brain/
sourcestock-market

Autoresearch: Has MP commercially shipped Tb/Dy, and do alternatives or Swarm break the chain? (19 Sep 2026)

Checkpoint on whether MP has commercially shipped Tb/Dy, whether Q3 FY2026 printed, whether Project Swarm binds volume, and whether Lynas/Iluka/Solvay/Carester/Canadian-Estonian HREE routes have reached operational scale that would break the Mountain Pass toll-gate.

Source

Autoresearch: Has MP commercially shipped Tb/Dy, and do alternatives or Swarm break the chain? (19 Sep 2026)

Generated by /autoresearch on 2026-09-19. Synthesized across 3 rounds from 12 fetched web pages, anchored by a Grokipedia MP_Materials primer that is stale for 2026 and is not used as a 2026 source. See Provenance. Treat as raw material — review before promoting into a project or thread. Context: vault/projects/stock-market (existing chain drone-warfare-demand-to-mp-hree-rerate; do not re-date MP; do not treat USAR Serra Verde close as this chain)

Summary

As of Saturday 19 September 2026, MP Materials has not published a Q3 FY2026 print, and no fetched primary source says MP has commercially shipped terbium or dysprosium. The last first-party checkpoint remains the 6 August 2026 Q2 call and 8-K exhibit: mechanical completion of the first HREE circuit in May, commissioning underway, feed introduction “imminent,” and Tb/Dy production plus Mountain Pass→Independence shipments still guided to later this year (Q2 FY2026 earnings release / 8-K Ex. 99.1; Q2 call transcript). A 9 September Jefferies fireside recap has CFO Ryan Corbett still talking year-end internal Dy/Tb for the magnet business, not a completed commercial shipment (MarketBeat recap, 12 Sep 2026; Daily Political recap, 12 Sep 2026).

Project Swarm is explicitly non-binding. MP’s own 30 July launch post says participants reserve 10X capacity “without making binding purchase commitments today” (Project Swarm). That matches the Q2 flexibility language and Adamas’s 26 August read (Adamas Intelligence).

Step 3 stays partial. No signed Tb/Dy offtake and no documented third-party feed contract into Mountain Pass were fetched. The commercial HREE confirmation on the books is still gadolinium, which Corbett described as the SEG leftover after Dy/Tb are pulled first (Q2 call). Lynas already advertises itself as a commercial non-China separator of Dy and Tb oxides (Lynas US Markets), and Neo has a small-scale Silmet line that has made first Tb/Dy process solutions (Neo, 10 Apr 2026). Those are live alternative routes, not proof that the Mountain Pass toll-gate is broken: Lynas is processing its own Mt Weld system; Neo is not yet at routine commercial deliveries; Iluka, Solvay industrial Dy/Tb, Carester/Caremag, SRC Saskatoon, and Energy Fuels commercial circuits remain 2026-commissioning / 2027-ramp claims, not demonstrated operational scale as of this date.

Priors check

  • Confirmed (directionally): Step 2’s last first-party words still stand — commissioning, not commercial Tb/Dy shipment. No fetched source contradicts the 6 August “later this year” guide.
  • Confirmed: Swarm flexibility language is not a purchase order. MP now says so in its own product page, not only on the call.
  • Untested / still open: Signed Tb/Dy offtake; third-party feed into Mountain Pass; Q3 FY2026 numbers; commercial Tb/Dy shipment volumes and realized pricing.
  • Refined, not contradicted: “Sole non-China HREE separator” was already too strong once Lynas listed separated Dy/Tb as current products. That does not by itself fire the 2028 falsifier (operational-scale alternatives that remove the toll-gate). Lynas is a parallel, own-feed separator; the other named Western projects are still pre-commercial or pilot.
  • Surprising vs priors: Corbett’s September wording frames first Dy/Tb as internal magnet use, which is consistent with Q2 “ship to Independence” but weaker than a third-party Tb/Dy sale. Rosenthal on the Q2 call also said the HREE circuit was designed to take third-party feed — capability talk, not a signed feed.

No /calibrate trigger: nothing here contradicts a load-bearing prior. Step 3 remains partial by design.

