negative for earnings quality.

Capital returns: Estimated ROIC ~1.9% (FY2025), far below any reasonable WACC over the long run; weighted ROE went from 14.85% in 2022 → 4.23% in 2023 → 0.18% in 2024 → 1.41% in 2025. Even on an optimistic normalized attributable profit of RMB 220 million under the "new center," ROE would be only ~3.3% — below even the 8% "pseudo-growth" warning line, and still weak. This is the core context for valuation judgment.

CapEx quality: CapEx/depreciation fell from 1.20 in FY2024 → 0.86 in FY2025; the 2022 rights-issue-funded projects (including phase one of the 5,000-ton project) closed out in December 2024, ending the investment cycle, with construction in progress of only RMB 56 million. Nominal capacity utilization is only about 50% (50.4% by 2025 sales volume, 53.8% by output), so there is no capacity bottleneck and no need for large maintenance capex — depreciation cash flow covers maintenance spending. The problem is not cash burn but return on invested capital.

Moat and red flags: Real assets (mass production of high-grade N55, grain boundary diffusion, 2–3-year automotive certification cycles, CAS background) are a barrier against commoditized low/mid-end capacity, but bargaining power is weak on both sides — against the upstream duopoly and downstream major customers. Red flags: ① profit is concentrated in the 56%-owned Cixing Gongmao subsidiary (its net profit of RMB 129 million exceeds consolidated net profit), leaving a low share of benefit for listed-company shareholders; ② customer concentration is acceptable (top five at 20.9%), but the top five suppliers account for 30.2% of purchases, leaving raw materials dependent on rare earth groups.

VI. Management Assessment

Consistency between words and deeds: partially delivered, with an aversion to quantitative commitments. The 2024 annual report pledged to "advance construction and expansion of production bases and continuously expand high-performance capacity" — actual CapEx declined for consecutive years (425 → 215 → 156 million RMB), capacity remained at the 25,000-ton level, the 2025 annual report stopped disclosing capacity figures, and investor relations activities explicitly stated "production will be arranged according to orders and market demand; we will not expand capacity blindly." Qualitative commitments were partially delivered (mainly technical upgrades), with no quantitative guidance (when asked for full-year guidance at the April 2026 results briefing, the answer was "please refer to announcements"). At the results briefing, investors directly questioned "impairments swinging wildly" and "all annual earnings go to paying wages"; management responded with standard templates and made no improvement commitments — a clear negative in the track record of words versus deeds.

Shareholder friendliness: neutral-to-weak. Payout ratio went from 22.1% in FY2023 → ~200% in FY2024 (RMB 0.2 per 10 shares in a marginal-profit year, drawing on retained earnings) → 33.0% in FY2025 (RMB 30.09 million); the current dividend yield is ~0.23%, with payout capacity constrained by low profitability. In 2024 the company bought back 1.22% of shares (RMB 108 million) but all were converted into 2026 option incentive shares rather than cancelled. The most glaring contrast: in 2025, controlled subsidiary Cixing Gongmao paid out RMB 340 million in cash dividends to its 44% minority shareholders, while dividends to all shareholders of the listed company were only RMB 30.09 million — cash is prioritized toward in-group minority shareholders, with listed-company shareholders coming last. After the March 2022 rights issue raised RMB 666 million, capacity stood still; capital allocation discipline is mediocre.

Risk signals: Related-party sales commissions exceeded the board-approved limit by 18% (TRIDUS RMB 11.83 million vs. approved RMB 10 million) — a flaw in internal limit management; third-largest foreign shareholder TAIGENE reduced its stake by 2.06 million shares in 2026H1 (0.17% of total share capital, modest); controlling shareholder Sanhuan Holdings' 2025 pledges have all been released with none outstanding currently, and no director/executive changes during the window period.

VII. Segment Breakdown

The company has a single main business: magnetic materials products account for 96.0% of revenue (FY2025: RMB 6.377 billion, gross margin 11.20%, −1.56pp YoY), with the remainder from minor other businesses; there is no disclosable multi-segment structure. The most informative breakdown is by geography:

Exports are the profit engine: For 2026H1, export gross profit contribution is ~55% × 14.0% ≈ 7.7 gross margin points vs. domestic ~44.7% × 10.8% ≈ 4.8 points — exports account for ~62% of gross profit, the company's actual profit engine, and hence triply sensitive to export controls, FX, and overseas demand (highly concentrated risk). The 3–4pp gross margin gap between overseas and domestic sales reflects overseas customers' certification barriers and a pricing environment superior to the domestic price-war market.

