The U.S. magnet buildout is mostly still a project

Only 4,000 reported units have shipment evidence

The census includes a U.S. manufacturing site or funded line when a company or government source discloses capacity. It leaves out exploration concepts, oxide projects without magnet conversion, and generic corporate targets.

I keep the announcements’ original units beside the normalized 37,584–38,048-tonne range. Ambiguous “tons” can mean metric tonnes or U.S. short tons. That uncertainty is small beside the question of when a plant can ship qualified products, but suppressing it would make the headline look more precise than its inputs.

At the cutoff, Noveon and eVAC account for the 4,000 stated-unit commercial-shipment class. MP Materials’ disclosed 3,000-tonne line and the first 600 tonnes of USA Rare Earth’s staged buildout fall into commissioning, ramp, or qualification. JS Link America, HyProMag USA, Vulcan Elements, the rest of USA Rare Earth’s project, and other disclosed firm lines remain future capacity until operating evidence supports reclassification.

A plant has to earn its capacity

The furnace, sintering line, and finishing equipment have to operate together at a repeatable rate. A ribbon cutting leaves bottlenecks, uptime, and scrap rates to be discovered.

Yield turns an installed rate into saleable output. During a difficult ramp, feedstock, labor, and working capital can flow through the plant while little leaves in a form a customer will accept.

Qualification ties the magnet to a buyer. Grade, geometry, temperature performance, coating, and traceability determine whether it can serve a motor or actuator. Approval of one product family leaves other specifications to qualify. In a safety-critical system, changing a magnet can require requalifying more than the magnet itself.

One launch customer can start shipments while exposing the plant to one program’s demand and bargaining power. Multiple qualified applications give the capacity alternative buyers. I would pay for that flexibility before paying for a larger line built around one specification.

Domestic pressing and sintering still need NdPr metal, alloy strip, dysprosium, terbium, and specialized processing. The buyer needs a qualified magnet with a dependable input chain. The U.S. building is one part of that product.

The 37,000-tonne demand figure includes embedded imports

Commerce’s roughly 37,000-tonne demand estimate includes magnets embedded in imported finished goods. The announced U.S. capacity measures magnet factories. Comparing the two totals skips the factories and purchasing decisions between them.

An imported motor, vehicle, appliance, electronics assembly, or industrial machine already carries its magnet. Selling that magnet from a U.S. plant may require relocating motor production, redesigning a component, changing a tier-one supplier, requalifying the system, or changing trade flows.

I would use direct domestic purchases to test near-term orders and utilization. Embedded imports measure a wider dependence and possible reshoring opportunity. Counting them as immediate orders gives the magnet factory credit for other factories’ future decisions.

The material flow reaches 46,400–70,459 tonnes

Finished-magnet nameplate measures a different product from oxide, metal, or alloy capacity. Strip, powder, and pressed compact describe other forms. A sintered block or machined piece also needs its product definition kept with the capacity figure.

The central project stack implies 46,400–70,459 tonnes of gross process flow under the model’s chemistry, conversion-loss, yield, machining, scrap-recovery, and recycled-feed assumptions. That is material handled through the chain to support finished output. I would compare each input disclosure at its own product stage and yield basis, without adding this flow to the 37,750-unit nameplate.

For a purchasing decision, I would keep the chemical form, original unit, saleable yield, and conversion route beside the supplier’s tonnes. That is how finished capacity acquires a matching input plan.

Grade and geometry determine which orders a plant can fill

A mass figure hides the product mix. Large blocks, thin arc segments, micro-magnets, high-speed rotor sleeves, grain-boundary-diffused grades, and complex assemblies place different demands on equipment. Machining intensity and coating area can change throughput while finished mass stays the same.

High-temperature and high-coercivity applications may need dysprosium, terbium, or a process concentrating them at grain boundaries. Their tonnage is much smaller than NdPr’s, yet they can be the constraint that leaves a motor program without a qualified domestic source.

Some buyers need coated magnets; others need magnetized segments, rotor subassemblies, or complete circuits. Moving downstream can add value, but also requires tooling, automation, quality systems, intellectual property, and customer-specific capital.

Expansion can move the bottleneck upstream

More sintering capacity needs consistent alloy strip and powder. More alloy capacity needs separated oxides and metal-making. More defense demand can tighten heavy-rare-earth additives even when NdPr is available. I would follow the added capacity upstream to the next constraint.

