MU is one of the main global makers of memory semiconductors, the other key producers being the South Korean firms Samsung Electronics and Hynix.

The memory business is in many ways similar to mining, in that both industries are highly capital intensive, their products don’t have many unique characteristics to differentiate one from another, and there are enough makers that customers, particularly of base metals, often have considerable market power. So these businesses–and their stocks–are prone to booms and busts.

We’re now in boom–the midst of an AI-chip generated surge in demand for tech hardware in general and DRAM in particular, with memory prices up by at least 4x over the past 12 months.

The stocks of the big three DRAM makers skyrocketed starting in the second half of last year, as a looming DRAM shortage became apparent, as well as when the Nvidia-Broadcom frenzy that preceded the memory chip craze was beginning to lose steam.

(As it turns out, I bought MU in the portfolio I’m actively managing during 4Q2025 and sold the last of it in June. This was partly due to price, partly because I wanted to reduce my exposure to IT.)

The three manufacturers are already starting to do what commodity product producers always seem to do when they’re flush with cash–expand. Korea has announced it intends to double DRAM capacity through an investment of half a trillion dollars. MU is expanding as well. But it is also asking customers for take-or-pay contracts in order to reserve output from new plant.

Generally, these contracts are agreements requiring a given customer to purchase specified amounts of output at specified prices from the new MU plant. The customer must pay whether it takes output or not, although if it chooses not to take output, it will sometimes be able to get a discount on later purchases. Presumably, MU will use these contracts to get better financing terms from lenders for the new capacity.

Two thoughts:

–the massive increase in capacity planned by the global DRAM industry creates the worry that demand may not keep pace with this new supply–and that any overcapacity that emerges will cause a plunge in DRM prices, and

–the take-or-pay structure conjures up images of public utilities rather than of gold mines, dimming the future possibility of MU having the kind of through-the-roof pricing we’re seeing at present. Arguably, this places a hard cap on future PE multiple expansion–as well as the appeal to more speculative investors.

In any event, all this would argue that the best for all these firms is at least temporarily in the rear-view mirror.