VIII. Accounting Practices and Inter-Period Consistency

Accounting red flags:

• Pro-cyclical swings in inventory write-downs (moderate): RMB 140 million provisioned in 2024 depressing the base → RMB 102 million reversed/written off in 2025 boosting profit, amplifying earnings volatility (2025 annual report asset impairment notes).
• Profit concentrated in a low-ownership subsidiary (moderate): Cixing Gongmao's net profit of RMB 129 million exceeds consolidated net profit of RMB 114 million, with minorities taking 20%–48%.
• Related-party commissions exceeding approved limits (moderate): TRIDUS sales commissions of RMB 11.8309 million exceeded the approved limit of RMB 10 million; the annual report itself disclosed "exceeds approved limit: Yes."
• Abnormal Q4 profit contraction (minor): Q4 2025 net profit attributable to parent was only RMB 0.9895 million, with recurring profit at −RMB 3.02 million; the full-year result effectively hinged on Q3, and the timing of year-end concentrated provisions warrants continued attention.
• Persistently high effective tax rate (minor, in the conservative direction): 36.5% (2025) → 31.2% (2026H1), due to RMB 251 million of deductible losses at loss-making subsidiaries (Wakehua, Ganzhou Sanhuan) with no deferred tax assets recognized — the reported tax burden is distorted but in a conservative direction.

Inter-period consistency: ① The sharp swings in OCF/net profit align with stockpiling cycles; management's explanations are consistent ("increased cash payments for procurement"), with no contradictions found; ② period-end magnetic materials inventory rose +77%/+66% for two consecutive years (712 → 1,260 → 2,096 tons), attributed by the company in the two years to "increased consignment goods" and "increased goods in transit," respectively — similar in nature, neither confirmed as finished-goods pileup, mutually corroborating the elevated inventory balance; ③ minority interest share of profit fluctuated at high levels, 47.7% → 20.0% → 30.4% — unexplained by the company; combined with the RMB 340 million subsidiary dividend payout, this is the most important governance-level anomaly to track.

IX. Valuation and Odds

Current market data (closing price 2026-10-20): Share price RMB 10.77, market cap RMB 13.09 billion, P/B 1.98x, PE(TTM) 135.6x, PS(TTM) 1.79x. Percentiles: P/B at the 21st percentile over 5 years and 14th over 10 years (5-year range 1.4–4.03x, median 2.35x); PE(TTM) at the 72nd percentile over 5 years — for a cyclical with razor-thin earnings, PE percentiles are mechanically distorted and serve as reference only, not headline; P/B is the primary anchor, but low percentiles must be read alongside ROE (TTM ROE 1.46% vs. peers' 5.5%–9.7%; at comparable P/B, earnings content is only 1/4–1/3 of peers').

Peer benchmarking (as of 2026-10-20 close):

What the market already implies: Back-solving the current price implies normalized attributable profit of ~RMB 420 million (market cap ÷ peer median PE of 31.4x) — 4.4x TTM earnings and more than 6x actual annualized recurring earnings (~RMB 65 million), and 50% above sell-side consensus 2027E (RMB 280 million). In P/B terms: the market gives Sanhuan nearly the same P/B as Zhenghai on the assumption of "profitability converging toward Zhenghai/Yunsheng," yet Sanhuan's ROE is one-quarter of theirs. The current price requires the company to deliver RMB 280–420 million in attributable profit within two years, versus an actual run-rate of ~RMB 100 million — the gap must be closed either by a jump in gross margin (needs 15%+, seen only in the 2021–2022 supercycle) or by a sustained thematic premium.

Three-tier value (EPV): Asset value RMB 5.43/share (book value per share as of 2026-06-30; inventory carried at cost with only RMB 87 million of write-down provisions remaining; upside from unbooked revaluation under rising praseodymium-neodymium prices, but raw material stock covers only 1–2 months of usage, limiting elasticity) — this is the floor reference, not additive with other tiers. EPV zero growth: using a 2023–2026E attributable profit center of RMB 120–150 million (weak-cycle basis), take RMB 135 million; cost of equity 10.5%; FCFE ≈ attributable profit (CapEx at 0.86x depreciation, working capital neutral cross-cycle); this report's computed EPV is ~RMB 1.06/share (range RMB 0.90–1.23; using the mid-cycle RMB 220 million basis incorporating the new center, RMB 1.48–1.94). ~90% of the current price is supported by the growth/option tiers — the market is paying almost entirely for "rising PrNd center + gross margin recovery + robotics/M&A themes," with zero-growth value nearly negligible. This extremely high proportion itself shows this is not a mispricing where "the market only prices zero growth," but rather the treatment of cyclicality and thematic options as the main course.

