How to Spot a Cyclical Peak

In a cyclical business, revenue and margin peak in the same quarter — which means the trailing multiple is lowest exactly when the earnings behind it are least durable. This is the structural trap, and it catches people who are otherwise disciplined about valuation.

The mechanism is simple arithmetic. Price rises flow almost entirely to the bottom line once fixed costs are covered, so margins expand fastest at the top of the cycle. A company earning record margins on record prices produces an earnings number that is real, audited, and unrepeatable. Divide the share price by it and you get a low multiple.

So the question is not whether earnings look good. They will. The question is which observable series turns before the income statement does.

Days of inventory outstanding is inventory divided by cost of goods sold per day. It converts a balance-sheet figure into a duration: how many days of production the company is holding as stock.

The conventional reading is that falling DIO means product is leaving faster than it is being made — tight supply, strong demand, pricing power. Rising DIO means the opposite. Because inventory builds before prices break, it is widely treated as the leading indicator for cyclical turns.

That reading is not wrong. It is just less reliable than its reputation, for three reasons that all happen to be active in the current memory cycle.

There is a fourth, more basic problem: direction is not level. A figure falling from a high base and a figure sitting at a genuine low are different states, and only the second is a tight market. Interpreting a decline requires knowing the normal range for that company across prior cycles — a baseline that a single quarter's figure cannot supply.

Rather than one metric, the more defensible approach is three that fail differently. When they disagree, the disagreement is itself the information.

Manufacturer DIO still belongs in the set — it is simply the least sensitive member of it, and should not be the one you act on alone.

The memory industry in 2026 is a clean illustration, because the three indicators are saying visibly different things at the same time.

The income statement is spectacular. Micron reported fiscal Q3 2026 revenue of about $41.5 billion with gross margin near 85%, up roughly 74% sequentially and 346% year over year. Ending inventory was about $8.6 billion, and the company reported days of inventory of 120. Read alone, that is a picture of a business that cannot make product fast enough.

The pricing second derivative disagrees. TrendForce's figures for conventional DRAM contract prices show the increases decelerating sharply:

Prices are still rising in every one of those quarters. But the rate of increase fell by roughly a factor of six in two quarters, while manufacturer inventory moved by a handful of days. One of those series is carrying far more information than the other.

Supply timing is the slow variable. Large new capacity in this industry arrives years after it is announced, which means the shortage has a scheduled end date largely independent of what demand does in the meantime.

Three indicators, three different readings: manufacturer inventory says tight, pricing momentum says decelerating, downstream buyers say saturated. The framework does not resolve that disagreement for you. It makes sure you have seen all three before concluding anything.

Being explicit about the limits is what keeps a framework from becoming a horoscope.

• —It has no timing. Decelerating price momentum can persist for many quarters before levels turn. “Late cycle” and “about to end” are different claims and this framework only supports the first.
• —It does not price anything. Where a cycle sits and what a security is worth are separate questions, and the second depends on what the market has already discounted.
• —Structural demand shifts can extend a cycle beyond its historical shape. A framework calibrated on prior cycles will read a genuinely larger market as a stretched one.
• —Every figure here is stale on arrival — quarterly reporting and research publication both lag the market they describe.

That last constraint should be familiar. It is the same problem as reading institutional holdings from quarterly disclosure: the data describes a moment that has already passed, and the useful skill is knowing which parts of it decay fastest.