Pieter Slegers — $100,000 Annual Cash Flow (Portfolio Update)

Evolution runs live-dealer casino games — real croupiers filmed in studios, streamed into online casinos which license the games rather than build them. It is the dominant supplier, and its first appearance in this archive is an unflattering one: second-quarter growth of 2.4% in constant currency, half the first quarter's rate, described as "worse than we hoped."

What makes the position interesting is what the company is doing with the low price. The finance chief has confirmed they are buying back shares as fast as the rules allow — up to a quarter of all daily trading volume in the stock. Because the shares have fallen so far, the arithmetic is extreme: the sum they are spending retires roughly 16% of the entire company in a year, and combined with the dividend that gives owners an 8-10% annual return before the business grows at all. In a single week in July they retired a million shares.

This is the "cannibal stock" idea in its purest form. When a company is shrinking its own share count that quickly, each remaining owner's slice of the profits grows even if total profits merely hold steady. The growth assumption in the portfolio model is correspondingly modest — 5.7% a year, among the lowest of any holding — because the return is not supposed to come from growth. Notably, it is one of the few holdings actually up this year.

Novo Nordisk makes Ozempic and Wegovy, and shares the obesity and diabetes drug market with Eli Lilly. Nine days before this update, Slegers passed on Lilly and said he was "more comfortable with Novo Nordisk" — this post shows why on price and, unintentionally, why the choice is not free.

On price the case is the strongest in the portfolio. The shares change hands at about 12 times next year's earnings against a five-year norm near 28 — the widest gap of any holding — and the fair-value calculation says they are worth roughly double the current price. The news is good too: oral Wegovy and a higher-dose injection are now approved across the EU, with the pill rolling out through more countries during 2026.

The uncomfortable number sits in the owner's-earnings table. Novo is the only holding whose profits are forecast to go nowhere: earnings per share of 23.03 in 2025 against 22.98 in 2028. Strip out the 3.5% dividend and the expected three-year return is 3.4% — the lowest in the book. So this is not a growth position at all; it is a bet that a company whose earnings are flat for three years is priced as if they were about to fall, and that the multiple recovers. Worth holding that tension in mind when reading the Lilly comparison, where the objection to Lilly was that its price assumed 18.9% growth.

Constellation buys small software companies that sell essential, unglamorous systems to niche industries, and never sells them. The engine is simply the rate at which it can keep buying, and this update quantifies it through RBC's estimate: $646 million deployed in the second quarter, $1.63 billion so far this year, and an expected $3.12 billion for 2026 — comfortably above the previous record of $2.46 billion set in 2023.

That matters because the shares are down 14% this year on fears that AI will make small software businesses obsolete. The acquisition figures say the opposite is happening operationally: the company is putting more money to work than ever.

The valuation gap is the largest in the portfolio. On the fair-value calculation the shares are worth $4,565 against a $1,909 price — more than double. The forward multiple is 15.8 against a five-year average of 31.5, so the market is paying half what it habitually paid for the same business. And the reverse calculation says the price only requires 7.5% annual growth, against the 15% expected. All three measures agree, which is the condition this archive requires before acting.

Topicus is Constellation's European arm, doing the same thing on a smaller base — buying vertical-market software businesses across Europe and holding them permanently. Smaller matters here, because a smaller acquirer has more room left to grow.

It is the second-largest modelled upside in the portfolio: $65.4 against a fair value of $133.5. Earnings are forecast to rise from 3.43 to 5.49 a share by 2028, a 17% annual rate and the second-highest expected return of any holding. There is no dividend — everything is reinvested, which is exactly the point of owning a serial acquirer.

The July acquisition named here, DiffusionData, is a small British company providing real-time data streaming for mobile, web and AI applications. It is worth noting the direction of travel: the AI fear that has taken 26% off the share price this year is the same technology the company is now acquiring infrastructure for.

Kelly Partners buys Australian accountancy practices and runs them under one roof. It is the most extreme position in the book in both directions: down 55% this year — the worst performer by a wide margin — while carrying the highest expected return of any holding, at 30.5% a year.

That figure comes from earnings per share forecast to rise from 9 to 20 by 2028, a more than doubling in three years. The fair-value calculation puts the shares at roughly double today's price, and the current multiple is 18 against a five-year average of 29.