Findings

Q3 FY2026 is not printed

MP’s last earnings event on the IR calendar recovered in search is the 6 August 2026 Q2 webcast (press release, slides, 10-Q listed as PDFs). No “Reports Third Quarter 2026 Results” IR page exists in search; the only Q3 results page is Q3 2025 (6 November 2025). Third-party calendars still show Q3 FY2026 as projected 5 November 2026 after the close (Public.com earnings page; NextEarningsDate; Yahoo quote “Earnings Date Nov 5, 2026”). The 6 August 8-K (accession 0001801368-26-000047) is Item 2.02 for the quarter ended 30 June 2026 (EDGAR index). No later 8-K fetched in this pass reports Tb/Dy production or shipments.

Do not treat Q3 2025 pages as a 2026 print.

Step 2 checkpoint: still commissioning / “later this year,” not a Tb/Dy shipment

On the 6 August call, Litinsky: “we are actively commissioning our Dy Tb circuit and remain on track to begin shipping product from Mountain Pass to Independence later this year” (Q2 transcript). Rosenthal: mechanical completion in May; punch-list, checkouts, “preparing to introduce feed to the circuit imminently”; ramp depends on commissioning and “quality over quantity”; “we remain on track to produce terbium and dysprosium later this year” (same transcript). Corbett: Dy/Tb “will make its way into inventory and then find its way down to Independence and ultimately, we'll recognize the value of those products through magnet sales” (same transcript).

The 6 August earnings release headlines NdPr volumes, the gadolinium offtake, and Project Swarm. It does not report a Dy or Tb sales KPI (8-K Ex. 99.1).

The freshest post-Q2 color is the 9 September 2026 Jefferies Global Industrials fireside. Two independent recaps of Corbett: MP “expects to produce dysprosium and terbium by year-end for internal use in its magnet business” (MarketBeat, 12 Sep; Daily Political, 12 Sep). Seeking Alpha hosts a transcript of that session; the fetch timed out, so these recaps are secondary. They are still inconsistent with a completed commercial Tb/Dy shipment as of mid-September.

Honesty for the chain: this does not re-date Step 2. It also does not retire the residual Step-2 falsifier on yields / commercial viability — those are untested until product is made and used. Q3 2026 not having printed yet is not a delay past Q3; the quarter is still open.

Gadolinium is still the only locked-economics HREE offtake — not Tb/Dy

Litinsky on the Q2 call: July long-term agreement to supply gadolinium oxide to “a leading U.S. aerospace and defense manufacturer”; “sizable 9-figure deal in total over multiple years” (Q2 transcript). Corbett: “we've locked in economics on this contract.” Sequencing: “the way that we will approach separating out the dysprosium and terbium first … leaves us with an SEG product”; gadolinium was “a very logical way to do that” (same transcript). Rosenthal: first samarium production planned 2028; gadolinium separation project moving to engineering/procurement on a similar timeline (same transcript).

That is a byproduct-pricing confirmation, not a Tb/Dy offtake.

Project Swarm subscriptions exist and do not bind volume

MP’s 30 July 2026 launch page is the cleanest primary:

“Unlike a traditional supply agreement, Project Swarm allows participants to reserve future production capacity without making binding purchase commitments today. As technologies mature and production programs evolve, participants can refine their requirements while retaining priority access to domestic supply.” (Project Swarm)

The same page: capacity is at 10X (Northlake, with the Department of War); participants “may receive preferential opportunities to secure earlier production from our Independence facility”; “multiple leading drone companies have already executed term sheets”; illustrative math is 50 g magnet per drone → 50 t magnets per million drones. No participant names, no contracted tons.

Q2 call (Litinsky): “We have already signed subscription agreements with a number of participants. … Project Swarm allows them to secure future manufacturing capacity today while preserving the flexibility to continue innovating” (Q2 transcript). Q2 release: “executed subscription agreements with multiple leading U.S. and allied customers” (8-K Ex. 99.1).

Adamas (26 August): restates the no-binding-purchase design; “none were named”; “reservations are capped at capacity equivalent to about two million drones a year” (Adamas). The two-million-drone cap is Adamas, not an MP primary number.

Demand-leg status stays partial. Subscriptions + term sheets ≠ binding volumes.

