08
Seven things I have tried to answer and can't. Public sources don't contain them, or contain contradictory versions. If you know any of these, I'd like to hear from you.
Last updated 27 August 2026. Answered questions get moved to the bottom with the answer.
Most research gets published as if the author has it figured out. The interesting part is usually the opposite. The specific places where the evidence runs out.
These are mine. Each one is something I've searched for properly and failed to find, or found in versions that contradict each other. Several of them are load-bearing for conclusions elsewhere on this site, and I'd rather flag that than let a weak claim sit behind a confident sentence.
DFARS 252.225-7052 extends upstream on 1 January 2027 to cover feedstock, including recycled material, for tungsten, tantalum and certain magnets. The clause text is clear that recycled material is in scope.
What no source I can find explains: what a recycler must actually maintain to certify that commingled scrap, material aggregated from many suppliers, is not of covered-country origin. Is there a de minimis threshold? A mass-balance provision? A chain-of-custody safe harbour?
I have searched DoD guidance, FAQs, class deviations and the rulemaking docket. Nothing addresses the documentation method. The silence is itself the finding, five months from the effective date.
Who would know: government contracts attorneys practicing in DFARS sourcing; DPC staff; anyone who has been through a prime's supply chain audit on a covered material.
This one was partly my own error and it's worth showing. I had two APT figures that couldn't both be right, and I said so here.
They were the same series. Fastmarkets assesses APT 88.5% WO3 min, cif Rotterdam and Baltimore duty-free, in dollars per MTU, weekly. It runs from $900–940 in January 2026 to $1,650–1,900 in mid-February to roughly $2,900–3,210 in late June. The figure I couldn't reconcile was that series quoted secondhand with the unit wrong. Per metric ton was never a real assessment.
What I still don't have is the series itself. Everything above is republished by third parties, at a lag, in fragments. So the question is narrower now, and more practical: is there a way for an independent buyer to get an assessed tungsten price without an enterprise subscription?
It got harder in August 2026, not easier. The Western and Chinese domestic benchmarks have decoupled far enough that Fastmarkets launched a separate China domestic APT assessment to track them apart. There are now two series to follow rather than one, and the interesting question, which of the two the West's price is really keyed to, is exactly the question the free fragments can't answer.
Who would know: anyone with an Argus, Fastmarkets or Asian Metal subscription; traders; procurement managers who actually buy the material.
I can demonstrate that the services layer around this supply chain is fragmented. Directory listings run to hundreds of firms across the relevant categories.
What I cannot demonstrate is the size distribution, the ownership pattern, or the asset profile of the businesses inside it. How many firms. What revenue bands. How many are owner-operated versus held by a parent. How old the owners are.
Those are different claims, and I conflated them in an early draft. Free sources do not carry market structure for businesses this size, and I have hit the same wall in four consecutive screens. It is the most persistent gap in this entire body of work.
One series does bear on it, and it points the wrong way for the optimistic reading. US cutting tool shipments were $239.8M in May 2026, up 15.2% year over year in dollars, with year to date up 16.8%. But units shipped fell for a second consecutive month, and the reported cost per unit on carbide drills and end mills rose. AMT and USCTI attribute much of the 2026 gain to inflation rather than units. The machine tool series says the same thing louder. June orders were $672.7M, up 56.8% year over year, and the first half hit $3.44B, up 36.0% and the strongest half since the series began in 1998. Machines ordered were 2.6% lower than the second half of 2025. Record dollars, fewer machines. If you care about how much material moves through these shops rather than what it costs, the units line is the one that matters, and it has been falling. Manufacturing capacity utilization was 76.0% in July, 2.2 points below its long-run average and a slight improvement on June's 75.7%, which is the one number here that does not fit the story I am telling.
I can read that two ways and I don't yet know which is right. Falling units means fewer tools consumed and fewer worn tools generated. Rising cost per unit is exactly the pressure that pushes a shop to recondition rather than rebuy. Both are live. Two months is not a trend.
Who would know: trade association staff who maintain member directories; distributors and consumable suppliers who sell into every shop in a region; regional business brokers.
Every operating benchmark in my models, revenue per employee, EBITDA margin, labor share, is borrowed from an adjacent industry, because no category-specific figures exist publicly for the services categories I care about.
This matters more than a normal proxy caveat, because in several cases I'm arguing the category behaves unlike the proxy while using the proxy to set the inputs. Both cannot be true.
Who would know: owners and former owners; accountants who serve them; anyone who has bought or sold one.
Recycling is the most promising Western supply route for several of these materials, because scrap is already enriched in exactly the elements that are scarce.
But the size of the available pool is unclear. Large integrated producers internalise their own scrap through closed-loop systems at rates that appear to be very high. If most material never leaves the majors, the independent pool is far smaller than an outside estimate would assume. Which changes the economics of every recovery business.
I have no reliable US figures on volumes or collection rates.
Who would know: scrap processors; procurement at integrated producers; anyone running a buy-back program.
A claim I rely on in more than one place: the compliance obligation arrives on 1 January 2027, but US-funded processing and magnet projects don't reach practical readiness until somewhere in 2028–2030. Leaving a gap where the requirement is live and the domestic supply to satisfy it isn't built.
That claim currently rests on a single source. It's consistent with what I see elsewhere, and litigation between the two largest domestic magnet players is weak indirect support for slippage rather than acceleration. But one source is one source, and this is load-bearing.
Who would know: engineers and program managers at the facilities under construction; DoD Office of Strategic Capital; primes doing their own supply planning.
Demand-share splits by end market, EV versus wind versus industrial motors versus electronics versus defense, vary widely between sources, and most trace back to modeled estimates rather than counted volumes.
The share attributable to general industrial motors is the one I trust least and the one that would move my thinking most, because it's the quietest and largest installed base of the lot.
Who would know: Adamas Intelligence and similar specialist houses; magnet producers; large motor OEMs.
If you can answer any of these, I'd like to talk. Twenty minutes, and I'll share what I've mapped on your segment in return. Nothing gets published with your name unless you say it's fine, and if you want the whole conversation off the record, say so and it will be.
I'm also glad to be told something here is wrong. That's the most useful call I can get.