[Back to writing](https://sulayman-bowles.dev/writing)

[Industrial systems](https://sulayman-bowles.dev/topics/industrial-capacity)

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

Plants with disclosed commercial shipments account for 4,000 of 37,750 announced capacity units. I would finance the product and its customers before the national headline.

[Sulayman Bowles](https://sulayman-bowles.dev/about)

Aug 17, 202612 min read23 sourcesUpdated Oct 27, 2026

Sources & data ↓

![Monochrome generative point study “Arc Current,” a mathematical form rendered in black and white.](https://sulayman-bowles.dev/images/generative-art/ink/yuru-20-bad7d4f3f05a.webp)

[@yuruyurau ↗](https://x.com/yuruyurau/status/2024489280064852228)

About four-fifths of the announced U.S. magnet capacity in this August 17, 2026 census still depends on construction, later phases, or production targets. I would give more weight to a smaller line with qualified products and credible customers than to a larger line still looking for its input chain and first orders.

The project list adds up to 37,750 reported capacity units. Announcements mix metric tonnes and U.S. short tons; normalization gives 37,584–38,048 metric tonnes. Just 4,000 reported units belong to plants with disclosed commercial shipments. Another 3,600 are commissioning, ramping, or qualifying, while 30,150 remain future capacity.

There is a reason to fund that ambition before ordinary buyers would. A second source has value when the cheapest incumbent can be disrupted. MP Materials’ filed government agreements show how price support and offtake can purchase that resilience. I would judge the return by the qualified magnets those contracts produce.

**Method & limits**

The capacity census and 2028 scenarios use the August 17, 2026 cutoff, original units, and disclosed operating classifications. Scenario factors and process yields are invented for illustration. MP’s July 2025 government price-protection, offtake, and transaction filings were checked October 25, 2026. Plant output and current contractual compliance weren’t refreshed; filed schedules contain redactions. This is analysis, not investment advice.

## 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.

U.S. NdFeB capacity stack used in this investigation

| Maturity class | Capacity | What the class requires | What it does not prove |
| --- | --- | --- | --- |
| Commercial shipments | 4,000 stated units | Disclosed nameplate plus evidence of commercial shipments | Full utilization, broad qualification, domestic inputs, or positive margins |
| Commissioning / ramp / qualification | 3,600 stated units | Installed or starting equipment with disclosed commissioning, first output, or customer qualification | Steady-state yield, recurring orders, or interchangeable product coverage |
| Future firm projects | 30,150 stated units | Named U.S. project or funded line with disclosed capacity | Completed construction, on-time startup, financing sufficiency, or customer acceptance |
| Firm-project total | 37,750 stated units | Arithmetic sum of the three maturity classes | Current domestic production |

U.S. rare-earth magnet capacity by maturity

A stacked bar divides 37,750 stated units of disclosed firm-project capacity into 4,000 with commercial shipments, 3,600 in commissioning, ramp, or qualification, and 30,150 in future projects. Mixed source units normalize to 37,584–38,048 metric tonnes.

FIGURE 01 / PHYSICAL-CAPACITY AUDIT

### 37,750 stated units is not current supply

Mixed source units; normalized total: 37,584–38,048 metric tonnes

**SHIPPINGCommercial shipments:4,000**

Disclosed nameplate plus public shipment evidence. Not full utilization.

**RAMP / QUAL.Commissioning / qualification:3,600**

Installed or starting lines without broad recurring shipment evidence.

**FUTURE FIRM PROJECTSFuture projects:30,150**

Construction, funded expansions, and later phases.

80% of the project stack

Evidence cutoff: 2026-08-17 · Capacity ≠ qualified output · Source: project ledger and cited primary disclosures

Eighty percent of the disclosed firm-project stack remains in the future-project class. Nameplate should not be read as current qualified output.[Source figure ↗](https://sulayman-bowles.dev/images/research/rare-earth-magnet-capacity-maturity.svg)

## 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.

Evidence ladder for classifying magnet capacity

| Gate | Strong evidence | Weak substitute |
| --- | --- | --- |
| Physical completion | Integrated line commissioned; equipment and site identified | Groundbreaking or building shell |
| Repeatable yield | Sustained saleable output, scrap data, or recurring production cadence | First furnace run or sample batch |
| Qualification | Customer-approved grade, geometry, coating, and process window | Laboratory magnetic-property result |
| Commercial depth | Recurring shipments to more than one program or product family | Non-binding memorandum or total addressable market |
| Input security | Origin and conversion path for oxide, metal, alloy, heavy rare earths, and recycled feed | Final assembly in the United States |

