The sanctioned export gateway of the Chinese magnet cartel — owns the Tesla/wind NdFeB franchise and a heavy-RE supply backstop no Western peer can match, but the equity already prices a 60kt robot dream at ~67x earnings while the actual P&L is a 9%-margin commodity processor levered to NdPr it cannot control.
No Friday close is on the record for 300748.SZ yet. The weekly job prices the covered universe; a name it cannot price is listed as missing rather than estimated.
The verdict
The sanctioned export gateway of the Chinese magnet cartel — owns the Tesla/wind NdFeB franchise and a heavy-RE supply backstop no Western peer can match, but the equity already prices a 60kt robot dream at ~67x earnings while the actual P&L is a 9%-margin commodity processor levered to NdPr it cannot control.
JL MAG Rare-Earth (江西金力永磁) is the largest producer of high-performance sintered NdFeB permanent magnets in China and, on its own claim, the world ``. Founded 2008, headquartered in Ganzhou, Jiangxi — the heartland of China's ion-adsorption heavy rare earth deposits. It is a midstream converter, not a miner: it buys rare-earth metal/oxide feedstock (chiefly NdPr, plus heavy REs Dy/Tb) and turns it into finished magnets and magnet assemblies.
What it actually makes. Sintered NdFeB magnets across the full grade ladder, plus magnetic assemblies, injection-/compression-bonded magnets, motor rotors (a higher-value assembly step), and a nascent rare-earth recycling loop . The company controls the entire process in-house — alloy/strip casting, block ("blank") production, grain-boundary diffusion, machining, surface treatment — which lets it sell finished components rather than raw blocks .
End-markets (the demand stack). New-energy vehicle traction motors and auto parts; energy-saving variable-frequency air-conditioner (VFAC) compressors; direct-drive wind turbines; humanoid-robot and industrial-servo motors; 3C; elevators; rail transit . The 2024 mix (Lens 4) is roughly **half NEV, a quarter VFAC, ~7% wind, ~21% other** .
Marquee customers. Tesla (Model 3 RWD permanent-magnet motor — a public, order-based parts-purchase agreement signed Sept 2021), BYD, Toyota, Nidec, United Automotive Electronic Systems (UAES); supplies 8 of the top-10 vehicle A/C compressor makers and 5 of the top-10 wind-turbine OEMs ``. This is a genuine tier-1 customer franchise, not a marginal supplier.
Scale. FY2025 revenue RMB7,717.5mn (~US$1.07bn), +14.1% YoY, on 34,400 tonnes of magnetic material produced . Capacity **~40,000t in 2025, targeting 60,000t by 2027** . Market capitalisationThe share price multiplied by the number of shares. What the market says the equity is worth. **~RMB47.4bn (~US$6.6bn) ** at RMB34.66 (8 May 2026) ``.
The one-line model: a vertically-integrated magnet processor whose margin is the spread between volatile NdPr input cost and a price it largely passes through — wrapped in a Chinese-state resource and export-licence moat that no Western converter possesses.
Map upstream → JL MAG → end-customer, with named stakeholders:
Upstream (feedstock).
Midstream (JL MAG itself). Five production bases: Ganzhou (HQ), Baotou (Inner Mongolia — co-located with China's light-RE Bayan Obo complex), Ningbo (~15,000t line), plus a fifth plant in Mexico under construction (a near-shoring hedge for North-American auto customers) ``. Grain-boundary diffusion (GBD) is the key in-house process step.
Downstream (buyers). Traction-motor and compressor tiers feeding Tesla, BYD, Toyota, Nidec, UAES; wind-turbine gearbox/generator OEMs (5 of top 10); A/C compressor makers (8 of top 10); and the emerging humanoid-robot actuator supply chain (Tesla Optimus-class programmes) ``.
Chokepoints / single-source dependencies.
Names present, chokepoints marked — this lens passes.
Four moats, in descending durability:
State-backed heavy-RE resource security (the deepest). The equity link to Ganzhou Rare Earth Group gives JL MAG protected access to Jiangxi heavy REs in a world where Dy/Tb are export-controlled and structurally scarce ``. No ex-China magnet maker — and even few Chinese peers — can replicate this. It is a political/structural moat, not merely industrial.
