Why fixed magnet pricing breaks
Rare-earth inputs have moved violently enough to make annual fixed pricing structurally unstable. In the first half of 2026, benchmark neodymium pricing in China rose to roughly 1,025,000 yuan per metric ton by mid-July, more than 80% above the same point a year earlier. A twelve-month fixed price agreed before that move is not a good deal for the buyer; it is a deal the supplier will try to reopen, honor badly, or absorb by quietly deprioritizing your orders.
That last outcome is the one commodity managers underestimate. A fixed price the supplier is losing money on does not stay a fixed price. It becomes longer lead times, allocation to better-paying customers, and a renegotiation conversation at the worst possible moment. Contractual certainty that the counterparty cannot afford is not certainty.
Indexation flips the problem. Instead of guessing the market, both parties agree in advance on how price responds to a published number. Volatility does not disappear, but it becomes calculable, and it stops being a matter of goodwill.
It is not a cost-plus open book, and it does not entitle a supplier to pass through everything. A well-written clause moves only the rare-earth material component and leaves conversion, margin, freight and overhead exactly where they were negotiated.
Choosing what to index
You can index at four points in the chain, and the choice determines how much basis risk each side carries. Basis risk is the gap between the thing you indexed and the thing you actually buy.
| Index point | Transparency | Basis risk to buyer | Best used when |
|---|---|---|---|
| Rare-earth oxide | Highest — widely published | Highest — excludes reduction, alloying, energy | You want a clean, auditable reference and will negotiate the rest separately |
| Rare-earth metal | Good | Moderate | Usually the best balance; closest published point to the supplier's real buy |
| NdFeB alloy | Limited, often supplier-quoted | Low | You trust the supplier's disclosure and want minimal argument |
| Finished magnet | Emerging | Lowest | The part is standard enough that a published magnet assessment genuinely reflects it |
The finished-magnet option became materially more practical in April 2026, when Benchmark Mineral Intelligence introduced regional rare-earth permanent-magnet price assessments covering China, North America, Europe, Japan and South Korea. That is a meaningful development: for the first time, contracts can reference a finished-magnet benchmark rather than reasoning from oxide prices and hoping the relationship holds.
For most portfolios, metal-level indexing on a two-element basket — NdPr plus whichever heavy rare earth the grade actually contains — is the right default. It is published, it maps closely to the supplier's purchasing, and it correctly distinguishes a standard N-series part from an EH-grade part carrying several percent dysprosium.
If you run standard grades alongside high-coercivity grades under a single NdPr index, the high-coercivity parts will be systematically mispriced in both directions. Index the basket that each part actually consumes, or accept that you are cross-subsidizing between part families.
Writing the adjustment formula
Two forms are in common use. They are not equivalent, and the difference is worth arguing about.
Proportional form
// k = rare-earth material share of the base price, e.g. 0.55
Simple to write and easy to audit, but it depends on k staying accurate. As the index moves, the true material share moves with it, so a fixed k drifts — under-recovering for the supplier on the way up and over-recovering on the way down.
Additive form
// m = kg of element i consumed per finished piece, on a buy-weight basis
// η = pass-through efficiency, typically 0.90 to 1.00
This is the stronger clause. It moves price by the actual dollar change in material consumed, so it stays correct at any index level and cannot over-recover. It also forces a useful disclosure: the supplier must state kilograms of each element per piece, which is exactly the input your should-cost model needs.
Insist that m is on a buy-weight basis, including machining loss net of scrap recovery. A supplier quoting finished mass is understating consumption; a supplier quoting gross blank mass with no scrap credit is overstating it. Both should be written down and both should be auditable.
The η term is where efficiency incentives live. Setting it slightly below one means the supplier absorbs a small share of every increase, which preserves their motivation to buy well and to improve yield rather than passing everything through.
Reference window, lag and reset cadence
Three timing parameters do most of the work in a real clause, and they are where inexperienced drafting causes the most damage.
