Has SI turned memory into a lasting franchise?

In fiscal 2025, Micron’s revenue was $37.4 billion. In fiscal 2026, it reached $133.2 billion, up 256%, a faster jump than Nvidia managed in any single year recently.

Last quarter alone, Samsung, the world’s largest memory maker, earned an estimated $80 billion in operating profit. Memory makers are now among the biggest beneficiaries of the superintelligence (SI) boom.

Yet the big memory stocks are still down roughly 22% to 65% from their highs, even as profits keep climbing.

Memory has historically been a commodity business, one of the hardest businesses in which to protect profits. But the industry has also passed through several distinct eras since the 1970s, which could be fundamentally changing the economics of these businesses (case in point: Micron's profit margin after production costs hit ~81% in FY2026, up ~90 pp from FY2023).

At the same time, the pattern that ended previous memory booms may also be showing up again. The makers are using their record cash to expand production capacity. New factories are scheduled to arrive from 2027 to 2029. China’s DRAM champion, CXMT, is expanding too, with its share of global DRAM revenue jumping from 4% to 10% in just a year, according to Counterpoint.

The next memory bust does not need SI to stop growing. The last two downturns began with customers working through inventory. Buyers who had stocked up paused orders, and suppliers were hit by falling prices as well as weaker shipments.

With customers putting up billions of dollars in deposits to secure supply, is memory becoming a franchise, or is it positioned for another boom-and-bust cycle that reverts it back into a commodity business?

In our view, this is the key question that determines the trajectory of this industry.

Answering it effectively means building understanding from the ground up: how a memory chip physically holds data, how each kind of memory is priced and sold, and who holds the power along the supply chain.

Each layer changes how the next one looks, all the way up to margins and stock prices, and the view that emerges is very different from the one built on headlines and market commentary.

Thus, our deep dive builds toward the answer in five parts:

The foundation: what memory physically is, and why SI made memory one of the most important bottlenecks in computing today.
Cycle or franchise? Why memory has historically been a brutal business, what SI has and hasn’t changed, and how bargaining power is split between memory makers, chip designers, and their customers.
The players and the map: how the three makers and a rising Chinese rival compete, why storage follows different economics, and how subsidies and export controls shape supply.
Will demand keep up? Who ultimately pays for SI, how much of the spending is borrowed, and the number that decides whether demand can absorb the new factories.
What could sustain or end the boom: how the cycle persists or breaks, and the signals to watch over the coming months.

Along the way, we also cover:

Why Nvidia depends on memory makers it can’t easily replace, and how qualification decides who wins the business.
What Samsung’s HBM comeback means for SK hynix.
Why storage has historically had the harsher cycle, and what that means for storage makers like SanDisk and Kioxia.
How to trace memory demand back to the end customers who ultimately pay for it.
What October’s selloff showed, and what the coverage left out.
Why memory stocks can look cheapest at peak earnings.