Actionable insights — Horizon Kinetics Q2 2026 Commentary

1. The planted-axiom test — never accept a headline multiple you haven't rebuilt

1. When you hear a valuation claim, ask what it rests on. A "planted axiom" is an unsupported assertion that makes an argument seem logical at the start, with everything after resting on a false or absent foundation ("it's pulled back from its highs and consensus estimates have bottomed out" tells you only what the price did and what others think).
2. Rebuild the multiple on cash, not accounting earnings. Where P/E is meaningless or misleading for a business type, say why, and name the alternative measure before using it.
3. Handle negatives honestly. Analysts normally drop negative-figure companies out of an average; instead, assign them a deliberately generous number and disclose it, so the group average is computable and the bias runs against your own conclusion.
4. Do NOT add stock compensation back. If a company didn't issue options, employees would demand cash — so it is an operating expense, and it dilutes shareholders even in share form. Size it in dollars, not as a percentage nobody feels.

• Any group multiple quoted without saying how negative-earnings members were treated; adjusted-FCF definitions that quietly capitalize stock comp; a falling P/E on rising capital intensity (the sign the E, not the price, is doing the work).

2. The royalty-company cycle test — buy when the commodity is below cost, sell when the premium is embedded

1. Find the moment the commodity price sits below the cash cost of production for a meaningful share of the industry. That is when producers are capital-starved and will sign away future output on terms they would never otherwise accept — high double-digit implied financing rates on 20-year, life-of-mine contracts on world-class assets.
2. Underwrite the purchase as a business valuation selection, not a commodity price wager: the contract discount rate, free exploration optionality on the surrounding land package, insulation from operating costs and balance-sheet risk. If the commodity also rallies, that is upside you did not need.
3. Track the bookend condition. When the commodity is comfortably above all-in sustaining cost, producers regain capital access and the royalty company can no longer sign contracts with the same earnings power or optionality. The reason you bought has expired even though the business has not deteriorated.
4. Name the model's real limits before defending it: no discretion over mine operations, and no pricing power — royalty companies are price takers on the metal.

• Industry-wide cash cost vs spot (the entry trigger); miners' access to conventional capital (the exit trigger); the terms on newly announced royalty deals — falling implied discount rates on new contracts mean the franchise's forward economics are shrinking while the reported ones still look great.

3. Audit the NAV before you accept the premium — and price the compression, not the growth

1. Open up the consensus net asset value. Three inputs decide it: the assumed commodity price path (typically today's price fading modestly), which production is counted (usually only producing or imminent mines — non-producing royalties are valued at zero), and the discount rate.
2. Convert the discount rate into a multiple you can feel. A 3%–5% discount rate on decades of cash flow is the equivalent of paying a 20x to 33x P/E. Compare it to the rate the company itself negotiates in its own deals — if the market discounts at 3% what management underwrites at 15%, you are on the wrong side of that spread.
3. Concede what the NAV misses in the other direction (perpetual gold optionality, exploration upside, future deals) — which is why such a company should trade above NAV. The question is never premium-or-not; it is how large a premium.
4. Then run the compression case, holding the business perfect: earnings growing 15%/yr for 15 years, but the multiple sliding from 2x NAV to 1.5x NAV over five years. If the answer is a single-digit return, the position is a sale on arithmetic alone.

• Sell-side NAV models re-based to a new "current" commodity price (the premium quietly resets); peer multiples at smaller, similarly profitable competitors (the floor the leaders can compress toward); any argument for a premium that relies solely on business quality with no number attached.

4. Watch TIME, not price — the drawdown-fractal check before you conclude "this time is different"

1. When a long-term holding is deeply off its high and the discomfort is "it's never been down this long," go and look it up. Pull the full price history and count the prior drawdowns by depth and duration.
2. Magnify an earlier window of the same chart. The smooth-looking early years are a mirage produced by compounding on a linear axis; blow them up and the volatility is identical — the pattern is fractal.
3. Ask what actually changed in the business model and the valuation. Absent those, "short-term price behavior offers no predictive information, and using price as a decision-making tool is likely to lead to the wrong long-term outcomes."
4. Understand why this is hard: brains built for the veldt spot sudden pattern changes and linear pursuit well, and handle geometric functions and long time frames badly. Under a power law, value arrives disproportionately at the very end of the interval — so the correct thing to monitor is elapsed time, not the balance.

• Your own reaction to a drawdown arriving before you've looked up the prior ones; log-scale vs linear-scale charts (a rocket ship on one axis is a straight line on the other); the 10x-interval stretch — each successive ten-fold gain takes longer, so re-anchor return expectations downward (90%/yr in the 2016 bitcoin era → ~33%).

5. The switching-cost test — price the tax and the years before trading good for better

1. Before selling an appreciated compounder for a better idea, compute the reinvestment haircut: after gains taxes you redeploy roughly 85¢ on the dollar.
2. Assume you are right — the new holding really does grow 25% faster — and solve for the break-even year. In the commentary's illustration that is year seven.
3. Now ask the real question: will you actually hold the replacement for seven years to find out? "It takes a lot of time to develop excess value through growth."
4. Apply the temperament filters: "the hardest temptation to resist is to trade something good for something average" (Peter Doyle), and "patient inactivity is hard, demanding, and exhausting work" (Murray Stahl).

• Switches justified by relative price action rather than a modeled growth-rate gap; your realized holding period on the previous switch (the honest input to step 3); tax-lot cost basis before, not after, the decision.

