Actionable insights — Paul Mann's helium deep dive

9:55 1. Supply-shock math — share of global supply offline × substitutability

1. Start with the concentration map: what share of world production comes from each source country? A commodity where the top two producers are ~50% of supply is a candidate; one with twenty suppliers is not.
2. Mark each concentrated source with its current status — producing, impaired, or embargoed — and add up the share that cannot reach a buyer today, for any reason. Physical damage and export controls count the same in the arithmetic.
3. Ask the substitutability question: can the buyer use something else, use less, or wait? Where the answer to all three is no, the shortfall converts to price rather than to demand destruction. Helium fails all three — it is inert, has no chemical analogue at cryogenic temperatures, and cannot be manufactured.
4. Check regional balance rather than the global total. A globally balanced market with a structurally short region still produces a price event in that region, because the deficit has to be shipped in.
5. Distinguish duration: outages measured in months get absorbed by inventory; outages measured in years re-rate contracts. Ask specifically whether damaged facilities are being repaired or written down.
6. Cross-check the conclusion against buyer behaviour, not producer rhetoric. Buyers worrying about volume rather than price is the tell that the shortage is real and that price has further to run.

• Qatari LNG/helium restart announcements — and, more usefully, whether the restart is described as repair or as replacement capacity.
• Any formalisation or relaxation of the Russian export controls.
• Semiconductor fabs disclosing gas-supply constraints in their own filings — buyer-side confirmation beats producer-side claims.

6:31 2. The grade screen — concentration, not field size, decides who is economic

1. For any by-product commodity (helium in gas, gold in copper, lithium in brine), get the concentration in the host stream and compare it directly with the industry-standard grade.
2. Convert the ratio into a processing-cost statement: a producer at 7× the normal grade moves one-seventh the host volume for the same product, so its unit cost is structurally lower regardless of scale.
3. Only then look at reserve size. Grade determines whether the project works at all; size determines how long it works for. A high-grade small field beats a low-grade large one.
4. Ask what fraction of the resource the current build actually consumes. If phases one and two use a small share of proven acreage, the expansion optionality is real rather than promotional.
5. Look for a third-party stamp that the concentration is what management says — a government strategic-asset designation, a certified reserve report, or a development-finance lender's diligence.

• The reserve analysis on the outer acreage promised for later this year — the event that converts "plenty of acreage" into a certified number.
• Any drop in produced helium concentration as more wells are tied in; grade is an average, and averages move.

15:13 3. The operating-leverage-to-price table — build it before you believe the pitch

1. Take the plant's fixed nameplate volume (here 70 mcf/d helium, 2,500 GJ/d LNG) — this is engineering, not forecasting, so treat it as the one hard input.
2. Take the base-case revenue and gross profit management gives at a stated price. The difference between them is your implied cash cost, which is largely fixed once the plant is running.
3. Re-run gross profit at every price scenario, holding cost constant. Because volume and cost are both fixed, the whole price delta lands in profit — so a 40% price rise can be a 60%+ profit rise.
4. Tier the price inputs by evidence quality and never blend them: contracted (signed, bankable), in negotiation (management's claim, unverifiable), reported spot (hearsay, and not the price they will realise on long-term contracts anyway).
5. Apply the same table to the expansion phase to see whether the base-case return is a function of the asset or of the price assumption.
6. Then reverse it: at what price does gross profit go to zero? A high-leverage asset cuts both ways, and the downside case is the one management will not present.

• The realised price in the first reported quarter of sales versus the $600 base case — the single number that validates or kills the whole table.
• Any gross-profit disclosure that implies cash costs above the ~$16m the base case implies.

25:52 4. Contract tenor and take-or-pay as a financeability signal

1. Ask for the contracted percentage of each product line separately, and the tenor in years. "We have customers" is not an answer; "100% of LNG on 5–15 year contracts" is.
2. Check for an escalation mechanism. A fixed-price fifteen-year contract in an inflationary jurisdiction is a liability; an inflation-linked one is an annuity.
3. Distinguish the contract types, because they carry different risks: take-or-pay (buyer pays whether or not they lift — revenue certainty), tolling (you are paid a conversion fee on someone else's feedstock — no commodity price risk at all), and spot (full exposure both ways).
4. Read the uncontracted slice as a deliberate position, not a gap. Leaving 25–50% of a product on spot in a shortage is a bet; leaving 0% is a financing decision.
5. Now connect it to the debt: lenders underwrite the contracted book, not the production forecast. When a facility's drawdown conditions include a contracting threshold, the contracting schedule is the funding schedule.

• Stage 2 contracting crossing the 50% threshold — the checkable trigger on $750m of debt.
• Whether new helium contracts are struck at $600 or nearer the $800–1,000 claimed, and whether tenor shortens as prices rise (a producer shortening tenor is telling you it expects higher prices).

