Actionable insights — how a contrarian sources out-of-favour ideas with improving fundamentals

48:31 1. Run toward areas of difficulty — then demand improving fundamentals

1. Start from the base rate: "the most money tends to go to the wrong place at the wrong time." Treat crowding — a sector at ~50% of an index, a jurisdiction everyone owns — as a valuation risk, not a signal.
2. Deliberately go where the news has been bad: a country that just had a constitutional crisis, an economy that "shot itself in the foot," a sector everyone dumped. "I have a tendency to encourage our team to run to areas of difficulty."
3. Difficulty alone is not the trade. Require the second half: improving fundamentals — a rate cycle turning up, dividends/buybacks starting, ROE inflecting, a policy regime changing — so you are buying a bad narrative with a better forward, not a value trap.
4. Separate spotting from allocating. He self-diagnoses as early ("the goal is to try not to invest or allocate too early") and lets valuation, not conviction, decide when to size up.

• A named, dateable bad event that explains the discount; then at least one fundamental line item that has turned (rates, dividend policy, buyback authorisation, occupancy, ROE). No inflection = keep waiting.

57:41 2. Screen for "supply down, demand up" — and insist on both legs

1. "One of the easiest ways to make money is if you can identify sectors or companies that are benefiting from a decrease in supply with an increase in demand. Economics 101."
2. Prove the supply leg with a mechanism and a duration: what physically stopped new capacity (cost of capital, permitting, bankruptcy, a country falling out of favour), and how long the lag is before it can restart.
3. Prove the demand leg with a new buyer class, not just a cyclical rebound — a tenant type or end-market that did not exist in the prior cycle.
4. Cross-check against replacement cost: if you can buy the existing asset for materially less than it would cost to build today, the supply leg is real in dollars, not just narrative.

• Construction starts / permits in the asset class; replacement cost vs market price per square foot; a genuinely new demand cohort you can name; whether high rates that killed supply are starting to fall (which would eventually restart it).

51:09 3. Track supply, not demand — it's the knowable half

1. "It's a lot easier in my experience in my career to track supply than to figure out demand." Build the position on the leg you can actually count.
2. Count suppliers first: two or three credible producers is an oligopoly, and an oligopoly changes what a "cyclical" business is worth. One supplier to the fastest-growing customer is better still.
3. Count the lead time to add capacity, including the tools: fabs take years, and the equipment queue (ASML, Lam) is itself the bottleneck — "get in line."
4. Then treat demand as the residual risk and monitor it with channel checks rather than forecasts: ask companies directly whether the spend is producing a return.

• New fab/capacity announcements and equipment order books; the number of credible suppliers changing; hyperscaler capex guidance; and the ROI check — whether AI spend is visibly lifting margins at big adopters (his stated test: "if they're not improving the margins of banks, they've got problems").

33:39 4. Mine the second derivative — buy the counterparty your research complains about

1. When you do a sector deep-dive, log not just the answers to your questions but the recurring complaint that shows up unprompted across every management call.
2. Ask who is on the other side of that complaint. A constraint for one industry is pricing power for its supplier or landlord.
3. Verify the constraint is structural (capacity that has been permanently removed) rather than a one-quarter bottleneck.
4. Then choose your side of the trade explicitly — and be willing to own the constrainer instead of the company you originally researched.

• The same complaint from three or more unrelated managements; leasing spreads / re-leasing rents at the landlord; vacancy trending down; and the reverse signal — tenants suddenly reporting space is easy to find (which would flip the trade back).

31:38 5. Build a lived-experience channel check you can run for free, forever

1. Identify a consumer cohort you have privileged, repeated access to, and turn access into a standing data feed — for a consumer business, the cohort's spending choices lead the reported numbers.
2. Ask the same question every time ("where are you going to spend your money?") so the answers are comparable across years, and note when a brand stops being named.
3. Know your blind spots: a cohort you have aged out of is a weaker signal, so discount it rather than pretend ("garage has been a trickier analysis process for me").
4. Pair it with on-site work — head offices, R&D tours, the actual product — so the qualitative read is anchored to management and technology, not vibes.

• A brand dropping out of the cohort's answers; store traffic your own eyes can verify; and the discipline test — whether the qualitative signal is confirmed by the free-cash-flow yield before you buy.

10:52 6. Use takeovers received as the scorecard for your own underwriting

1. Keep a running count of portfolio companies that receive takeover offers. It is an independent third party paying cash to confirm your valuation work — the cleanest feedback a discount-to-value process can get.
2. Understand the mechanic you are renting: an acquirer trading above its own asset value can pay a premium for a target trading below asset value and still create value for itself. Screen for that spread rather than for cheapness alone.
3. Read the acquirers' passports as a macro signal: a shift toward non-domestic buyers tells you where the relative-value gap has opened.
4. Never buy for the takeover — "I don't buy a stock because I think it's going to get taken over… but would I be surprised? Hell no." Underwrite the standalone, treat the bid as free optionality.