Customers may require traceability, exclusion of specified foreign entities, recycled content, or compliance with procurement rules. A plant able to run imported alloy may still need a domestic input route to qualify for a defense or subsidized automotive order.

Factory scrap has a known composition and a short path back into production. End-of-life magnets first need collection, dismantling, recovery, and qualification. I would place the recycled tonne in the supply model where it becomes an accepted input, after the associated costs and losses. Counting it again as new magnet demand would duplicate the same material.

The 2028 base case yields about 16,700 reported units

I apply maturity factors to classes of capacity to test how much the national result depends on execution. Commercial-shipment capacity gets the highest factor. Commissioning and qualification get less. Future projects carry the widest range.

The base case produces about 16,700 reported units, with factors of 90% for commercial-shipment capacity, 70% for ramp and qualification, and 35% for future projects. The conservative and high-execution cases appear in the table. I use them to frame the execution risk around the headline.

These scenarios keep the mixed-unit basis of the announcements. The maturity factors combine schedule, yield, qualification, and utilization limits, so another generic haircut would count some of the same risks twice. I would decompose the original factor before adding a new one.

A buyer needs a smaller model. Suppose a plant has 1,000 tonnes of nameplate, 500 tonnes of available output, 300 tonnes compatible with the qualified product and process, and firm orders for 250 tonnes. The delivered-order ceiling is 250 tonnes. Adding the four constraints would invent 2,050 tonnes of supply from the same material.

I would replace the national bucket factor with the purchase’s grade, accepted process, input availability, delivery schedule, and binding order. Those terms determine whether a motor program can buy the supply that the national scenario suggests.

The state can buy demand as well as construction

MP’s July 2025 price-protection agreement covers specified upstream NdPr products through an NdPr-equivalent kilogram calculation. Its $110-per-kilogram mechanism applies at that stage. A finished magnet still needs conversion, yield, machining, and qualification priced around it. Price-protection agreement.

The 10X agreement commits the government to magnet output on its terms and allows syndication to other buyers. MP’s filing describes production-cost purchases and facilitation payments supporting a defined EBITDA threshold after full capacity. I read this as a purchase of demand and price-risk protection, extending beyond a construction grant. Transaction filing.

The purchaser also specifies what it will accept. Sections 3, 5, and 6 provide for agreed magnet characteristics, remedies for nonconforming delivery, and funding and reporting. I would underwrite the filed allocation of risk alongside evidence of funding, conforming production, and payments. The offtake agreement supplies the contract; operating performance supplies the return.

Redundancy has value before the disruption arrives. Demanding that a new domestic plant match the incumbent’s spot price immediately can prevent the second source from being built, because public support can buy the learning, qualification, and standby capability that an individual motor buyer would underfund. I would judge that purchase against its resilience deliverables.

A protected contract can make a plant financeable before broad commercial demand appears. Its return depends on performing the agreement and surviving its conditions while improving yield and conversion cost. For a motor buyer, I would still require the accepted grade, qualified input route, and delivery schedule. National nameplate divided by demand skips both decisions.

A plant needs a product-and-order record

The project list shows where industrial effort is going. The record below connects a plant’s specification, inputs, financing, and customer orders to the delivered supply I would finance.

I would give a smaller line with repeatable yield and credible buyers more credit than a larger line with unresolved qualification and feedstock. A government contract can deliberately absorb those risks. I would value the agreement on its own terms and judge whether it produces the promised capability.

The capacity belongs to its source date

The project ledger identifies each company release, government award, filing, facility page, or technical publication. It keeps the original unit and product description with the site and phase. Target dates and operating status complete each entry.

Classification follows the strongest evidence at the August 17, 2026 cutoff. A first production run belongs with ramp and qualification until recurring shipments support the commercial-shipment class. Construction and later phases stay future capacity.

I would finance qualified supply, including a government’s deliberate purchase of resilience, before financing the national headline. The workbook preserves the reported units and calculates its scenarios from them. The next proof is repeat business in the products customers need.

The order is more specific than the tonne

I would value the announced buildout by how quickly its plants earn recurring orders for qualified products. Input origin, saleable yield, and customer demand tell a motor buyer which tonnes it can purchase. That is the supply I would finance.