Alternative non-China HREE routes — operational vs not, as of 19 Sep 2026

Lynas (Malaysia) — already a commercial separator, own-feed, not a Mountain Pass toll customer. Lynas’s US-markets page states it is “the world’s only commercial producer of separated Light and Heavy rare earth oxides outside of China” and lists Dysprosium (Dy) and Terbium (Tb) as current Heavy Rare Earths products (separated Sm is still listed under “Future Products” with an H1 2026 target footnote) (Lynas US Markets). That is a first-party commercial claim. Lynas ASX quarterly PDFs that would give Dy+Tb tonnes were not fetched (off the *.gov PDF whitelist). Even on the HTML record, Lynas is a parallel integrated route (Mt Weld → Malaysia SX), not evidence that third-party miners are skipping Mountain Pass.

Neo / Silmet (Estonia) — commissioned small-scale, not routine commercial. 10 April 2026: “successful commissioning” of a “small-scale production line”; “operating at nameplate capacity” with work still on “stable product purity prior to transitioning to routine production capacity”; first separated Tb and Dy process solutions (precursors for metal making), all processing in Europe (Neo release). This is a real Estonian line. It is not operational scale that retires a US toll-gate.

Solvay / La Rochelle — industrial Dy/Tb still a 2026 target, not a fetched start confirmation. 1 June LOI with Viridis: An Nuyttens, “industrial scale separation of [dysprosium] (Dy) and terbium (Tb) by September 2026” at La Rochelle; Viridis Brazilian feed is aimed at 2028 and the LOI is not a definitive offtake (Solvay–Viridis LOI). 29 July Q2 results: “In autumn 2026, Solvay will also begin industrial-scale separation of dysprosium (Dy) and terbium (Tb)”; plus €15–20 million more SX investment (Solvay Q2 2026). Today is 19 September — early autumn, not a post-start operations report. No tonnes, no “we have begun” sentence.

Carester / Caremag (Lacq) — under construction; Q4 2026 commissioning claim. USA Rare Earth’s 23 July 2026 8-K exhibit: Caremag “scheduled for commissioning in late 2026” with design ramp 800 tpa NdPr, 500 tpa Dy, 100 tpa Tb; Humpton: “one of the few facilities outside of China capable of separating heavy rare earths beginning in 2027”; safe-harbor: “Caremag … is under construction and has not commenced commercial operation” (USAR Ex. 99.1). That exhibit also discusses Serra Verde feedstock access for Carester. Per the task: do not treat USAR Serra Verde close as this chain. Cite Caremag only as a French alternative separator that is not yet operating.

Iluka Eneabba — 2027 commissioning (ASX PDFs not fetched). Search indexed Iluka 22 June 2026 and HY26 releases (“commissioning … mid 2027” / “construction 60% complete”). Those PDFs are off-whitelist and were not fetched. Treat Iluka as not operational in 2026 on search-index titles only; do not cite tonne or offtake figures from unfetched PDFs.

Canadian SRC Saskatoon — commercial operations targeted early 2027. Saskatchewan government 8 December 2025 release: SRC facility “fully operational in early 2027”; five-year offtake in which REalloys buys “the vast majority” of NdPr metal and Dy and Tb oxides (saskatchewan.ca). SRC’s own campaign page Cloudflare-blocked on fetch. This is a signed Canadian offtake that does not route through Mountain Pass — but the plant is not claimed as operating today.

Energy Fuels (Utah) — pilot kilograms, commercial circuit “as early as 2027.” 25 March 2026: first kilogram of 99.9% Tb oxide at White Mesa plus “nearly 30 kg” of 99.9% Dy; pilot ~1 kg Tb/week; commercial recovery “operational as early as 2027” for up to ~35 t Dy and ~12 t Tb per year, subject to permits and monazite feed (Energy Fuels). Pilot ≠ operational scale.

