The companies most exposed to China’s rare-earth export controls today are close to the inverse of the companies that will be most exposed after 11 November 2026.
Anyone positioned on today’s exposure map is positioned for a regime that may not exist in four months. The inversion turns on a single suspended measure, and the switch has a date.
Why today’s map looks the way it does
China’s rare-earth controls are usually described as having been paused in November 2025. That is not accurate, and the inaccuracy is what makes the current exposure map counter-intuitive.
The stand-down suspended six measures, including the extraterritorial re-export regime and the 0.1% de minimis rule. It did not touch Announcement 18 (2025), which placed terbium and dysprosium — with their metals, oxides, alloys and magnet materials — under export licensing on 4 April 2025. That licensing has run continuously for fifteen months and has since been expanded.
Terbium and dysprosium are what let a magnet hold coercivity at temperature. Standard NdFeB loses roughly 0.65% of its coercivity per degree Celsius, which is untenable in a robot joint under load; adding dysprosium lifts the working ceiling from around 80°C to over 220°C.
The binding constraint on robotic actuation was never paused.
But the live perimeter is narrow. MOFCOM’s September 2025 clarifications explicitly exempt motor components, assembled products, consumer goods, catalysts and phosphors. Only Tb/Dy-containing NdFeB and samarium-cobalt are controlled — general NdFeB is not.
The consequence runs against intuition: exposure attaches to magnet buyers, not product shippers. A Japanese motor maker importing dysprosium-grade NdFeB is inside the perimeter. A Chinese firm exporting a finished humanoid robot is outside it, because the assembled product is exempt.
What changes on 11 November
Among the six suspended measures is Announcement 61 — an extraterritorial rule requiring a MOFCOM licence for foreign-made products containing 0.1% or more Chinese-origin rare earths.
That rule is precisely what removes the finished-goods exemption. If it resumes, exposure stops sorting by where a company buys magnets and starts sorting by what is inside anything it ships — wherever it was made.
| Company | Today | If No. 61 resumes |
|---|---|---|
| Nidec 6594.T | Inside — imports Dy/Tb-grade NdFeB | Worse |
| Tesla TSLA | Inside as a magnet buyer | Materially worse — US-assembled units with Chinese-origin content |
| Yaskawa 6506.T | Inside — servo grades need Dy/Tb | Worse |
| Fanuc 6954.T | Inside — same | Worse |
| Sanhua 002050.SZ | Outside — domestic supply | Largely unaffected |
| Tuopu 601689.SH | Outside — same | Largely unaffected |
| UBTECH 9880.HK | Outside — assembled robots exempt | Inside. De minimis captures finished robots |
The Chinese OEM sitting most comfortably outside the perimeter today is the one a de minimis rule captures most completely.
And the US moves first. The corresponding American measure — the BIS Affiliates Rule — reimposes automatically on 10 November, one day before China’s package can resume. The two suspensions were explicitly traded for one another. Washington’s position on the 10th is an observable signal of Beijing’s on the 11th, and it is the highest-information day in the calendar.
Resumption is not automatic. This is a negotiated stand-down, not a timer, and extension is at least as plausible as snap-back. But the asymmetry is that the market is watching one date when it should be watching two, in order.
How long the constraint binds
The escape from Dy/Tb licensing is domestic magnet capacity, and it is being built — slowly.
ULVAC of Japan supplies the magnet production line globally: vacuum sintering furnaces, strip casters, hydrogen decrepitation units, jet mills. In May 2026 it established new Japan-based production for rare-earth magnet vacuum melting furnaces explicitly to diversify beyond China, and expects orders to roughly triple year on year, driven by European and North American magnet makers. Bunting runs grain boundary diffusion in the UK; MP Materials has its own process and a new campus in Texas.
So the constraint has a duration rather than being permanent. On a two-to-three year build cycle, the current dependency is a medium-term condition, not a structural one. Any analysis treating rare-earth exposure as a permanent feature of the humanoid supply chain is over-extrapolating.
The pattern worth carrying forward
Two of the six suspended measures target the upgrade path rather than the current state.
Announcement 56 controls rare-earth production equipment across 26 categories — including grain boundary diffusion equipment, the process that makes high-coercivity magnets economically viable using 70–80% less heavy rare earth.
Announcement 58 controls lithium cells at 300 Wh/kg and above. Mainstream humanoid cells run 250–300 Wh/kg today, just below the line. The industry’s stated upgrade path — semi-solid at 350–400, CATL’s all-solid-state for humanoid robots at roughly 450 — sits entirely above it.
Neither measure binds much on current production. Both would bind on the next generation. That is a consistent design, and it suggests where subsequent controls are likely to land: not on what the industry uses now, but on what it is moving to.
What we are not claiming
No company named here has reported a production disruption. This is a change in regulatory posture and its forward implications, not an observed supply failure.
Every regulatory claim above was verified against primary Chinese-language source text. Two claims drawn from secondary sources — a July move in the dysprosium price, and a widely-quoted figure for the magnet content of a Tesla Optimus unit — did not survive that check and were removed rather than softened. A third, our own initial argument that Announcement 56 closes the substitution route, was weakened by our own follow-up research and is stated above in its corrected form.
We keep a permanent record of every alert, including the ones that turn out to be wrong, and every alert receives a dated review at 90 days regardless of outcome.
USINO AI PTE. LTD. · Singapore
Supply chain research, not investment advice. USINO AI is not a financial adviser. Company names appear as supply-chain exposure analysis, not as recommendations.
Sources: Global Times · SCMP · S&P Global · Jones Day · Federal Register · Skadden