Pieter Slegers — Buy-Hold-Sell List: August 2026

Adyen is the plumbing behind card payments for large merchants — one platform that handles the whole transaction end to end, which is unusual in an industry normally stitched together from several vendors. That simplicity is the product, and it is why the letter treats any move away from it as the main risk.

The shares now trade at about 25 times next year's expected earnings against an average of 64 over the past five years. Something that dramatic usually has a reason, and here it is named: this year Adyen made the first two acquisitions in its history — a loyalty platform for €750m and a billing platform for €335m — and some investors read buying growth as a signal that growth has stopped arriving on its own, and worry that the clean single platform is about to get messy.

The counter-argument is that neither purchase touches the payments engine; they sit alongside it and give Adyen more to sell to customers it already has. One of them, usage-based billing, matters more as software companies move to charging per unit of consumption. And management has already raised its full-year guidance to more than 20% revenue growth in constant currency, which is not what a business running out of momentum looks like.

Alphabet is upgraded from hold to buy this month. It is worth being clear about what that upgrade rests on, because two of the three valuation methods the letter publishes argue against it: the shares trade at about 30 times forward earnings against a five-year average of 22 — expensive by their own history — and the reverse-DCF test says the price already assumes 22.6% annual growth while only 15% is expected.

Only the third method supports it, projecting a 13.7% annual return from growth plus dividend plus a modest change in multiple. No written reasoning is given beyond a one-line description of the business. So this is a judgement call presented alongside a spreadsheet that disagrees with it, and a reader should treat it that way rather than assuming the numbers backed it up.

Hermès makes Birkin bags, silk scarves and other very expensive things, and it deliberately makes fewer of them than people want. That single choice is the business model: no discounting, no outlet stores, no leftover stock, and margins other luxury houses cannot match.

It has always been too expensive to buy. This upgrade is a statement that it has become less so — 36 times forward earnings against a five-year average of 48, after the shares fell about 26% over the year. It is still not cheap in absolute terms: the expected return of 10.0% is the lowest of any name on the buy list, and the reverse-DCF test says the price still demands 18.4% growth against 11.9% expected. Read the upgrade as "finally within reach" rather than "a bargain".

Teqnion is a small Swedish holding company that buys profitable niche industrial businesses and leaves them to run themselves — the same model as Constellation Software, applied to workshops and manufacturers rather than software. Its long-run record is strong: about 24% a year over a decade.

The upgrade to buy is the weakest-supported on the list. The multiple has fallen a long way (31 times earnings against a five-year average of 45), but the letter's own expected return for it is 5.96% a year — barely half what it demands elsewhere — and the reverse-DCF check fails. As with the other two upgrades this month, the change is a judgement the published arithmetic does not endorse.

Dino Polska runs small supermarkets in Polish towns, opening new ones relentlessly out of its own cash flow. It has been the archive's favourite illustration of the ideal business: high returns on capital and somewhere to put every zloty it earns.

It is downgraded from buy to hold this month "due to increasing competition" — the first time in this archive that a rating is cut for a reason about the business rather than the price. That is the important part. A reinvestment machine is only as good as the runway in front of it, and competition arriving in those towns attacks the runway directly, not the multiple.

One caveat for the reader: the downgrade is stated twice in the text, but the portfolio spreadsheet in the same issue still marks Dino a BUY, and it still appears in the post's list of nineteen owned names on Buy. The stance recorded here follows the explicit downgrade.

Novo Nordisk makes insulin and the GLP-1 weight-loss drugs. It is downgraded from buy to hold for the same stated reason as Dino Polska: increasing competition.

What makes this the most interesting decision in the issue is that every number the letter publishes screams the opposite. The shares are shown 54.5% below the letter's own fair value, at 12 times forward earnings against a five-year average of 28, with the highest expected return of any holding at 18.9%. The downgrade overrules all of that on a judgement about rivals — which is either the framework working as intended (a cheap price cannot fix a shrinking moat) or the framework being abandoned at the worst moment. Both readings are live, and the issue offers no argument to settle it.

The same bookkeeping caveat applies: the portfolio sheet still says BUY.

Williams-Sonoma sells kitchenware and home furnishings. It is moved from hold to sell "due to valuation concerns" — the first outright sell rating this archive has published on a name in its watch universe.

The useful lesson is not about the company, on which no figures are offered at all; it is about how the rating scale actually works. Being too expensive is enough to earn a sell if the firm does not own the stock — but it is explicitly not a reason to sell something it does own: Games Workshop appears in the same issue at 51% above its own fair value and is simply held. Two different standards, applied to the same input, depending on whether money is already in the position.

Microsoft sits on the buy list this month, and the same issue opens by explaining why Chris Hohn — whose fund made nearly $20bn last year, and whose tollkeeper approach the archive openly admires — sold every share of it. His argument is that artificial intelligence could erode Office and Azure faster than the market believes.

The letter's own figures are mixed: the shares are about 19% below their five-year average multiple, but the reverse-DCF test says the price already assumes 17.5% annual growth against 15% expected. No attempt is made to answer Hohn's disruption case. For a reader, the honest summary is that this issue contains a well-argued bear case for Microsoft and a rating that ignores it.

Martin Marietta digs gravel, crushed stone and sand out of quarries. The reason a serious investor would swap a software giant for that is entirely about geography. The rock itself is worth only a few dollars a ton, so it cannot be shipped far before freight costs exceed its value — which means whoever owns the quarry nearest a construction site effectively owns that site's business. Peter Lynch's phrase for it is a local tollbooth.

New quarries are close to impossible to permit, so the position cannot be competed away, and no software can replace crushed stone. That is the whole argument: a business with no technological risk at all, at a moment when technological risk is what the buyer was trying to escape. No Compounding Quality rating or valuation is attached — it appears here as somebody else's decision.

Vulcan is the other big American aggregates producer, and the same local-monopoly logic applies. The letter adds one striking fact: over the 98 years from 1925 to 2023 it was the second-best performing US stock of all, behind Altria — a dollar invested at the start would have become almost $400,000.

The point being made is about what actually compounds. Neither of the top two long-run performers was exciting; both simply survived and kept charging a little more each year. That is the same durability argument the archive makes elsewhere through the Lindy series, arriving here with a price attached to it. As with Martin Marietta, this is reported as Hohn's purchase, not rated.

Games Workshop makes and sells Warhammer miniatures. It is the one holding in the portfolio that the letter's own spreadsheet shows as expensive on every single measure: the shares trade about 51% above the estimated fair value, at 33 times forward earnings against a five-year average of 23, and the price implies 16.4% annual growth against 7% expected. The expected return is 6.4% a year, the lowest in the book.

It is rated hold and kept, which is consistent with the house rule that valuation is never a reason to sell. Whether that rule is wisdom or an excuse is exactly the question this row poses — especially in an issue that puts a different company on a sell list for the same offence.

LVMH owns Louis Vuitton, Dior, Moët and dozens of other luxury brands. On the numbers published in this very issue it looks like a buy: about 39% below the letter's own fair value, a forward multiple 20% below its five-year average, a positive reverse-DCF check, and an expected return of 14.9% — better than several of the seven names rated Strong Buy.

It is nevertheless rated hold, with no reason given anywhere in the issue. The same unexplained gap appeared in the 2 August portfolio update. Whatever is driving it — a view on Chinese luxury demand, on the founder's succession, on the multiple never re-rating — it is not written down, and a reader should not fill it in on the firm's behalf. Ten days later the 1 September letter says it plainly for the first time: LVMH is one of the three names to be reduced.