Toll-gate read: alternatives that bypass Mountain Pass are being built (Lynas own-feed already running; SRC→REalloys contracted for 2027; Caremag designed for 2026/27; Solvay targeting industrial Dy/Tb this autumn; Energy Fuels 2027). None of the 2027-class plants are fetched as operating at scale on 19 Sep 2026. Lynas is the one commercial non-China Dy/Tb seller on the HTML record, and it is not a third-party feed into MP. That refines “sole separator” rhetoric; it does not graduate Step 3 and does not fire “operational scale before 2028 that removes the toll-gate.”

What this pass did not find (and must not invent)

  • No MP 8-K, IR release, or fetched conference primary that says Tb or Dy has been sold or shipped to a customer.
  • No named Project Swarm subscriber and no binding volume.
  • No signed third-party HREE feed into Mountain Pass (Rosenthal said the circuit can take it; Q2 transcript).
  • Reuters.com is hard-blocked in SOURCE_RELIABILITY.md; not fetched.
  • Lynas/Iluka ASX PDFs not fetched.

Contradictions and open questions

  • Lynas vs “only non-China HREE facility.” Lynas’s own US page claims commercial separated Dy/Tb outside China (Lynas US Markets). The chain’s Step 2 one-liner is too strong if read as “only separator on Earth.” It can still be read as “only US scaled circuit nearing operation,” which this pass does not refute.
  • Solvay “by September 2026” vs “in autumn 2026.” Same company, six weeks apart. As of 19 Sep there is no fetched start confirmation.
  • Caremag “Q4 2026 commence” vs “HREE beginning in 2027” vs “under construction.” All in the same USAR exhibit (Ex. 99.1). Do not treat it as operating.
  • Jefferies recaps vs official transcript. Corbett “year-end / internal use” is dual-sourced in recaps; Seeking Alpha fetch failed.
  • Open: realized Tb/Dy shipment date, volumes, Independence consumption vs third-party sale, Q3 print, Swarm conversion from reservation to PO, Solvay first industrial Dy/Tb batch, Lynas FY26 Dy/Tb tonnes from a fetchable HTML.

Provenance

Rounds run: 3 of 3 (full). No early exit — Q3-print and Carester/Solvay start status needed a third round.

Sub-questions by round:

Round 1 (broad survey):

  1. Has MP commercially shipped Tb/Dy as of 19 Sep 2026?
  2. Is the HREE circuit still commissioning / “later this year”?
  3. Have Lynas, Iluka, Solvay/Carester, Canadian/Estonian routes reached operational scale?
  4. Do Project Swarm subscriptions bind volume?
  5. Has MP Q3 FY2026 printed?

Round 2 (drill-down):

  1. Any post-6 Aug MP primary (8-K / IR / conference) claiming Tb/Dy shipment? — targeting the Step 2 checkpoint
  2. Do Lynas + Neo + Solvay constitute scale that breaks the toll-gate? — targeting the Step 3 falsifier
  3. Are Iluka / Carester / SRC / Energy Fuels still pre-operational? — targeting the 2028 falsifier

Round 3 (resolve remaining uncertainty):

  1. Caremag operating or still under construction as of Sep 2026? — targeting the French-route gap
  2. Jefferies 9 Sep wording: internal Dy/Tb vs commercial sale? — targeting the shipment question after Q2
  3. Swarm: any post-launch binding-volume disclosure? — targeting the demand-leg open question

Anchor source (Grokipedia, fetched before round 1):

  • MP Materials — 8,017 chars extracted (capped) — company primer through 2024–2025; not used as a 2026 source per this pass’s rule.

X sources: not run (--include-x not passed).

URLs fetched (12 successful, 5 failed or blocked, plus PDF skips):

Round 1:

Round 2:

Round 3:

Not fetched (PDF, off *.gov whitelist): Lynas ASX quarterlies / FY26 results; Iluka 22 Jun 2026 and HY26 PDFs; MP Q2 10-Q PDF on IR. Note “not fetched (PDF)” rather than inventing tonnes.

Hard-blocked, search-only: www.reuters.com (per SOURCE_RELIABILITY.md).

Tools used: WebSearch, WebFetch, grokipedia-fetch helper (_lib/grokipedia.py). No x-fetch. Generated: 2026-09-19 21:09 UTC

Referenced by
Mechanisms
Drone-warfare shift → HREE magnets as binding constraint → MP Materials sole qualified non-China supplier → re-rate