Three scenarios and odds (anchoring rationale: base-case P/B 1.6–1.8x = peer median 2.03x discounted 15%–20% for ROE, cross-checked against the lower end of its own 5-year median of 2.35x; bear-case P/B 1.3–1.4x anchored to the 2024 actual trough trading range and its own 5-year low; bull-case P/B 2.4–2.8x, below the 3.05x implied by the 52-week high of RMB 16.55):

Probability-weighted fair value is ~RMB 10.1 (expected return ~−6%), with a safety margin of ~−18% (on the base-case fair-value lower bound, this report's computation). Base-case cross-check: RMB 11.3 billion market cap ≈ next-cycle peak attributable profit of RMB 400–450 million × PE 20–25x (50–60% of the 2022 peak of RMB 800 million) + net cash — the two paths are self-consistent; normalized EPS of RMB 0.181 × PE 45±5x (retaining a sector-option premium) = RMB 7.2–9.1, connecting with the lower bound of the base-case range. External anchor comparison: sell-side target price median RMB 18.7 (3 brokers), minimum RMB 12.87 — both this report's bear and base cases sit below the sell-side floor; the divergence stems from sell-side anchoring to 3.4–3.6x P/B and gross margin recovery to 15%+ (thematic assumptions), whereas this report anchors to a discounted peer median and realized earnings. Sell-side coverage is only 3–6 brokers, all Buy-rated, so their floor does not constrain this report. Our forecasts vs. third parties: this report projects FY2026E revenue of RMB 7.8–8.1 billion and attributable profit of RMB 115–150 million (23%–41% below sell-side consensus); FY2027E revenue of RMB 8.2–9.0 billion and attributable profit of RMB 180–270 million (4%–36% below consensus); management gives no quantitative guidance, and the closest internal anchor is the option vesting thresholds (2026 ≥ RMB 98 million, 2027 ≥ RMB 112 million) — our forecasts fall "above the vesting thresholds and below consensus."

Conclusion: Mildly overvalued; target price range RMB 8.8–9.9 (base-case fair value). Separate quality from price — the assets are real and the channels scarce (medium quality), but the price has already front-loaded "peer-level earnings convergence + theme delivery." Evidence of a bubble is insufficient (P/B is only half the 2021–22 supercycle peak of 4.03x, and the stock has already retraced 35% from its 52-week high), so it is not classified as a bubble, but rather a "expensive but not broken" wait-and-see zone.

X. Industry Landscape and Competitive Structure

Industry size: Global NdFeB demand is expected to grow from 168,300 tonnes in 2021 to approximately 250,000–330,000 tonnes by 2025 (251,600 tonnes per Huajing Industry Research Institute; over 330,000 tonnes per CCX Pengyuan, corresponding to roughly 110,000 tonnes of praseodymium-neodymium oxide consumption), a CAGR of approximately 10.6%; China's output in 2025 is expected to be around 285,000 tonnes, accounting for over 70% of the global total, with approximately 85% of global high-performance NdFeB produced in China (BCG, as cited by CCX Pengyuan). Over the next 3–5 years: CCX Pengyuan expects high-performance demand CAGR to exceed 10%, driven by the triple overlap of new energy vehicles, wind power, and robotics; broker estimates put NEV magnetic material demand at 69,000 → 85,000 tonnes from 2025 → 2026, and wind power at 11,000 → 15,000 tonnes. The other side of demand quality: inverter air-conditioner compressors (47,000 tonnes of domestic consumption in 2024, per Antaike) already have high penetration and contribute a stable base rather than incremental growth; declining direct-drive penetration in wind power and lower unit consumption constitute structural drags.

