Notes
Thinking out loud, on the record.
How I think about businesses, time, and where the world is actually heading. The mind behind the ratings, rather than the ratings themselves.
Buying a share of a company is buying a claim on a future you cannot see. That is the whole activity. Every model, every multiple, every discount rate is a way of dressing up one unavoidable question: what will still be true in thirty years?
So I think that question deserves to be asked directly, rather than smuggled into a spreadsheet as a terminal growth rate.
A few things I believe, and argue about here.
Durability beats excitement. The businesses worth owning are usually the ones whose advantage compounds quietly, where each additional customer makes the next one cheaper to win. That mechanism is boring to write about and difficult to dislodge. Most of what gets called disruption is a company borrowing tomorrow's customers at a discount and calling the invoice growth.
Being right about the technology is not the same as being right about the stock. Almost everyone who correctly called the internet in 1999 still lost money, because they owned the wrong equity in an industry they had correctly identified. The gap between a true thesis and a profitable position is where most conviction goes to die. When I write about where the world is going, that distinction is the thing I am most careful about.
Price is a separate question from quality, and confusing the two is the most expensive mistake available to a good analyst. A wonderful business bought carelessly is a mediocre investment. Half the discipline is admiring something and not buying it.
The infrastructure usually keeps more of the value than the application does. Whoever owns the scarce physical thing, the fab, the power, the land, the network, tends to collect regardless of which brand wins the consumer. That belief shapes a lot of what I cover.
Time does more of the work than selection. I pick stocks in public and I still think most people should own an index and go do something else. Both can be true. This is where I defend that, rather than hoping nobody notices.
Nothing in this section carries a rating or gets scored. Notes are longer-horizon and lower-confidence than the research by design, and when a claim here can be proven wrong, I try to say exactly what would do it. A vision with no failure condition is enthusiasm with a longer time horizon.
Longer horizon. Lower confidence. Same arithmetic.
The notes
19 notesNewest first. Nothing here is rated or scored.
-
The average return is not the return you get
Volatility comes straight off the compound return, roughly half the variance a year. A daily 3x fund multiplies the average by three and the drag by nine: over ten years its average outcome is 10.0 times the money and its typical outcome 2.33 times, before any costs.
Read the note → -
A growth rate is a flow divided by a stock
A 36.5 percent fall in net additions moves the headline growth rate 1.25 points, because year-over-year growth is a trailing average of the flow. The stock tells you about last year. The flow tells you about next year.
Read the note → -
A company with a spare dollar has exactly five options
Reinvest, acquire, pay down debt, buy back stock, or pay a dividend. Priced on one consistent set of assumptions the same dollar becomes anywhere from $0.67 to $1.67, and almost nothing in the system rewards getting the choice right.
Read the note → -
The rideshare map in 2031 is a manufacturing question
Autonomous vehicles are 0.9 percent of US rideshare today. Extrapolating Waymo's recent growth five years out gives 234 percent of the entire market, which proves the growth curve is the wrong tool. Twenty percent share needs 107,000 vehicles and a fleet doubling every nine months.
Read the note → -
Free cash flow is not a fact
Operating cash flow is defined by accounting standards. Free cash flow is defined by whoever is quoting it. One set of books produces four defensible answers, and the sign flips between them: on one definition the company burns cash, on another it generates it.
Read the note → -
A premium is a promise, paid in advance
Do not ask whether a deal works, because the synergy you assume does all the work. Reverse it. A 30 percent premium commits the acquirer to finding 5.3 percent of the target's revenue in permanent savings, every year, forever.
Read the note → -
Management gets paid for the wrong thing
An EPS target met by retiring 6 percent of the shares. A growth bonus paid for burning capital. Pay measured on earnings that exclude the cost of the pay. The proxy statement explains more of a company's behavior than its strategy deck, and takes twenty minutes.
Read the note → -
Most of a discounted cash flow is a number you made up
Terminal value carries 60 percent of a ten-year model and 76 percent of a five-year one, and two assumptions nobody can verify swing the whole valuation by 74 percent. A criticism of our own primary method, which is why it is worth making.
Read the note → -
The one-time item that happens every year
A tariff recovery worth 72 percent of Nike's quarter. An $860 million reserve release inside UnitedHealth's medical care ratio. Both real, both disclosed, both gone next year, and every multiple built on top of them inherits the flattery.
Read the note → -
Leverage is a calendar, not a ratio
Same debt, same business, one ordinary cycle, and net debt to EBITDA moves from 1.80 times to 5.97 times without a dollar being borrowed or repaid. The ratio is at its most reassuring exactly when the company is most exposed, and solvency turns out to be a set of dates.
Read the note → -
A low multiple is not a cheap stock
One company, one price that is fair in every year, one ordinary cycle. The reported multiple runs from 70 times earnings at the trough to 7 times at the peak while the value never moves. The screen has the sign backwards at both ends, and the money is in the uncomfortable half.
Read the note → -
Growth is not a virtue
Two companies with the same profit, the same growth rate, and the same cost of capital, worth five times different. Growth is funded, the bill is the growth rate divided by the return on new capital, and below the cost of capital the faster a company grows the less it is worth.
Read the note → -
A hit rate proves nothing
A coin flipping eight calls runs the table three-and-a-half percent of the time, and a five-for-eight record carries a confidence interval from below random to nearly perfect. It takes 153 scored calls to prove even a 60 percent edge. Why our own scoreboard is not yet evidence, and why we keep it anyway.
Read the note → -
Ask what the price assumes
A fair value is an opinion you announce. An implied assumption is a fact you extract. The reverse valuation method behind our recent calls, shown pushing in both directions: it overruled us into one rating and talked us out of another. With a worked example and the honest limits.
Read the note → -
The price makes the buyback
A repurchase is a purchase of ownership from the shareholders who leave, on behalf of the ones who stay. Below intrinsic value it creates value and above it destroys value, dollar for dollar, and the popular test, accretion to earnings per share, approves both. One worked company, one value, three prices.
Read the note → -
Non-cash is not free
Stock compensation is paid in one of exactly two currencies: a share count that rises, or the cash spent buying that dilution back. Both transfer identical value, which is why one looks ten times cheaper and is not. Why the addback is contested, and the one adjustment mistake that charges the cost twice.
Read the note → -
What money cannot buy quickly
Most moat analysis is circular: it observes high returns and infers an advantage. A stricter test asks whether attacking the business would be irrational for a well-funded competitor, rather than merely difficult. What passes, what only looks like it passes, and how moats actually die.
Read the note → -
Right, and still poor
Cisco investors were right about the internet, right about Cisco, and waited twenty-five years and eight months to break even. Three separate questions hide inside every investment argument, and getting the first two right tells you nothing about the third.
Read the note → -
The bottleneck moves
Advanced packaging is the binding constraint on AI right now, not chip fabrication. By the end of this decade it is electricity, because a data center takes two years to build and a grid connection takes seven. What the technology needs at five, ten, and twenty years, and the five things that would prove it wrong.
Read the note →