## 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 demand behind each utilization case

| Denominator | What it includes | Use in a capacity model |
| --- | --- | --- |
| Direct magnet demand | Magnets purchased by domestic motor, component, and system manufacturers | Near-term addressable orders and utilization |
| Embedded-import magnet content | Magnets inside imported finished goods and subassemblies | Strategic exposure and possible long-run reshoring |
| Replacement / aftermarket demand | Maintenance, repair, spares, and remanufacturing | Often fragmented; grade and geometry specific |
| Export demand | Qualified foreign customers served from U.S. plants | Potential utilization, subject to cost, logistics, and trade rules |

Demand denominators for U.S. rare-earth magnets

The diagram separates direct domestic magnet purchases, magnets embedded in imported finished goods, replacement demand, and export demand. Only direct domestic purchases are the near-term order book immediately contestable by a new U.S. magnet factory.

FIGURE 02 / DEMAND BOUNDARY

### “U.S. demand” is not one addressable market

Strategic exposure is broader than the orders immediately available to a domestic magnet plant

NEAR-TERM CONTESTABLE

#### Direct domestic magnet demand

Magnets purchased by U.S. motor, component, and system manufacturers.

not automatic

STRATEGIC EXPOSURE / LONG-RUN RESHORING

#### Magnets embedded in imported finished goods

Capturing this volume may require relocating motor production, redesigning components, or requalifying a complete system.

#### Replacement / aftermarket demand

Fragmented, geometry-specific, and often tied to installed equipment.

#### Export demand

Potential utilization subject to qualification, cost, logistics, and trade rules.

Do not divide announced domestic nameplate by an embedded-import denominator and call the supply gap closed.

The broad strategic exposure is larger than the order book immediately available to a domestic magnet plant.[Source figure ↗](https://sulayman-bowles.dev/images/research/rare-earth-magnet-demand-denominators.svg)

## 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.

Material-flow interpretation used in the workbook

| Measure | Modeled range | Interpretation |
| --- | --- | --- |
| Finished magnet nameplate | 37,584–38,048 t | Normalized range around the 37,750 stated-unit project sum |
| Gross process flow | 46,400–70,459 t | Cumulative material handled before saleable finished output under modeled yields and recycle loops |
| Commercial-shipment class | 4,000 stated units | Nameplate attached to plants with disclosed commercial shipments |
| Qualified output | Not publicly disclosed as a complete national total | Must be measured by customer-approved products and sustained shipments |

Rare-earth magnet finished capacity versus gross process flow

A process diagram traces oxide and metal through alloy strip, powder, pressing, sintering, machining, coating, internal recycling, and 37,584 to 38,048 tonnes of normalized finished magnet nameplate. Gross process flow is modeled at 46,400 to 70,459 tonnes.

FIGURE 03 / MATERIAL BALANCE

### Gross flow is not a second capacity total

Every disclosure must preserve its chemical form, process stage, unit, and yield boundary

1.

#### Oxide / metal

NdPr + additives

origin and purity matter

2.

#### Alloy / strip

chemistry + casting

oxygen control

3.

#### Powder / press

milling + orientation

handling loss

4.

#### Sinter / machine

shrinkage + cracking

geometry loss

5.

#### Finished magnets

**37.6–38.0 kt**

normalized nameplate

#### Internal scrap recovery

reduces virgin feed but does not erase process loss

Modeled gross process flow: **46.4–70.5 kt**

throughput requirement across stages · not additive to finished capacity

The 46,400–70,459-tonne range is a throughput requirement, not a second capacity total.[Source figure ↗](https://sulayman-bowles.dev/images/research/rare-earth-magnet-material-flow.svg)

## 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.

Where a domestic chain can still fail

| Stage | Failure mode | Observable evidence |
| --- | --- | --- |
| Separation | Insufficient NdPr purity or missing heavy-rare-earth products | Qualified oxide specifications and recurring deliveries |
| Metal / alloy | Scale, consistency, or oxygen-control problems | Lot consistency, strip-cast output, customer acceptance |
| Powder / sintering | Low yield, oxidation, cracking, or unstable magnetic properties | Saleable yield and process-capability data |
| Machining / coating | Geometry bottlenecks, corrosion failures, or high scrap | Dimensional yield and coating qualification |
| Assembly | Customer-specific tooling or rotor integration delays | Production-part approval and serial shipments |
| Recycling | Collection and purification cost exceeds recovered value | Closed-loop contracts and qualified recycled feed |

## 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.

Illustrative maturity-adjusted output scenarios

| Case | Commercial shipments | Ramp / qualification | Future projects | Illustrative output |
| --- | --- | --- | --- | --- |
| Conservative | 85% | 45% | 10% | ~8,000 stated units |
| Base illustration | 90% | 70% | 35% | ~16,700 stated units |
| High execution | 95% | 90% | 70% | ~28,100 stated units |
| Headline nameplate | 100% | 100% | 100% | 37,750 stated units |

