This week MSN ran a piece arguing Micron looks like a top AI investment right now, resting on strong multiyear guidance from the company. Memory makers don't usually hand out multiyear forecasts. When one does, it deserves a slow read.

The bull case is simple enough. AI data centers need high-bandwidth memory, and Micron is one of the few companies that makes it. Per secondary reporting, CEO Sanjay Mehrotra has guided for fiscal 2027 to top a record fiscal 2026. When a CEO stands up and says demand looks strong for years, not quarters, that's real information from someone with a better view of the order books than any of us will ever have. Multiyear guidance in a cyclical business is worth something. Micron has had its share of noise lately too, insider sale filings from the CEO among them, though none of that tells you the business is bad. It tells you the stock is noisy, which memory stocks have always been. I owned a memory name briefly in the 1990s and sold it after about eighteen months, because I got tired of guessing which direction the cycle was spinning.

But my caution here is an old one. In memory, the good times plant the seeds of the bad times. High prices attract capacity, capacity catches up with demand, and prices fall until somebody blinks. Anyone who has watched this business over the decades has seen that movie more than once, across generation after generation of memory chips. High-bandwidth memory may be different, because it's harder to make and tied to a small set of giant customers. Or it may not be. Nobody knows yet, and anyone who says they do is selling something.

There's a concentration question worth asking too. When a large share of your demand comes from a handful of hyperscalers building data centers, their capital spending decisions become your earnings. We've seen what happens when one of them pauses. Ask anyone who owned semiconductor suppliers in 2022.

None of this means Micron is a bad investment. It means it's a cyclical stock wearing an AI costume, and you should size it like one. If you want AI exposure without betting on where memory prices sit in eighteen months, a total market fund gives you Micron at its actual weight, along with a few hundred other companies that don't need the cycle to turn kindly.

If you do buy it, know your reason, write it down, and decide in advance what would make you wrong. That discipline is what separates investing from hoping.

As always, this is my perspective, not professional financial advice.