Sanctioned export channel (the most topical). JL MAG was in the first batch of three magnet makers — with San Huan (Zhongke Sanhuan) and Yunsheng (Ningbo Yunsheng) — granted MOFCOM "general licences" to ship controlled REs to pre-cleared customers under year-long, multi-shipment permits (Dec 2025); it had already received an early medium/heavy-RE export licence in June 2025 . While most of the industry is throttled (yttrium/Dy/Tb exports ran ~50% below pre-restriction baseline ), JL MAG is a licensed gateway to Tesla/European OEMs. This is a powerful but revocable moat — its durability is a function of Beijing's posture.
Process / scale moat (genuine but commoditising). In-house grain-boundary diffusion lets JL MAG hit high-coercivity grades with 70–100% less Dy/Tb, cutting heavy-RE cost meaningfully ``. Combined with the largest sintered-NdFeB output base globally, this yields a cost and grade-breadth edge. Caveat: GBD was first developed in Japan and is now diffused across all the major Chinese houses — it is table-stakes among the top tier, not a proprietary secret.
Switching-cost / qualification moat (real, customer-specific). Magnets in a Tesla traction motor or a wind generator are designed-in and safety-/performance-qualified; re-qualifying a supplier is slow and costly. Once spec'd into a platform, JL MAG is sticky. But the Tesla agreement is explicitly order-based (Tesla "can adjust purchases according to actual needs") `` — so the lock-in is moderate, not contractual take-or-pay.
Bargaining power. Over customers: elevated right now by the export-licence scarcity (a licensed magnet is worth more than an unlicensed one). Over suppliers: weak — NdPr/Dy pricing is set by the upstream Chinese RE complex and the state, not by JL MAG. Net: the moat is real and currently widening, but it is largely granted by the Chinese state rather than owned by the company — which is the single most important thing to understand about this name.
our figuresis empty (header-only); all figures/. JL MAG reports product-end-market splits, not clean EBIT-by-segment, so operating income per segment isn/a — not disclosed at segment level.
FY2024 revenue by end-market (total RMB6,763mn) ``:
| Segment | FY24 revenue (RMB mn) | Share of total `` | Trend |
|---|---|---|---|
| NEV / auto parts | 3,314 | ~49.0% | Core engine; volume-led despite NEV sales softening |
| Variable-frequency A/C (VFAC) | 1,540 | ~22.8% | +61.8% YoY — the breakout grower `` |
| Wind power | 497 | ~7.3% | Lumpy, project-driven |
| Other (3C, robotics, servo, elevator, rail, export) | ~1,412 | ~20.9% | Includes the nascent robot-rotor line |
``
What moved and why. FY2024 revenue was flat (+1.1%) but net profit fell 48% to RMB291mn because NdPr prices collapsed and squeezed the processing spread — a vivid demonstration that this is a margin-on-feedstock business, not a unit-growth business, when prices fall ``. VFAC was the bright spot (+61.8%), reflecting China's energy-efficiency A/C mandates.
FY2025 inflection. Revenue RMB7,718mn (+14.1%), output 34,400t (+~17% vs 29.3kt blanks in 2024), and NP +142% to RMB705.6mn as NdPr stabilised then recovered and operating leverage kicked in . Net margin moved from **~4.3% (FY24) → ~9.1% (FY25)** — the whole bull case is that this margin keeps climbing as NdPr rises and robot/EV volume scales; the whole bear case is that ~9% is what a price-taking processor earns at good times.
Latest audited: FY2025 (released 25 Mar 2026) ``:
; reported gross margin around **27%** .Quarterly cadence (the recovery arc):
vs consensus / guidance. FY2025 NP landed at RMB705.6mn vs guided RMB660–760mn `` — squarely in the range, top half. No clean Street EPS-vs-actual beat/miss is sourceable for an A-share name → n/a on consensus delta.
Balance-sheet / cash flags. Management explicitly cited delivering growth "despite a YoY decline in total NEV sales and sharp short-term NdPr volatility" `` — i.e. margin, not volume, did the work. The RMB300mn "small-scale fast-track" financing authorisation and the RMB1.05bn Baotou Phase-III spend signal heavy ongoing capex — watch Free cash flowCash left after paying to run and maintain the business. Unlike profit, it is hard to flatter with accounting choices. and working capital (magnet inventory swings violently with NdPr). Hard cash-flow / net-debt figures: n/a (no filing on the shelf).
