Skip to content
Peculiar Materials
Tools · Beta

What does this magnet cost, and where?

A landed cost per kilogram for a finished sintered NdFeB magnet, from five places at once, broken into the pieces that actually move when you change where you buy.

Short versionA landed cost per kilogram for a finished sintered NdFeB magnet from five places — Europe appears twice, as Estonia and as Germany, because they share a duty lane and differ by a factor of two and a half on labor. Broken into material, processing, tooling, logistics and duty. Built to show you where the money sits and how it moves when you change where you buy. It is a beta, the numbers are modeled from public data rather than anybody's books, and it does not calculate margin — so treat the comparison as the output and the absolute figures as indicative.

What this tool skips, so nobody mistakes it for complete. Supplier margin and overhead. Payment terms, tooling ownership, minimum order quantities, and every other commercial term that moves a real quote. Quality cost — scrap at your incoming inspection, not the plant's. Currency hedging, and currency itself: everything here is US dollars. Inventory carrying cost and the working capital tied up in a ninety-day ocean lane. Export licensing lead time, which is flagged where it applies but not priced. Anything about SmCo, ferrite or bonded magnets: this is sintered NdFeB only.

Beta. This tool compares regions. It does not price your supplier.

The part

Finished dimensions after coating. The yield engine runs in the background to work out how much block each part actually consumes.

The lane

Where it lands, how it travels, and the one trade-policy question the model cannot settle from outside CBP.

Landed cost by source

 

Material Processing Tooling and NRE Logistics Duty and fees
Beta. This tool compares regions. It does not price your supplier.

 

What this is, and what it is not

This estimator is in beta. It is built to show the shape of a sourcing decision: how much of a landed cost is the metal, how much is the processing, how much is tooling and NRE, and how much is freight and trade policy. The comparison between regions is the output worth trusting. That is what the model is good at, and the direction and rough magnitude of the gaps are defensible.

The absolute numbers are a different matter. They are modeled from public data, published rates, government filings and one audited producer disclosure. No individual supplier's cost structure has been validated against this model, and no supplier's quote was used to calibrate it. Real costs turn on plant age, utilization, contracted power, yield on your specific geometry, and a dozen commercial terms that are not public and never will be.

Margins are not calculated. There is no supplier overhead line and no profit line anywhere in this tool. Everything it shows you is cost. What a supplier adds on top is a commercial question, and it is yours to answer.

So do not take a number from this page into a negotiation. It will not survive contact with someone who knows their own cost base, and trying will cost you more credibility than the number is worth. Use it to work out which questions to ask, and which region deserves the RFQ.

Where the numbers come from

Every figure is a default you can overwrite.

  1. The four regional conversion costs do not share a method, and you should know which is which. The Chinese figure is derived from one audited producer filing and is the best-evidenced number in the model. The European figure is a bottom-up build that an independent top-down happens to corroborate. The Japanese figure has no producer disclosure behind it at all and is interpolated between its neighbors. The American figure is anchored on announced greenfield capex, whose high end is a subsidized first-of-a-kind plant with a wider process scope than this boundary. One assumption runs through all four: the division of plant cost between block-making and machining is taken as 60/40, which is an estimate rather than a source. At 50/50 the Chinese figure falls from $3.10/kg to $2.58. A number this load-bearing and this thinly sourced is exactly the kind a reader should push back on.
  2. Material is priced on the block a part consumes, less the swarf that goes back. A finished 30×15×4 rectangle consumes 1.49× its own mass in block, so pricing finished mass understates material by about a third. But grossing the raw material up by the machining yield and stopping there overstates it by roughly 1.7× against the one audited direct-materials line available. Both a feedstock discount and a swarf credit are needed to reconcile the two, and both are fitted to a single company's filing rather than measured. They are editable for that reason.
  3. Japan and the EU are combined-rate caps, not additive layers. If the column 1 rate is already at or above the threshold there is no additional duty; if below, the combined rate is set to the threshold. Japan's base is 0% under the US–Japan Trade Agreement, so Japan lands at 12.5% flat. The EU's base is 2.1%, so the EU lands at 10% flat — an effective adder of 7.9%, not 10%.
  4. The largest unresolved input is whether 8505.11 sits on the Section 301 forced-labor exclusion annex. The evidence that it does not is strong but indirect: a USITC-synced tariff service resolves the heading by applying the forced-labor heading rather than an exemption, no published summary of the annex mentions 8505, and USTR expressly rejected pleas to spare magnets in the 2024 four-year review. The default is 39.6%; the toggle is there because the evidence is indirect.
  5. A correction worth carrying. An article circulating in the magnet trade says 8505.11.00.70 is annex-excluded. It refers to the IEEPA reciprocal Annex II, a different instrument under an authority struck down in February 2026, and the same article asserts 8505.11 is SmCo and 8505.19 is NdFeB, contradicting both itself and USTR. It is wrong on both counts.
  6. No carrier or forwarder publishes a magnetized-material dangerous-goods surcharge. That is a researched negative rather than a gap, so the surcharge here is an honest $0 with the real cost carried where it actually falls: shielding weight, at a 1.30× multiplier, plus an origin inspection fee. The UN 2807 classification floor is 0.159 A/m at 2.1 m and the carrier acceptance ceiling is 0.418 A/m at 4.6 m; a widely cited compliance page presents the ceiling as though it were the floor.
  7. Three rates have a known change date inside seven months. The US merchandise processing fee re-indexes 1 October 2026, the UK network-charge compensation uplift lands 1 April 2026, and China's suspended October 2025 escalation expires in November 2026. A Section 232 critical-minerals report also came due in July 2026 with an unknown outcome, and a minimum-import-price remedy is expressly reserved — that would not behave like an ad valorem duty and would break a percentage model outright.
  8. Import VAT is excluded in the EU and UK as recoverable, which is right for a VAT-registered owner-importer. Recovery attaches to the owner, so under DDP or consignment it becomes a real sunk cost.
  9. Energy is real but small. The whole China-to-Germany spread is about $1.20/kg, an order of magnitude below the feedstock line, so it is not given a slider. Worth knowing anyway: Texas beats coastal China on industrial power.

What this page does with your data

The page carries a security policy that lets it contact this site and one other address, and the browser enforces it. It loads its calculating engine from this site when the page opens, and it asks api.frankfurter.dev for the European Central Bank's daily reference rates so it can show you a total in your own currency. That request is a plain read: it asks for a table of exchange rates and sends nothing about you or your part — no dimensions, no quantity, no rates you have edited. If it fails, the page falls back to rates baked in on the date below and says so on screen. After that, open your browser's developer tools and watch the Network tab: while you work, you'll see nothing, and each run adds a single empty request to /api/route-run. What is recorded, and for how long.