When Peak Earnings Look Cheap

An 84% margin is a price, not a moat

Look at where the growth came from, and almost all of it is price, not volume. Micron did not ship dramatically more memory; it sold roughly the same bits for far more money, as a shortage in the kind of memory that feeds AI servers sent prices sharply higher. An 84.6% gross margin on a near-flat cost base is what pricing power in a shortage looks like — which is a different thing from the durable, structural margin of a business with a moat. It tends to be a feature of a cycle near its peak, rather than a sign that the rules have changed.

Why the multiple can mislead near a peak

Commodity earnings do something awkward to a multiple. Near a trough, earnings are small and the ratio looks high; near a peak, earnings are large and the ratio looks low. Nine times peak earnings is not the same as nine times normal earnings — and a low multiple often reflects a market that doubts the earnings will last, rather than a bargain it has overlooked.

Micron’s own history is a useful reference here. A few years ago it earned an operating margin close to 50% at the last peak; one cycle later it was loss-making and consuming cash. The memory cycle has not gone away — it simply moves hard in both directions.

Demand leads supply — and demand is borrowed from capex budgets

The bull case rests on very strong AI demand. That demand is real today — but, near term, it largely mirrors the capital-spending budgets of a handful of hyperscalers, on the order of three-quarters of a trillion dollars this year. When the customer’s demand is a capital budget, that budget sets the pace: it can pause, it can double-order, it can air-pocket. Management itself has begun to mention a “moderation in the rate of price increases,” even as the cost of making each bit rises and meaningful new supply is due around 2028. It is worth noting which side of the case the 2028 horizon actually serves: it is when supply catches up — a timing point for the cautious view rather than the optimistic one.

The one thing that could raise the floor

There is a genuinely new feature this cycle, and it deserves credit. A meaningful slice of the book is now under multi-year, take-or-pay supply agreements — long contracts with minimum volumes, large customer deposits and floor pricing. Memory has never really had this. If those contracts hold, they lift the trough for the contracted portion of the business — the next downturn could be a higher low than the brutal ones of the past.

Two caveats keep it honest. The customer deposits are a form of financing that gets returned later, not permanent profit; and the contracts cover part of the book, not all of it. So they could soften the next down-cycle — they do not abolish it.

The two-sided question

The cautious case is mostly gravity: memory has always mean-reverted, costs are rising, fresh supply is due around 2028, and a low multiple at a peak tends to warn rather than tempt. The other side deserves a fair hearing: if the take-or-pay contracts, plus a structurally larger pool of AI demand, genuinely raise the trough, then the next low is shallower than history — and a stock priced for collapse would instead re-rate for durability. The aim here is not to call the top. It is to keep the distinction between a good business and a good trade, and to remember that near a peak a multiple can mislead in both directions.

What to do with this — and what we are not doing

What to look for. The price-versus-volume mix in the next set of guidance (is the growth still almost all price?); the trend in the cost of making each bit; how much of the book is genuinely under floor-priced contracts; and — the real clock — whether hyperscaler capital-spending plans are being revised up or quietly trimmed.

What to be wary of. Anchoring on a low headline multiple at a cycle peak; extrapolating shortage-driven margins as though they were permanent; and treating a single memory name as a proxy for the whole AI trade. One practical note: the options market is pricing this stock at extraordinarily high implied volatility — the move can be violent in either direction, and high volatility cuts both ways.

What we are not doing. This is general market commentary, not advice or a recommendation on any security. Micron is named only to illustrate a publicly reported market development — how a commodity business is valued at the top of its cycle. Anything that touches your own portfolio is a suitability conversation — one that needs a person, not a page. We are always glad to have it.