## 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.

Recommended fields for each U.S. magnet project

| Field | Question answered |
| --- | --- |
| Nameplate and unit | What finished product mass is the line designed to make? |
| Product form | Sintered, bonded, block, segment, assembly, or rotor? |
| Operating evidence | Construction, commissioning, first output, qualification, or recurring shipment? |
| Qualified capacity | How much output is approved for named product families? |
| Utilization | What share of nameplate is producing saleable product? |
| Saleable yield | How much input becomes accepted finished magnets? |
| Grade coverage | Which magnetic and temperature-performance grades are qualified? |
| Input origin | Where do oxide, metal, alloy, heavy rare earths, and recycled feed originate? |
| Customer concentration | How much output depends on one customer or program? |
| Offtake quality | Binding volume, price floor, indexation, and duration? |
| Schedule confidence | What critical equipment, permits, financing, or qualification steps remain? |
| Policy dependence | Which grants, loans, tax credits, procurement rules, or price supports are required? |

## 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.

## Sources

1. [BIS: NdFeB magnet Section 232 report (June 2022; Federal Register publication February 14, 2023)](https://www.govinfo.gov/content/pkg/FR-2023-02-14/html/2023-03078.htm)Verified Aug 17, 2026

Access rechecked October 19, 2026. BIS’s June 2022 investigation is reproduced in the February 14, 2023 Federal Register. The original press-release URL is unavailable.

2. [U.S. Geological Survey—Mineral Commodity Summaries: Rare Earths](https://www.usgs.gov/centers/national-minerals-information-center/rare-earths-statistics-and-information)Verified Aug 17, 2026
3. [U.S. Department of Energy: 2023 Critical Materials Assessment](https://www.energy.gov/cmei/ammto/articles/2023-doe-critical-materials-assessment)Verified Aug 17, 2026

Access checked October 19, 2026. DOE’s August 1, 2023 assessment links the technical report; plant shipment data come from the dated project sources.

4. [U.S. Department of Defense: MP Materials rare-earth supply-chain agreement](https://www.defense.gov/News/Releases/Release/Article/4232241/dod-signs-historic-agreement-with-mp-materials-to-strengthen-us-rare-earths-supply-chain/)Verified Aug 17, 2026
5. [MP Materials: commercial production of NdFeB magnets in Fort Worth](https://mpmaterials.com/)Verified Aug 17, 2026
6. [MP Materials: investor filings and manufacturing disclosures](https://investors.mpmaterials.com/financials/sec-filings)Verified Aug 17, 2026
7. [Noveon Magnetics: U.S. magnet manufacturing and recycling platform](https://noveon.co/)Verified Aug 17, 2026
8. [Vacuumschmelze / eVAC: U.S. permanent-magnet manufacturing](https://vacuumschmelze.com/)Verified Aug 17, 2026
9. [USA Rare Earth: Stillwater, Oklahoma magnet facility](https://www.usare.com/)Verified Aug 17, 2026
10. [Vulcan Elements: U.S. rare-earth magnet manufacturing](https://vulcanelements.com/)Verified Aug 17, 2026
11. [JS Link America: U.S. permanent-magnet project](https://www.jslink.com/)Verified Aug 17, 2026
12. [CoTec: HyProMag USA feasibility results (November 25, 2024)](https://www.cotec.ca/news/hypromag-usa-feasibility-study-demonstrates-robust-economics-and-the-opportunity-to-develop-a-major-new-domestic-source-of-recycled-rare-earths-magnets-for-the-united-states)Verified Aug 17, 2026

Access checked October 19, 2026. November 25, 2024 feasibility plan. The capacity census keeps its August 17, 2026 cutoff.

13. [General Motors: domestic rare-earth magnet supply initiatives](https://news.gm.com/)Verified Aug 17, 2026
14. [International Energy Agency: Global Critical Minerals Outlook](https://www.iea.org/reports/global-critical-minerals-outlook-2025)Verified Aug 17, 2026
15. [Government Accountability Office: critical-mineral supply-chain work](https://www.gao.gov/critical-minerals)Verified Aug 17, 2026
16. [National Renewable Energy Laboratory: rare-earth magnet supply-chain research](https://www.nrel.gov/materials/critical-materials.html)Verified Aug 17, 2026
17. [Ames National Laboratory: Critical Materials Institute](https://www.ameslab.gov/cmi)Verified Aug 17, 2026
18. [European Commission JRC: rare-earth permanent magnet material-flow research](https://joint-research-centre.ec.europa.eu/)Verified Aug 17, 2026
19. [U.S. Department of Defense: Industrial Base Policy critical-minerals programs](https://www.businessdefense.gov/ibr/mceip/index.html)Verified Aug 17, 2026
20. [Federal Register: neodymium-iron-boron permanent magnet trade and security actions](https://www.federalregister.gov/)Verified Aug 17, 2026
21. [MP Materials–DoD Price Protection Agreement](https://www.sec.gov/Archives/edgar/data/1801368/000119312525157310/d43796dex106.htm)MP Materials / SEC Exhibit 10.6 · 2025-07-09Verified Oct 25, 2026

July 9, 2025 agreement: the NdPr-equivalent mechanism covers specified NdPr products. Filed schedules contain redactions; finished-magnet economics include subsequent conversion.