Industry chain and value distribution: Rare earth raw materials (praseodymium-neodymium oxide/metal) → magnet manufacturing (blank → grain boundary diffusion → machining → surface treatment) → motors/OEMs/home appliances. The upstream is monopolized by a "northern-southern duopoly" (Northern Rare Earth controls roughly 75% of light rare earth mining quotas; China Rare Earth Group monopolizes medium and heavy rare earths), so magnet makers, without mines, have weak bargaining power upstream; the downstream consists of major automakers and home appliance giants, generally with annual price-reduction clauses — the magnetic materials segment is squeezed from both ends, earning processing fees and inventory gains, with profits concentrating upstream during price upcycles (already realized this round on the mining side: Baogang's 2026Q4 concentrate related-party price of RMB 38,769/tonne). Zhongke Sanhuan partially offsets this through 100% centralized procurement plus equity participation in long-term contracts, but its 2026H1 gross margin of 12.6% vs. Jingli's 21.7% and Yunsheng's 20.8% confirms that profit divergence relies on technological premium and componentization, not industry-wide price rises.

Supply-demand and competitive landscape: China supplies roughly 90% of global rare earth permanent magnet materials. Leading high-grade capacity (end-2024, CCX Pengyuan figures): JL MAG 32,000 tonnes (38,000 tonnes on a blank basis, planned 60,000 tonnes by 2027), Zhenghai Magnetic 30,000 tonnes (up to 36,000 tonnes), Zhongke Sanhuan 25,000 tonnes (management: "unchanged for now, depending on orders and market demand"), Ningbo Yunsheng 21,000 tonnes (blanks already at 26,000 tonnes plus 10,000 tonnes of grain boundary diffusion), Dadi Xiong 18,000 tonnes, Innuovo 13,000 tonnes. Concentration: the industry as a whole is highly fragmented (over 180 sintered NdFeB enterprises, top three at roughly 10%), but the high-performance sub-market is concentrating toward the first tier; entry barriers lie in automotive qualification (2–3 years), high-grade formulations and grain boundary diffusion know-how, batch consistency, and raw material supply security. Competitive intensity: homogeneous price wars in the low-to-mid end (industry profits bottomed in 2024), while the high end competes on grades/qualification/componentization; head players' counter-cyclical capacity expansion in 2025–2026 implies post-2027 high-end capacity competition risk — Sanhuan is standing still, with a declining relative share trend. Substitution threats: ferrite/SmCo limited to low-end and ultra-high-temperature scenarios; Western "de-China-ification" (the US Department of Defense taking equity in MP and locking in ten-year purchases at a floor price of USD 110/kg, equivalent to RMB 890,000/tonne including tax; Lynas commissioning dysprosium-terbium separation) is a medium-to-long-term share threat, but US magnet capacity was only about 5,500 tonnes in 2024, unable to shake China's 90% share in the short term.

Cycles and regulation (cycle timing detailed in the next chapter): four main regulatory threads — ① The Rare Earth Management Regulations effective October 2024 plus the August 2025 interim measures for aggregate volume control, with quotas plus full-chain traceability (2025 mining quota of 300,000 tonnes REO, roughly 64% of the global total); ② Export control timeline: April 2025 controls on seven categories of medium/heavy rare earths including samarium, gadolinium, terbium, and dysprosium (still in effect, directly restricting export licenses for Dy/Tb-containing magnets) → the four expanded announcements of October 2025 suspended from 2025-11-07 until 2026-11-10 (the single largest policy variable: extension means continued high export growth; reinstatement means a repeat of the Q2 2025 license queue); the 2026-10-08 China-EU consultations reached consensus on a "green channel" for rare earth and permanent magnet exports to Europe, partially hedging EU-direction risk; ③ Mandatory national energy-efficiency standards for motors (GB 18613-2020, IE3 entry floor) plus equipment renewal policy provide a long-term floor for permanent-magnet adoption in industrial motors (48,000 tonnes of domestic industrial motor NdFeB consumption in 2024, second only to NEVs); ④ The "15th Five-Year" Plan lists key specialty materials as breakthrough directions. Regulation on net benefits the first tier (quotas clearing out disorderly supply, licensing barriers rising), but for Sanhuan, with the highest export share, it is a two-way exposure.