The size of both numbers is the warning. A projection of 30% annual growth from a company whose shares have more than halved is either a genuine mispricing or a forecast that has not yet caught up with reality — and results due on 6 August will settle a good deal of it. Note also that this update does not revisit the governance problem documented earlier in this archive, where the founder was forced to surrender a third of his stake after a margin call. The rating is BUY; the risk is not discussed here.

Games Workshop makes Warhammer — the miniatures, the rules, the shops and now the Amazon television rights. It has been the portfolio's best long-term investment by a distance, compounding at nearly 47% a year over ten years, and it is up 16.5% this year while almost everything else has fallen.

Which is precisely the problem, and the reason it is the only holding rated HOLD on business grounds. It trades at 33 times forward earnings against a five-year average of 23, and the model's fair value is roughly 40% below the market price. The expected return over the next three years is 6.3% — below the 10% threshold that defines fair value here — and the reverse calculation says the price requires 10.9% annual growth against the 7.0% expected. That negative gap is unique in the portfolio.

What is instructive is the response: it is held, not sold. The archive's stated policy is that a good business becoming expensive is not automatically a sale — only two positions have been sold since 2023 — and this is that policy being applied to the clearest case for selling that the numbers have produced.

LVMH owns Louis Vuitton, Dior, Moët and dozens of other luxury houses, selling mostly through more than five thousand of its own stores. It has fallen 22% this year and its five-year record is now slightly negative.

The strange thing about this entry is that the numbers and the rating disagree. Every valuation column says it is cheap: 19.9 times forward earnings against a five-year average of 24.9, a fair value roughly 38% above the price on one sheet and 56% above on the other, an expected return of 14.7% a year, and a reverse calculation requiring only 6.2% growth against the 9.6% expected. On its own criteria this should read BUY.

It says HOLD, and the post gives no reason. The most likely explanation is a position-sizing decision rather than a valuation one — the holding may already be at its intended weight — but that is inference, not disclosure. The standing thesis in this archive is that LVMH is a coffee-can name meant to be owned indefinitely, so the HOLD is not a warning about the business.

Medpace runs clinical trials for small and mid-sized biotech companies. It gets paid for running the trial whether the drug works or not, which is why it is owned instead of the drug developers — the same exposure to biotech activity without the binary outcome.

It is the oldest position in the portfolio, bought in October 2023, and it is the only holding shown as trading above its fair value: $526.6 against $439.4, roughly 17% too expensive on that sheet. The separate valuation sheet reaches a milder conclusion — 5% undervalued — and still rates it BUY. The two tables use different assumptions about the multiple, and the disagreement is not explained.

Its own multiple, 29.7 times forward earnings, is almost exactly its five-year average, so there is no re-rating to collect. Earnings are forecast to grow about 11% a year with no dividend, giving an expected return of 10.4% — right at the threshold. This is what a fully-valued quality holding looks like in the model: not a sale, but not a source of the portfolio's upside either.

Kinsale insures the risks ordinary insurers decline, mostly for smaller businesses, and does all its own underwriting and claims handling with its own software. The July update argued that the current soft market — falling premium prices as competitors pile in — would eventually clear out the weaker participants and let prices recover, with Kinsale surviving as the low-cost operator.

This update quietly moderates that. The multiple case is intact and striking: 17.2 times forward earnings against a five-year average of 28.6, nearly 40% below its own norm. But the earnings forecast now assumes only 6% annual growth to 2028, giving a three-year expected return of 6.2% — the second-lowest of any holding, for a company with a 34.7% ten-year compound rate.

Read together, those two facts describe the position honestly: the recovery in premium pricing is not being assumed in the numbers. If the soft market persists, 6% is what you get; if it turns as argued, the multiple has a very long way back.

Berkshire is not owned in this portfolio. It appears as the historical argument, and this time with three specific windows rather than a general appeal to Buffett's reputation.

In 1999, at the height of the internet boom, Berkshire fell 18.9% while the index rose 23% — a 42-point gap, and Buffett was widely written off. Over the past year the same pattern is visible in miniature: Berkshire up 3.5%, the index up 19.5%. Then the third window, which is the point: from 2000 to 2003 Berkshire rose 53.7% while the index fell 19.2%.

The claim drawn from it — "periods of underperformance are always followed by periods of outperformance" — is stronger than the evidence supports, and the word "always" is worth noticing. But the structure of the argument is sound, and it is being made alongside live evidence rather than in isolation: the semiconductor ETF fell 20% in mid-July while the index was flat, and quality names held up over the same month.