Actionable insights — Position management through a resource cycle (and the bust that sets up the buy)

1. Sequence the bust: pop → panic → Fed print → hard assets rally

1. Separate the demand shock from the supply position. A bursting capex bubble removes marginal demand for the input; it does not add a single new mine, well or smelter.
2. Assume the first move is indiscriminate: "everything goes down in a panic," the leadership index can fall 80–90% (the Nasdaq precedent), and your commodity equities fall with it regardless of their fundamentals.
3. Then ask the only question that dates the entry: what does the central bank do in a deep recession? Answer from history rather than forecast — it prints, "more money than any," because a deep recession creates "a lot of problems."
4. Anchor the buy level on the commodity's cost curve, not on a chart: the median cost of production is where supply destroys itself and where the downside is bounded by the producers' willingness to keep operating.
5. Position for both legs: trim what has already run several hundred percent as the bubble cracks, hold the cash through the panic, and redeploy into the same hard assets after the policy response.

• Spot copper approaching the industry's median cost of production (his marker: $3.50–4.00/lb); credit spreads widening on AI-capex borrowers; the first emergency easing/QE facility after the drawdown — that is the redeployment trigger, not the low print.

2. Read a capex boom by how it is funded, not how big it is

1. Track the funding source of the spend in order: internally generated cash flow → equity issuance → debt issuance. The order is a maturity clock on the boom; each step down means the previous source was exhausted.
2. Test whether the spend is a one-time asset or a treadmill. Ask: does this become maintenance capex after year one, or does a competitor's product cycle force a full re-spend? A treadmill can never earn back its capital.
3. Size the boom against GDP — not against the spenders' own revenue — so it can be compared with prior manias (fiber/telecom 1999–2000, housing 2004–2007) that are known to have ended in bust.
4. Apply the return test bluntly: at some point cash flow must justify the outlay. If no one can show the model where it does, treat "it's an arms race" as the admission it is.

• The next rung down the ladder: vendor financing, SPV/off-balance-sheet structures, private-credit data-center paper; rising interest expense at hyperscalers; the first capex guidance cut framed as "discipline."

3. The copper supply checklist — four questions that decide whether a crash is a buy

1. Producer trend: is the largest producing country's output rising or rolling over? (A structural decline in the incumbent is worth more than any demand forecast.)
2. Where is the marginal deposit? If it is in a conflict or expropriation-risk jurisdiction, the price signal cannot reach it.
3. Who can actually sponsor it? Test the marginal project against a Western major's board: an $8bn capital request in a war zone dies in committee, whatever the copper price. If only state-backed or politically protected developers will build, treat the Western supply response as absent.
4. Scale the requirement against history — if the tonnage needed over the next two decades equals everything ever mined, and the capex isn't visible, the deficit is arithmetic, not opinion.
5. Then separate the demand sources: strip out the speculative one (AI) and check whether the durable ones (electrification, global-south development, grid transformers and wiring) still carry the case on their own.

• Chilean and Peruvian monthly output; new-mine FIDs by Western majors in high-risk jurisdictions (their absence is the thesis); grid/transformer order books as the non-AI demand tell.

4. Keep a thesis logbook — so a 60% drawdown is a decision, not a feeling

1. Before buying, write down in a physical logbook why: what type of company it is (developer vs producer), what has to happen, and who is running it — with a bias toward operators who have "serially been successful before."
2. When the position falls hard, do not consult the price. Consult the logbook and ask one question: has anything in the written thesis changed?
3. If nothing has changed and the fall is a market blip, the correct action is to add — this is the only way the volatility becomes an asset rather than a tax.
4. If you cannot answer "why did you buy this?" in one sentence, you do not have a position, you have a trade someone else made for you. Close it or don't open it.

• Thesis-breaking events as distinct from price events: a failed drill result, a permit refusal, a covenant breach, a management change, a commodity-cycle turn. Everything else is a blip.