23:04 5. Root-cause triage of a commissioning delay — bought-in component vs process physics

1. When a plant is late, refuse the schedule answer and ask for the failure mode: which specific piece of equipment, from which supplier, failing which specification.
2. Sort the answer into one of three buckets, which carry completely different risk:
   • Bought-in component (a defective part from a vendor) — bounded, fixable, repeatable failure across units but a known remedy.
   • Integration / commissioning — the parts work, the system does not; usually months, sometimes a redesign.
   • Process physics — the method does not achieve the claimed result; potentially terminal.
3. Demand the evidence that discriminates between them. The decisive question is whether the already-repaired portion of the plant performs to spec. If it does, the physics is proven and the remainder is logistics.
4. Check whether the fix is serialised — how many units, at what rate, in what facility. A rework performed by hand in a clean room is a throughput problem with a countable end.
5. Ask what changes on the next build: a different supplier, a re-specified part, or incoming inspection. A delay with no process change is a delay that recurs.
6. Separately, ask whether the shortfall is capacity or quality. "Running at nameplate" for a partial plant is not the same claim as "producing saleable product."

• Stage count disclosed at each update (48 → 96/100) — a hard, countable progress metric rather than a narrative one.
• Whether the second plant (Iceland, "about 12 months from starting construction," 29:06) uses the same compressor supplier.
• The first shipped and invoiced Si-28, not the first enriched Si-28.

26:58 6. Derived-demand chaining — price an upstream input off the downstream product

1. Identify the end product that consumes the input and get its revenue today plus its consensus growth. That is your demand curve, already forecast by people with better information.
2. Establish the conversion ratio — is one unit of end-product demand one unit of input demand, or is it diluted by yield, recycling, thrifting or substitution?
3. Look for physical constraints that block stockpiling. A short half-life, a perishable form or a boil-off rate means demand must be met continuously and cannot be smoothed by inventory — which converts a supply gap directly into a shortage.
4. Count the pipeline behind the lead product to gauge whether today's demand is the peak or the base.
5. Map the supply side to its concentration (method 1) and see whether one jurisdiction controls it.
6. Look for third-party distress that corroborates the shortage independently of the seller — a customer publicly blaming the input for its own delays is worth more than any producer claim.

• Pluvicto quarterly revenue and label expansions into further cancer indications — the direct read-through to Yb-176 volume.
• Any second non-Russian commercial Yb-176 enricher appearing; that is what ends the scarcity premium.
• The continuous processing vessel completing (September) and the first commercial Yb-176 shipment.

13:48 7. Treat logistics as a line in the cost of goods, not a footnote

1. For any product that degrades in transit — boil-off, spoilage, decay, moisture — get the loss rate per unit time and multiply by realistic voyage length. That is a direct, unavoidable cost of goods.
2. Compare producers on delivered economics rather than at the plant gate. A location advantage of a few shipping days can be worth more than a few dollars of operating cost.
3. Check what the customer actually receives, since the degraded form is usually worth far less: liquid versus boiled-off gas, fresh versus decayed isotope.
4. Ask whether the advantage is structural (geography, which competitors cannot copy) or contractual (freight rates, which they can).

• Whether customers pay a delivered premium for the higher liquid fraction, or simply pocket it — the difference between a margin advantage and a marketing point.
• Freight and ISO-container availability for liquid helium; a boil-off advantage evaporates with a delayed vessel.

5:09 8. Turn the pitch into a dated scorecard — six months, item by item

1. Extract every dated, falsifiable commitment management makes and write it down with its date. Discard everything undated ("growing rapidly," "great runway") — it cannot be scored.
2. For each item, decide in advance what evidence would count as met — first invoiced shipment, a filed document, a signed contract — not a press release describing progress.
3. Rank the items by informational value. Any milestone that converts a capital project into revenue tells you more than five that do not.
4. Score the list at the deadline and use the hit rate itself as your prior for the next set of guidance. A management team's calibration is a measurable, reusable input.
5. Track the dependency chains, so a slip early in a chain is recognised as a slip in everything behind it.

• September as the concentration point — three separate "firsts" (helium shipment, Yb-176 vessel, and the earliest possible QLE date) land in the same month.
• The distinction between "producing" and "shipped and paid for" in every update; the second is the only one that scores.

19:38 9. Read a management guide by its internal redundancy, not its headline

1. When a long-dated target is given, insist on the division-by-division build-up. A single group number is unfalsifiable; a decomposition can be attacked piece by piece.
2. Check the width of each range — management's own uncertainty, stated. Narrow ranges sit on contracted or visible revenue; wide ones sit on hope.
3. Sum the low ends and the high ends. If the low-end sum already exceeds the headline, the target is sandbagged; if only the high-end sum reaches it, everything must go right.
4. Look for redundancy: does any single division reach the group target on its own? Redundancy is what makes a plan robust to one division failing.
5. Set the whole thing against a near-term checkpoint from the same management (method 8) so the far-dated number inherits a measured credibility.

• Any revision to the division ranges — the first place a 2031 target erodes is in one segment's low end, long before the headline changes.
• Whether the "helium and LNG alone can do $300m" claim survives contact with the first year of actual realised helium pricing.