Unusual vs own history. The +142% NP rebound is real but base-effect-flattered — FY2024 was a trough. Normalised, FY2025 NP (RMB706mn) is only modestly above FY2022-era peak earnings — this is a cyclical recovery, not a step-change in earning power.
No transcripts on the shelf (
transcripts=0); sentiment is inferred from public results commentary ``.
Recurring management themes across the 2025→Q1-26 prints:
Tone shift. 2024 commentary was defensive (absorbing the NdPr crash, profit halved). Through 2025 it turned confident (record results, margin recovery, licence wins). The thing they stopped leaning on: pure NEV-volume growth (since NEV sales softened) — replaced by margin recovery + robots + export-gateway status. The risk: the robot story is doing a lot of valuation work for what is still <~5% of revenue ``.
| Company | Ticker | Layer | Mkt cap | P/E | EV/EBITDA | Note |
|---|---|---|---|---|---|---|
| JL MAG (A) | 300748.SZ | Magnet maker | ~RMB47.4bn (~US$6.6bn) `` | ~45.8x / ~67x on FY25 NP | ~76.7x `` | The subject |
| JL MAG (H) | 6680.HK | Magnet maker | (same co.) | — | — | Avg PT HK$25.33 `` |
| Zhongke Sanhuan | 000970.SZ | Magnet maker | n/a | n/a | n/a | Largest sintered-NdFeB output peer (~30–35kt) `` |
| Yantai Zhenghai | 300224.SZ | Magnet maker | ~RMB16.2bn `` | high (EPS TTM RMB0.07 ``) | n/a | 18–22kt capacity |
| DMEGC Magnetics | 002056.SZ | Magnet + ferrite | n/a | n/a | n/a | 15–20kt NdFeB |
| MP Materials | MP | US miner+magnet | n/a | n/a | n/a | DoD-backed; halted China shipments Aug 2025 `` |
| Lynas | LYC.AX | Ex-China miner | n/a | n/a | n/a | Largest non-China NdPr |
| Arafura / USA Rare Earth / Almonty / Energy Fuels | — | Developers/miners | n/a | n/a | n/a | Pre-/early-revenue; not earnings-comparable |
Read. On every sourceable metric JL MAG is the premium-valued name in the group — ~46x trailing P/E and ~77x EV/EBITDA are growth-stock multiples on a commodity-processor P&L. Yantai Zhenghai (the nearest listed magnet peer with a sourced cap) is roughly a third of JL MAG's market cap on far thinner earnings. The Western miners trade on resource/strategic-premium narratives, not comparable earnings. Bottom line: the market already awards JL MAG a large premium for its scale + licence + robot optionality — there is little "cheap" here to discover. Where multiples aren't sourceable I have left n/a rather than fabricate.
Pattern of what actually moves this stock ``:
What the tape reveals: this name trades on (a) NdPr price, (b) marquee-customer/export-licence headlines, and (c) the robot narrative — far more than on quarterly EPS. It is a policy-and-commodity beta with a robotics call option, not a steady-compounder. Macro/geopolitical headlines move it more than its own income statement.
Granular bios/comp are thin in English sources; flagged where unsourced.
. The structural flags are *category* ones: (i) **state-linked ownership** can subordinate minority interests to national policy; (ii) heavy related-party proximity to the China RE Group complex (feedstock supplier ≈ affiliated ecosystem) warrants scrutiny on transfer pricing — **not sourced as abusive, but unverifiable from English disclosure**; (iii) serial equity-raising dilutes (share count ~1.323bn (2025) → ~1.38bn (Jun 2026) ).
our figuresempty; no filings on shelf. Forensics is necessarily directional and ``-based — this is a known limitation of analysing a PRC issuer with no EDGAR trail.
Accounting-risk surface (where to look hardest):
n/a from English disclosure).Regulatory findings (required sub-section). Per regulatory/regulatory-findings.md (generated 2026-06-30): JL MAG has no CIK — no SEC EDGAR enforcement search is possible; total_sec_findings = 0 ``.
"JL Mag" (FTC OR DOJ OR FDA OR consent decree OR settlement OR fine OR penalty)): no material enforcement actions, fines, or consent decrees surfaced ``. The relevant "regulatory" reality is the inverse of enforcement — JL MAG is a beneficiary/licensee of China's MOFCOM export-control regime, not a target of it.n/a — no 10-K exists (PRC issuer, no EDGAR).Bottom-up `` off FY2025 actuals + FY2026 run-rate. No
our model createin watchlist mode (per skill). Years are fiscal (Dec).
