A Blowout Quarter That Signals a Bigger Shift
Micron recently delivered one of the strongest quarters ever seen from a memory company. The results were impressive not only on their own merits but also because the guidance came in well above the company's own previous guidance and far above what consensus had forecast. The interpretation here is that this was not merely a single outstanding quarter but a genuine proof point that the memory semiconductor space is at the beginning of a re-rating cycle.
Historically, memory semiconductor names have traded at incredibly low price-to-earnings multiples relative to the broader semiconductor sector — and for good reason. They have long been subject to severe boom-and-bust cycles, which justified the discount investors applied to them. What appears to be happening now is a structural shift: the fundamentals supporting memory names have changed, even if the share prices have not yet caught up to that reality.
Despite the strong earnings, Micron's stock pulled back — falling nearly three and a half percent on the day of this discussion. The view here is that this pullback and the accompanying volatility are being driven largely by positioning rather than by anything in the fundamental case for memory. In other words, the selling reflects how investors are managing their exposure, not a deterioration in the underlying business.
Has the Cyclical Thesis Fundamentally Changed?
Memory has traditionally been viewed as one of the most cyclical sectors in all of technology, defined by the recurring boom-and-bust pattern. The question, then, is whether the most recent Micron earnings and guidance have fundamentally altered that investment thesis.
The answer is yes — at least within the next roughly 24-month window. The forward multiples currently being applied to these stocks do not account for the new dynamics; instead, they are anchored to the old way that memory cycles used to work. The market is, in effect, still pricing these companies as though the historical pattern remains intact, when the evidence suggests it has broken.
The Role of the Korean Investment and the Pullback
Part of the recent share-price weakness is attributed to unrest and unanswered questions surrounding a major new Korean investment. South Korean heavyweights are preparing to invest more than 500 billion dollars into new production facilities. This raises a critical question: what does that massive capacity addition do to the favorable position that Micron, SK Hynix, and Samsung currently enjoy — a position built on severely limited inventory? Investors are uncertain whether these investments will eventually address the supply constraints and, by extension, soften the pricing power these companies now hold.
The key consideration, however, is timing. Those investments fall into the 2028–2029 timeframe before they even come online and begin to have any impact. As a result, even though a huge wave of investment is coming from the two South Korean heavyweights, it does little to fix the supply challenges facing the industry in 2026, 2027, 2028, and 2029. The long-term outlook over the next 12 to 24 months remains unchanged by these announcements.
On the question of whether investors should welcome this new capacity or fear it as the seed of the next downturn: after such an incredible run in the stocks, it is reasonable for investors to reassess their positioning, particularly during what amounts to an "air pocket" in the AI space now that Micron's earnings are behind us. Over the longer term, the added capacity is viewed as a positive thing. But it does not change the nearer-term outlook, precisely because the new supply will not arrive in time to relieve the current squeeze.
Margins That Rival or Exceed Nvidia
One of the most striking elements of Micron's results was the margin figure. The number — roughly 85 percent — was so extraordinary that it warranted leading the earnings coverage with the margin first, an unusual departure from typical earnings analysis. Memory companies are now producing operating margins that rival, and in some cases outperform, those of companies like Nvidia.
How sustainable is this? Because of supply constraints that are likely to persist not just for this year but for multiple years, margins are expected to remain higher than average going forward. That said, 85 percent is incredibly high, and there is probably some room for it to come down over time. But the structural drivers supporting elevated margins are durable.
The most fascinating illustration of the supply problem is this: even if a company wanted to build a brand-new memory semiconductor factory today, it would not be able to equip it — because the manufacturing equipment itself is sold out through the end of this year and into next year. This captures the core dynamic of the moment. Demand has skyrocketed because of the AI data center buildout, while the supply side has remained very, very low, held back by years of caution born from the painful boom-and-bust cycles of the past.
Additional Pressures: Apple and China
Beyond the South Korean investment questions, another factor weighing on the sector is uncertainty around Apple's sourcing strategy — specifically, the prospect of Apple trying to source memory from China. This is adding to the noise around memory stocks. The consumer side of demand is also relevant here: none of that consumption is being slowed down, which reinforces expectations that prices will remain high and that demand on inventory will stay very strong. Yet from the standpoint of the fundamental long-term thesis, nothing observed in the volatility of the last couple of days has changed the underlying picture.
High Bandwidth Memory and Permanent Scarcity
High bandwidth memory (HBM) is a part of the story that is arguably being underdiscussed and underappreciated. HBM is sold out years in advance. Compounding the problem, the industry cannot get the equipment, cannot get many of the necessary component pieces, and cannot get all the chips involved.
Is this scarcity becoming a permanent fixture of the market? HBM uses a great deal more wafer capacity than standard DRAM, which intensifies the strain on available production. The scarcity is about as permanent as one can reasonably project — looking out into 2028 and 2029. The investments discussed earlier fall within that very timeframe before they even come online and have any impact, meaning the results of today's shortage are already beginning to show. The expectation is that prices will remain high and that demand on inventory will stay elevated for the foreseeable future.
Could Today's Shortage Become Tomorrow's Glut?
A natural concern with any supply-constrained market is whether new capacity could eventually overshoot demand and turn a shortage into a glut. Realistically, how much new supply would be needed for that to happen — and is it even possible?
Not within any short timeframe. Demand is still increasing. Even though the new Korean investment reportedly aims to double DRAM capacity over the next five years, that effort would still leave the market in a deficit. In other words, even a doubling of DRAM capacity is not enough to outpace the growth in demand, which underscores just how severe and durable the current imbalance is.
The Bottom Line
The overarching picture is of a memory sector in transition. AI-driven demand has fundamentally reshaped the economics of memory, driving margins to levels that compete with the most celebrated names in semiconductors and breaking the historical boom-and-bust pattern, at least over the next 12 to 24 months. Supply cannot respond quickly — equipment is sold out, HBM is committed years ahead, and even hundreds of billions of dollars in new Korean fabrication investment will not arrive in time to relieve the squeeze before the end of the decade, nor enough to erase the deficit even when it does. The recent stock volatility, including Micron's pullback, is best understood as a function of investor positioning and a handful of open questions — around Korean capacity and Apple's China sourcing — rather than any crack in the long-term fundamental case.