22. [MP 10X–DoD magnet Offtake Agreement](https://www.sec.gov/Archives/edgar/data/1801368/000119312525157310/d43796dex105.htm)MP 10X Development / SEC Exhibit 10.5 · 2025-07-09Verified Oct 25, 2026

Sections 3, 5, and 6 cover offtake, quality, acceptance, funding, and reporting. Performance, appropriations, and completion require later evidence.

23. [MP Materials July 10, 2025 transaction report](https://www.sec.gov/Archives/edgar/data/1801368/000119312525157310/d43796d8k.htm)MP Materials / SEC Form 8-K · 2025-07-10Verified Oct 25, 2026

Issuer summary of the upstream price protection, financing, and 10X magnet offtake agreed in July 2025.

## Supporting material

- [Capacity and maturity model (CSV file)](https://sulayman-bowles.dev/research/us-rare-earth-magnet-capacity-model.csv)The machine-readable maturity-bucket model, normalization range and illustrative qualified-output scenarios behind the analysis.
- [Methodology and definitions (MD file)](https://sulayman-bowles.dev/research/us-rare-earth-magnet-methodology.md)Definitions for project nameplate, commercial shipment, customer qualification, future capacity, gross process flow, and denominator handling.
- [Source ledger (CSV file)](https://sulayman-bowles.dev/research/us-rare-earth-magnet-source-ledger.csv)Primary-source index for the companies, government programs, technical references, and demand estimates used in the audit.

## When a source link won’t open

Links checked Oct 19, 2026.

**Commerce: NdFeB Section 232 investigation**

BIS’s June 2022 investigation is reproduced in the February 14, 2023 Federal Register. The original press-release URL is unavailable.

- [BIS: NdFeB magnet Section 232 report (June 2022; Federal Register publication February 14, 2023) ↗](https://www.govinfo.gov/content/pkg/FR-2023-02-14/html/2023-03078.htm)

The original link was unavailable when checked.

**CoTec: HyProMag USA**

This November 25, 2024 feasibility release describes a historical project plan. Later capacity updates exist; the article’s dated capacity census has not been refreshed.

- [CoTec: HyProMag USA feasibility results (November 25, 2024) ↗](https://www.cotec.ca/news/hypromag-usa-feasibility-study-demonstrates-robust-economics-and-the-opportunity-to-develop-a-major-new-domestic-source-of-recycled-rare-earths-magnets-for-the-united-states)

The original link was unavailable when checked.

**Department of Energy: materials assessment**

DOE’s August 1, 2023 assessment page links the report. It supplies technical context, not current plant shipment data.

- [U.S. Department of Energy: 2023 Critical Materials Assessment ↗](https://www.energy.gov/cmei/ammto/articles/2023-doe-critical-materials-assessment)

The original link was unavailable when checked.

**GAO: critical minerals**

The generic topic URL returned 404. GAO-24-106395, published July 31, 2024, discusses recovery from nontraditional sources. It provides related background, rather than an archive of the missing topic page or an update to the capacity inventory.

- [GAO: July 31, 2024 critical-minerals report ↗](https://www.gao.gov/products/gao-24-106395)

The original link was unavailable when checked.

[Sulayman Bowles](https://sulayman-bowles.dev/about) studies finance at UT Austin and writes about software, AI systems, and markets.

[Biography](https://sulayman-bowles.dev/about)[Writing feed](https://sulayman-bowles.dev/feed.xml)

## Read next

[An AI megawatt is not a megawatt](https://sulayman-bowles.dev/research/ai-systems/the-ai-megawatt)

The modeled facility fits 402,574 B300 GPUs. Its power date, workload, and customer contract decide what that equipment is worth.

[Where do online returns go? Restock, resale or disposal](https://sulayman-bowles.dev/research/financial-systems/where-online-returns-actually-go)

A retailer may restock an unopened item, sell it open-box, send it to a liquidator or let the buyer keep it. The choice comes down to what the item can earn after handling.

---
Source: https://sulayman-bowles.dev/research/data-systems/us-rare-earth-magnet-manufacturing-capacity · Sulayman Bowles