• Premium-to-NAV acquirers circling discount-to-NAV sectors; private capital buying listed assets (a mispricing tell); and the count itself — a process producing zero bids over years is not being validated.

44:16 7. Buy the "two-quarter problem" — refuse to pay the price of certainty

1. Define the target shape: a company with "a temporary dislocation that isn't structurally going to be challenged longer term" — Morrison's "two-quarter problem," or a "bear trap."
2. Separate the two kinds of unknown: who will run it (management turnover) and how it will be priced (business-model transition) are usually resolvable; a broken competitive position is not. Only take the first kind.
3. Name the marginal buyer who is sitting out and why. "A lot of size investors" will wait to meet the new CEO/CFO/IR — that queue is your discount, and it clears mechanically once the meetings happen.
4. Price the trade-off explicitly: "if you wait for the certainty, you pay a high price for the certainty." Substitute your own work (get on a plane, tour the site) for the certainty you're declining to buy.
5. Know your capacity edge: at $2.2B you can change your mind; at $200B or $2T you cannot. Only run this play if you can actually exit.

• The resolving events themselves (new CEO/CFO named and met, pricing model disclosed); evidence the dislocation is operational rather than competitive (win rates, renewals); and re-rating as the sidelined buyers return.

45:52 8. Read management conviction off balance-sheet actions, not language

1. Insist on the precondition: net cash (more cash than debt). "Why do we like owning companies with clean balance sheets, net cash? Because when they think their stock gets too cheap to be true, they step up and buy the stock."
2. Grade buybacks by timing and circumstance, not size. A buyback supersized at a specific low price band, executed before a new CFO was even hired — the executive who normally signs off on capital structure — is a much stronger statement than a routine authorisation.
3. Weight insider purchases at strength: a CEO buying 10,000 shares just before a blackout window, with the stock at a 52-week high, says the insider still sees a gap to value.
4. Prefer founder-led / founder-owned businesses, where the decision-maker's own money is the alignment mechanism.
5. Sanity-check quality alongside the signal (e.g. software's "rule of 40": growth + margin ≥ 40) so you're not just buying a cheap bad business that likes its own stock.

• Buyback execution prices vs the tape (disclosed in filings); insider buys at highs rather than lows; net-cash position maintained rather than levered up; and the anti-signal — heavy equity issuance instead of repurchase.

53:02 9. Arbitrage the same industry across jurisdictions — and go where the puck is going

1. Pick an industry whose structure you already understand deeply in your home market, then find the structurally-identical version in another country and compare the same metrics side by side (price/book, P/E, ROE, dividend policy).
2. Accept that some discount is permanent (governance, currency, index inclusion) — then ask whether the current gap is at a normal or extreme point of its own history. Extremes are the trade: "the gap… is at one of the widest junctures I've ever seen in my career."
3. Look for the capital-returns catalyst rather than an earnings catalyst. Buyback and dividend culture spreads country to country by imitation — Apple's 2012 template → Japan for a decade → Korea copying Japan → the dividend-cultured UK now repurchasing. Position where it is arriving, not where it is finished: "go where the puck is going as opposed to where the puck is."
4. Watch for domestic-capital repatriation as the flow mechanism — investors in each jurisdiction bringing money home is what closes the gap.

• Buyback authorisations and dividend-policy changes at non-US companies; governance reform programmes; local-investor flows repatriating; and the gap metric itself — track it, because the trade is done when it normalises, not when the story feels good.

56:13 10. Rotate from the enablers to the enabled — buy the adopters, not the vendors

1. Assume the obvious infrastructure winners are already priced ("it's obvious the AI winners are Nvidia… that ship sailed a few years ago") and ask your team for the non-obvious version: "tell me something I don't know."
2. Screen for the adopters: companies with proprietary data or scale that new technology makes more productive, whose margins should expand as they deploy it. His historical precedent: in the 1990s, buy the non-tech company that knew how to use an Oracle/Sybase database — the banks were early adopters and it "added basis points to ROE."
3. Add two hard filters so you're not paying for a story: a clean balance sheet / net cash, and a founder still in the seat.
4. Track adoption directly as a research pipeline — maintain a list of who is implementing the enabling technology, and treat a partnership with the leading vendor as third-party validation (without paying the vendor's multiple).
5. Time it by capex payback, not by the narrative: if the enablers' spending is genuinely productive, the margin expansion has to show up at the adopters next. He explicitly draws on running money through the NASDAQ bust — "not the same, but there are similarities."

• Margin expansion at named adopters (the actual proof); a tracked list of implementations/partnerships; commoditisation at the vendor layer (his example: "the fourth or fifth LLM"); and honest ROI feedback from company calls, which he says is still "mixed."