Demand math (EV mainline and robotics option): EV mainline — China's 2025 NEV output of 16.626 million units (CAAM, +29%) × 2–3 kg NdFeB per vehicle (midpoint 2.5 kg) ≈ 42,000 tonnes; global ~20.19 million units × 2.5 kg ≈ 50,000 tonnes, nearly 30% of global demand — the largest application and the mainline of certainty. Robotics option — 3.5–4 kg of high-performance NdFeB per humanoid robot (Bernstein, 2026-10, 1.5–2× that of an EV); Tesla Optimus shipments of ~25,000 ± 10,000 units in 2026 correspond to only 88–200 tonnes (under 0.1% of global demand); a one-million-unit scenario in 2030 would be ~3,500 tonnes ≈ 1.4% of global demand and 3.5% of high-performance demand — robotics is a low-base, high-elasticity option, not the current mainline, and at the company level there are no design-win or volume supply announcements whatsoever (on the investor platform, the company itself acknowledged cooperation with Panggu Power involves "relatively small order volumes"); media-circulated claims of "entering the Optimus Gen-2 supply chain" are unconfirmed by the company. Sanhuan's 2025 sales were 12,600 tonnes; even capturing 20–25% of robot magnet share by 2030 would correspond to revenue of roughly RMB 200–300 million per year, about 4% of current revenue.

Company positioning in the industry: an established first-tier player (earliest industrial producer of sintered NdFeB in China, Chinese Academy of Sciences background, undertaken multiple national projects), third in capacity, second in market cap, first in export share (major overseas customers Bosch/TDK + mass production of N55 high grades + rare earth enrichment technology with grain boundary diffusion); but its share and profitability trends are weaker than JL MAG/Yunsheng (capacity stuck at 25,000 tonnes vs. peers' 36,000–60,000 tonne expansion, with gross margin/net margin/ROE all at the bottom). Positioning conclusion: "high-beta cyclical stock + robotics/export theme option" — large recovery elasticity when industry sentiment rises, but negative standalone alpha — with the same PrNd price increase, JL MAG earns a 21% gross margin while Sanhuan only gets 12.6%.

XI. Cycle Positioning and Through-Cycle Earnings

Historical cycle template (praseodymium-neodymium oxide, RMB 10,000/tonne): the 2010–11 speculative round surged past RMB 1 million before collapsing; the 2016–17 policy stockpiling round went 24 → 52 (about 2 years); the 2020–22 new energy round went 24 → 110 (upcycle of about 18 months, 4.6×), then drifted down for 2 years to a 2024 low of about 35; this round started with the April 2025 controls (RMB 453,000) → peaked at 90–95 in January–February 2026 (upcycle of about 10 months, ~1.1×) → now pulled back to around 74, in the 8th month of decline. Current position: approximately 59th percentile of the 2016–2026 range, −32% vs. the 2022 historical peak, +113% vs. the 2024 low, −22% vs. this round's peak — in a "post-rally high-level oscillating decline phase," neither a trough nor a top.

Supply response: the peculiarity of this round lies in the hard constraint of the quota system — 2025 quota of 300,000 tonnes REO with controlled increments, shrinking Myanmar imports (Jan–Jul 2025 REO −23.6% YoY), and absent overseas capacity in the short term, supply elasticity is significantly weaker than the 2020–22 round; but head magnet makers' expansion (JL MAG 60,000 tonnes by 2027, Zhenghai/Yunsheng 36,000 tonnes) will create a wave of high-end capacity coming online after 2027; if robot demand materializes slower than supply, processing fees will independently come under pressure — the mining side depends on quotas, the materials side on the 2027 commissioning wave; the two cycle clocks are not synchronized.

Through-cycle earnings: normalized (mid-cycle) EPS of approximately RMB 0.181 (normalized net profit midpoint of RMB 220 million, implying ROE of 3.3% — this already includes "new normal" recovery expectations; excluding the 2022 super peak, the average actual ROE over the past four years is only about 1.9%); true earnings trough EPS of RMB 0.0099 (2024 net profit of RMB 12 million, adjusted net loss of RMB 17 million; note that the price trough and the earnings trough coincide in the same year, with the stock price bottom leading by about two quarters). The current 135.6× TTM P/E is the classic "high P/E in cyclical recovery" pattern — using the current forward P/E as the sole valuation anchor commits both mirror errors: "buying at the top on low P/E" and "selling at the bottom on high P/E." Normalized valuation bands: sector option-preserving approach PE 45 ± 5× × RMB 0.181 = RMB 7.2–9.1; steady-state cyclical approach PE 20 ± 5× = RMB 2.7–4.5 (extremely conservative floor, lower bound only) — the gap between the two bands is roughly RMB 4.5/share, i.e., the entire option value the market pays for "a second cyclical reversal + robotics theme." Asset floor reference: PB 1.4× × RMB 5.43 ≈ RMB 7.6.