5. Measure the volatility before you own it — and size by sleep

1. Before entering a resource name, pull its yearly high and low for every year of available history and compute the range. Expect 50–60% peak-to-trough swings within a bull market.
2. Set the position size so that the historical drawdown, applied to your actual dollars, is survivable — meaning you would still add at the bottom rather than sell.
3. Use the physical test as the binding constraint: if the position keeps you awake or makes you check the price compulsively, it is too large — reduce it until it doesn't, regardless of conviction.
4. Accept the gate honestly: if you have neither the knowledge nor the temperament for a 50–60% swing, the sector is not a size problem, it's the wrong sector.

• Your own behaviour as the indicator: checking quotes intraday, seeking reassurance on social media, or an urge to "just get back to even" — all signals the position is oversized before any fundamental has changed.

6. Profit-taking rules that fit the asset — 30% is not one of them

1. Reject rules imported from other asset classes. A 30% gain in a resource equity is noise inside an annual range that routinely exceeds 50% — "if you're selling after 30% gains, you need to get out of resource stocks."
2. The one mechanical rule he'll endorse: after a 100% gain, sell half. Original capital is recovered, the remaining position is free-carried, and it can run for the multi-year outcome.
3. After several hundred percent, switch from mechanical to cycle-aware: trim on evidence the cycle's demand driver is cracking, expecting to buy back lower.
4. Judge the sell on remaining upside and where you are in the cycle, never on the return already earned or a round-number rule.

• Position sizes that have doubled by appreciation alone (the un-trimmed double is the one that later forces a panic sale); the cycle's demand driver showing funding stress before price does.

7. The canned-peaches inversion — price down, thesis intact, buy more

1. For every holding, define in advance what a sale looks like (cheaper price, unchanged product) versus what a defect looks like (the thesis broke).
2. When the masses sell and nothing fundamental has changed, treat it as the grocery-store discount it is: buy more, and be glad of the price.
3. Immunise against the emotional amplifier: social media turns a price move into a narrative, and the narrative is what causes the bottom-tick sale. Do the reading instead — "this is a knowledge-based business."
4. Reverse the test for the up-move too: an unearned price rise with an unchanged thesis is the same signal in the other direction.

• Sector-wide selling with no company-specific news; sentiment indicators (bullish-percent at zero, capitulation volume) coinciding with unchanged operating results.

8. Position for the policy that the debt arithmetic forces — and for how it will be enforced

1. Start from the refinancing schedule, not the rhetoric: the volume to be rolled plus the new issuance needed to fund the deficit. That number, not the Fed's stated intent, is the constraint.
2. Find the yield at which the real economy breaks (his estimate: 7–8% on the long bond). Everything above that is politically unavailable, so the policy response is pre-determined.
3. Expect the historical exit — cap the yield below inflation and inflate the debt away — and expect it to be enforced through regulation of captive buyers (insurers, pensions, 401(k) menus) rather than voluntary demand.
4. Position early in the Cantillon chain: own the assets the new money reaches first, not the wage or the fixed coupon it reaches last. Diversify the political basket itself — invest in one country, bank in a second, live in a third.
5. Treat a "new Volcker" appointment as theatre until the refinancing arithmetic changes: the office cannot deliver what the balance sheet forbids.

• Long-end yields pressing 6%+ with issuance still rising; any regulatory proposal setting minimum "safe asset" allocations for retirement plans, insurers or banks; capital-flow frictions (withholding, reporting, transfer limits) appearing after a yield cap.

9. Audit a supply disruption for the losses nobody is counting

1. When a chokepoint closes, list everything that transited it, not just the headline commodity — LNG, refined products, petrochemicals, fertilizer, and the industries co-located there for cheap energy.
2. Check whether the workarounds address those too. A pipeline or truck convoy can move crude; it cannot relocate an aluminium smelter or an ammonia plant.
3. Quantify the offset that is masking the shortfall (strategic and commercial stock draws, demand rationing) and treat it as a countdown, not a solution: "this is all finite."
4. Ask what the offsetting party was paid — a large, voluntary, uncompensated drawdown by a rival is a signal that something was traded, and that the trade has a political expiry.

• Chinese crude import volumes turning back up (the drawdown ending); Gulf aluminium and nitrogen-fertilizer prices and utilisation; LNG contract diversions; and any softening of US declaratory policy on Taiwan as the tell that the theory has legs.