Anchors: FY2025 NP RMB705.6mn; Q1-26 NP RMB192.8mn (+20% reported, +66% core); shares ~1.38bn → EPS FY25 ≈ RMB0.51 . Q1-26 EPS RMB0.14 .
Swing variables: (1) NdPr price (sets spread); (2) volume ramp toward 60kt by 2027; (3) robot-rotor mix (margin-accretive if it scales); (4) operating leverage as utilisation rises; (5) dilution from ongoing raises.
| Path | FY2026E NP | FY2027E NP | FY2028E NP | EPS FY28E `` | Logic |
|---|---|---|---|---|---|
| Bear | ~RMB650mn | ~RMB600mn | ~RMB650mn | ~RMB0.45 | NdPr round-trips lower; spread compresses; volume grows but margin gives it back (the FY24 lesson) |
| Base | ~RMB900mn | ~RMB1.15bn | ~RMB1.45bn | ~RMB1.00 | NdPr holds elevated; 60kt ramp on track; robots reach mid-single-digit % rev; margin ~10–11% |
| Bull | ~RMB1.2bn | ~RMB1.8bn | ~RMB2.6bn | ~RMB1.80 | NdPr stays high on Western-supply detour; robot rotors inflect to a real segment; margin →13–14% on mix + leverage |
``
Reality check on valuation. At ~RMB47bn cap, even the bull FY28 EPS ~RMB1.80 implies ~19x P/E three years out; the base ~RMB1.00 implies ~35x FY28. So the stock is discounting the bull volume+robot case landing in full. This is not a cheap stock on any honest forward path — the multiple is the bet. Base-case Brier forecast (not logged in watchlist): "300748.SZ FY27 net profit ≥ RMB1.15bn", p≈0.45.
Bull case. JL MAG is the single best-positioned name to monetise the rare-earth weaponisation of the 2020s. It owns: the largest global sintered-NdFeB capacity; a state-protected heavy-RE feedstock backstop (Ganzhou RE Group); a sanctioned MOFCOM export licence that lets it sell to Tesla and European OEMs while rivals are throttled; and a call option on humanoid robots (~3.5kg NdFeB per robot × a market scaling to tens of thousands of units ``). NdPr is in a structural up-cycle (Western supply detour, electrification demand > ex-China supply). FY2025's +142% NP shows the operating leverage when the spread turns. If robots inflect and NdPr stays high, earnings could ~3x by 2028 (bull). It is, functionally, the publicly-tradeable equity expression of "China controls the magnets."
Bear case (2–3 potentially-permanent impairments).
Pre-mortem (18 months out, thesis broke). NdPr round-tripped after the 2025 spike; China widened general licences so JL MAG's gateway premium evaporated; humanoid-robot volumes slipped right (Tesla Optimus redesigns/delays ``) so the rotor segment stayed immaterial; margins drifted back toward mid-single digits; and the ~46x multiple compressed to ~20x — a 50%+ de-rate even with revenue still growing. The stock fell not because the business broke, but because the premium did.
Multiples too high? Yes, on any earnings-based view. The only frame in which they're defensible is "strategic-scarcity asset / robot-supply-chain pure-play," which is a narrative multiple.
Contrarian view (what the market refuses to see). Bulls treat the export licence as an unalloyed moat. The contrarian read: being the Chinese state's chosen export conduit is a double-edged sword — it makes JL MAG indispensable today but also a policy instrument, fully exposed to being throttled, taxed, or out-competed-by-decree the moment Beijing's incentives shift. The market is pricing the upside of state favour while underpricing the fragility of depending on it.
Dismantling the bull case:
| Industry | Critical Materials |
| Size | Public Company |
Where JL MAG Rare-Earth sits against the other names we cover on this beat. Each line is that company’s verdict, not a summary of it.
The bear case fired — but through a door the prior dossier wasn't watching, and the stock overshot it.
Cash $1.4B
The bear case got tested on its own terms and lost
Cash $7.3B
The June thesis inverted on the arithmetic, not the story
Cash $568M
A fully-funded, government-backed option on the first US commercial-scale lithium mine
A single-asset, equity-accounted 44.8% call option on Cauchari-Olaroz