Downside stress test (impact on 2026E gross profit of ~RMB 910 million): PrNd −10% (to ~RMB 660,000): gross profit −15% to −26% (high-priced inventory drag for 1–2 quarters); −20% (to ~RMB 590,000): −31% to −47%, net profit approaching breakeven; −30% (to ~RMB 510,000): −46% to −68%, plus further inventory impairments (exposure: inventory of RMB 3.115 billion, impairment provision of only RMB 87 million, coverage of 3.7%; the provision reached RMB 170 million during the 2024 downturn), with adjusted net profit highly likely turning negative in a single year — already tested in 2024: the 350,000–450,000 price band corresponded to net profit of only RMB 12 million. Trough liquidity: net cash of ~RMB 966 million, cash at 3.3× short-term interest-bearing debt, extremely low capex — solvency is not at risk; what the trough kills is the income statement and valuation, not the balance sheet; the real cash leaks are minority shareholder dividends at subsidiaries (RMB 340 million in 2025) and the proposed cash acquisition (estimated at RMB 500 million–1 billion), which together would compress net cash to the RMB 200–300 million range.

Management's cycle discipline: pro-cyclical contraction type — rights-issue-funded expansion at the 2021–22 sentiment peak (followed by two years of price declines; capacity digestion relied on demand recovery in 2025–26); after project closure at end-2024, CapEx fell consecutively below depreciation, with no counter-cyclical expansion during the trough ("no blind expansion"). Verdict: prudent capital allocation, no cash-burning, but also missed the counter-cyclical positioning window — compared with JL MAG's counter-cyclical binding of Tesla/wind magnet customers and expansion to a 60,000-tonne plan during the trough, Sanhuan's "inaction" is defensively adequate but offensively absent among cyclicals.

XII. Overall Conclusion and Monitoring

Overall judgment: Neutral (avoid/watch), confidence 0.50, time horizon 6 months. The combination of mediocre quality (real assets and channels, clean balance sheet, net cash of RMB 966 million) and a relatively expensive price (baseline fair value RMB 8.8–9.9, current price 9%–22% above, safety margin of about −18%), compounded by the binary nature of the 2026-11-10 policy node, does not constitute a basis for a directional position; expected return of approximately −6%, with odds skewed to the downside but large two-way volatility. Strategically: do not chase at current prices, and do not use it as a short target either (the right-side risk of a theme revival is real); leave the decision to two verifiable nodes — the Q3 report (scheduled disclosure 2026-10-24) to check Q3 single-quarter gross margin (Q2: 12.14%) and net profit progress (first three quarters ≥ RMB 90 million as the assessment safety line), and the expiry of the controls suspension (2026-11-10) to watch extension or not and its two-way transmission to export volumes and PrNd prices.

Conditions to turn bullish (any one): share price falls below the baseline fair value range (below RMB 8.8) without the earnings trajectory being broken; magnetic materials gross margin ≥14% for two consecutive reporting periods with single-quarter adjusted net profit ≥ RMB 40 million; the acquisition lands at a reasonable price and significantly reduces minority shareholder leakage.

Conditions to turn bearish (any one): praseodymium-neodymium oxide falls below RMB 550,000/tonne for two consecutive months; controls are reinstated and overseas revenue declines for two consecutive quarters; acquisition pricing implies a price for Cixinsheng's 44% stake more than 100% above book net assets; share price falls below RMB 7.8 (entering the bear-case zone — requires re-examination of the cycle-bottom judgment rather than mechanical stop-loss).

Risk warnings and monitoring points: ① PrNd prices (SunSirs/Baichuan weekly reports; a fall below RMB 550,000 is a strong signal; direction of Baogang's 2027Q1 concentrate pricing); ② export control milestones and monthly magnet export volumes; ③ FX (forward hedging coverage below 4%; a single-quarter RMB appreciation over 2% would swallow the full year's core profit); ④ minority interest share and subsidiary dividend pace; ⑤ post-2027 peer capacity ramp-ups and processing fee trends; ⑥ company-level proof of robot magnet design wins (currently zero). On liquidity: daily turnover of ~RMB 160–300 million (0.12%–0.23% of market cap) allows easy entry/exit, but volume has shrunk ~85% from the July theme peak, raising impact costs as the theme fades.

Data in this report as of 2026-10-20; financial data from company periodic reports and announcements; price and industry data from SunSirs, Baichuan Yingfu, CCX Pengyuan, CAAM, and other public sources; valuation estimates are per this report's methodology.