6. The ETF Divide screen — hunt where index mechanics impose a non-economic discount

1. Screen for the shapes index organizers structurally cannot buy: market caps below the smallest index constituents; companies shrinking their own float through buybacks; heavy insider or parent-company ownership that collapses the float-adjusted cap; conglomerates whose best asset sits inside a "wrong" industry designation; illiquid names an institution could neither build nor exit quickly.
2. Separate the two discount sources and name which one you are being paid for: the Equity Yield Curve (value realization too distant or indeterminate for a relative-return industry) and the ETF Divide (discounts imposed for non-economic, non-fundamental reasons).
3. Verify the fundamental record independently — long-run per-share earnings growth, margins, the buyback's actual share-count effect (retiring 2.5% of total shares is 7.5% of the traded float when insiders hold two-thirds).
4. Then honestly test for NFL risk (No Free Lunch risk): is the market pricing a real risk you are dismissing? If not — and the discount traces to flows, coverage and liquidity rules — the lunch is free.

• Analyst-coverage counts collapsing (fewer eyes = wider discount); buyback pace vs float; a subsidiary IPO or parent take-private as the value-realization catalyst; index-inclusion rule changes that would suddenly make the name buyable.

7. Separate the perpetuity from the depletable asset — then measure it per share

1. Inside any resource company, split the balance sheet in two: the perpetuity (surface acreage, which can be repurposed to higher uses and compounds forever) and the depletable minerals that came with it. They deserve different valuation multiples, and management that understands the difference can create value by trading one for the other.
2. Judge the company on units of the perpetuity per share, not on total growth. Acreage can triple while acres per share falls if it was bought with stock.
3. Prefer a land position large enough that cash flow does not have to be reinvested to stand still — that is what frees the budget for dividends, buybacks and opportunistic expansion.
4. Check management's own cash commitment, distinct from granted equity: a requirement to purchase stock worth a multiple of salary within a fixed window is evidence the team expects to be present for the long compounding stretch.

• Acreage growth funded with equity (dilution disguised as expansion); repurposing optionality — data centers, power, transmission, water, carbon capture on the same surface; whether royalty-contract purchases are consuming the free cash flow.

8. Underwrite the growth that needs no capital

1. Isolate revenue growth that arrives with zero capital spending and no management effort, and quantify it separately from discretionary growth. Two sources qualify: a physical/geological trend in volumes, and contractual inflation escalators.
2. Model the physical trend explicitly (ratio today → ratio in N years → implied annualized volume growth), and state what it does not depend on — customer expansion, new projects, commodity price.
3. Add the contract escalators from a published index to get a floor revenue growth rate; treat repricing of new contracts above the current rate as separate upside.
4. Then value the optionality on top as a call, not as a base case (new use classes for the same asset that generate recurring, high-margin, royalty-like revenue).

• Water-cut disclosures per operator (the physical trend's confirmation); contract renewal pricing vs the current fee; escalator indexes vs realized inflation; the first data-center or power lease signed on the acreage (the option struck).

9. The commercial-utility test — decide whether a new product is a threat before the tape does

1. When a headline says a new venue or instrument threatens an incumbent, first separate the shares going down from the business going down. Check the latest reported revenue, operating margin and per-share earnings, and the incremental margin on new volume.
2. Ask what economic need the new product serves. A hedging instrument requires an expiry date, delivery of the underlying, and an exchange enforcing collateral on both sides — that is what lets a commercial user offset a real exposure.
3. If the instrument lacks expiry and delivery, it cannot hedge a business commitment; it is a leveraged wager (gamification) competing for retail speculators, not for the institutional flow that is most of the incumbent's revenue. Size the customer mix to prove it.
4. Ask the reverse question: can the incumbent simply offer the new product? If nothing precludes it — and the incumbents are already launching versions — the "threat" is a demonstration of demand, i.e. an expansion of the incumbent's addressable market.
5. Finally, check the historical pattern: has this industry converted crises and technological shifts into new products before?

• Customer mix (retail vs institutional) at the "disruptor"; regulatory definition fights (a perpetual as swap vs future) as a timing risk, not a business risk; incumbent product launches — 24/7 futures, tokenized settlement, prediction partnerships — as the tell that the threat is being absorbed; new trade routes or new physical exposures that need contracts that don't yet exist.

10. Triangulate a price with two unrelated methods — a supply floor and a demand curve

1. Build the supply-side floor bottom-up: identify the dominant input cost (here electricity, ~60% of the total), construct a network-wide consumption schedule, divide by known network output to get average fleet efficiency, then divide by units produced per day to get cost per unit. Sanity-check by flexing the inputs (±$0.01/kWh, 55% vs 60% cost share) rather than presenting a single number.
2. Project the floor forward through the known schedule change, and add the empirical premium producers have historically required to keep operating (they could do something else for a living).
3. Build the demand-side curve from an independent first principle — here Metcalfe's Law (network value ∝ users², a t^2 function) combined with network size expansion (t^3, the snowball's cubic volume), giving a power function a touch under t^6.
4. Backtest the demand curve against realized prices over the full history before using it forward, and report the misses.
5. Only act on the convergence. Two methodologies with nothing in common landing in the same range is the signal; either one alone is a model.

• Mining rig efficiency generations and electricity contract pricing (both move the floor); hashrate as the demand-side proxy (up 10x in five years — and note the open question of whether price follows hashrate or the reverse); the stretching 10x interval (1.1 → 1.6 → 2.4 → ~8 years) as the reason to lower, not abandon, expectations.