Averaging is the single most valuable provision. A spot reference invites both parties to game the reference date and exposes the price to one anomalous print. A quarterly average of monthly settlements is nearly impossible to manipulate and produces a price that behaves like the market rather than like a snapshot of it.
Lag deserves more thought than it usually gets. Suppliers buy material weeks or months ahead of the production they will use it for. A clause with no lag charges you today's index for material bought at last quarter's price on the way up — and hands you an undeserved windfall on the way down. Matching the lag to the supplier's actual procurement rhythm makes the clause fair in both directions, which is what makes it survive a downturn.
Cadence is a trade-off between accuracy and planning. Monthly resets track cost closely but make budgeting miserable and generate administrative overhead on every release. Quarterly is the practical default for most magnet programs and pairs naturally with the release schedule under a blanket order or VMI arrangement.
Collars, caps, floors and reopeners
Indexation without limits transfers all volatility to the buyer's P&L. The controls below shape the distribution of outcomes, and each has a price.
| Device | What it does | Typical setting | Trade-off |
|---|---|---|---|
| Deadband | No adjustment unless the index moves beyond a threshold | ±3–5% | Removes administrative noise; almost always worth including |
| Cap | Limits total increase within a period | 10–15% per year | Supplier will price the risk in, or seek a matching floor |
| Floor | Limits total decrease | Mirror of the cap | The honest counterpart to a cap; expect to concede one for the other |
| Collar | Cap and floor together | Symmetric band | The most negotiable structure; both sides give up tails |
| Sharing band | Splits movement beyond a threshold | 50/50 above 15% | Keeps the supplier engaged in extreme scenarios rather than defaulting |
| Reopener | Triggers renegotiation on a defined event | Index moves >25%, or a licensing event | Essential escape valve; must be drafted narrowly to avoid abuse |
Reopeners are the clause most worth drafting carefully, because magnet supply carries risks that no price index captures. Export licensing, origin restrictions and allocation events can make a contract impossible to perform at any price. A reopener that names those triggers explicitly — rather than relying on a general force majeure clause — gives both sides a defined path instead of a dispute.
Before signing, run the formula backwards. If the index falls 30%, does your price fall? By how much, and after what delay? A clause that rises quickly and falls slowly is common, rarely deliberate, and almost always accepted because nobody modeled the downside.
Regional benchmarks and the end of one world price
For most of the industry's history, a single Chinese reference was a reasonable proxy for global magnet pricing. That assumption is now breaking down, and contracts written on it are quietly acquiring basis risk.
As non-Chinese capacity comes online — supported by government investment in domestic refining, magnet manufacturing and recycling — regional prices are diverging. A magnet produced in North America or Europe carries different energy costs, different labor, different amortization on new plant, and a different demand profile. It should not be expected to follow a Chinese index, and increasingly it does not. The introduction of regional magnet assessments in 2026 formalized what buyers were already observing.
That last point is routinely missed. Indexing a dollar contract to a yuan-quoted benchmark means every adjustment carries an exchange-rate move that has nothing to do with rare earths. State the conversion method and the rate source in the clause, or convert the index to your contract currency using a defined averaging method before applying the formula.
The clause checklist
Before an index-linked magnet agreement goes to signature, every item below should have an unambiguous answer written into the document rather than assumed.
That final row matters more than the rest combined. Indexation solves price. It does nothing for availability, and a perfectly drafted formula is worthless if the supplier cannot obtain an export license or has allocated your capacity elsewhere. Price mechanics belong alongside supply-risk monitoring, a qualified second source, and inventory cover sized to the real qualification timeline.
None of this argues that fixed prices are always wrong. For stocked standard parts at moderate volume, a distributor holding inventory can often offer a genuine fixed rate for up to a year because the inventory position is already taken. Indexation earns its complexity on contracted custom production, where the material has not yet been bought.