- Micron Technology ($MU) has been one of the strongest performers in the semiconductor space, driven by surging demand for high-bandwidth memory (HBM) used in AI accelerators.
- HBM capacity for 2026 is largely sold out under long-term contracts, giving the company unusual revenue visibility for a memory maker.
- Despite the AI-driven upcycle, the stock trades at a discount to many AI-linked peers, suggesting the market remains skeptical of memory’s durability.
- Memory has historically been a boom-and-bust commodity business, which explains persistent valuation skepticism even amid record demand.
- The bull case rests on HBM changing Micron’s earnings profile from cyclical to structurally higher-margin.
Few corners of the semiconductor market have been as polarizing as memory. Micron Technology has ridden the artificial intelligence buildout to record results, yet the stock continues to trade as if investors expect the cycle to turn at any moment. That disconnect — between what the business is delivering and what the multiple implies — is the core of the bull argument.
The HBM Story Is Real
High-bandwidth memory is the key enabler of modern AI training and inference. Accelerators from Nvidia and others require enormous memory bandwidth, and HBM is the technology that delivers it. Micron is one of only a handful of companies globally capable of producing HBM at scale, alongside SK Hynix and Samsung. That oligopoly structure is a meaningful change from the fragmented, brutally competitive DRAM markets of the past.
What makes this cycle different is contracting. Much of Micron’s HBM output has been committed under long-term agreements, with capacity for the coming year largely spoken for. For a company historically at the mercy of spot pricing, that visibility is transformative. It reduces the risk of a sudden pricing collapse and gives management the confidence to invest in additional capacity without fearing a glut the moment demand softens.
Why the Market Remains Skeptical
The skepticism is not irrational. Memory has burned investors before. Every prior upcycle — in PCs, in smartphones, in cloud servers — eventually gave way to oversupply and sharp margin compression. Investors who lived through those cycles carry scar tissue, and they apply a lower multiple to memory earnings than to, say, logic or software earnings. That is the “disbelief” the bulls point to.
There is also a legitimate question about how long AI infrastructure spending can grow at its current pace. If hyperscaler capital expenditure decelerates, demand for HBM could cool, and pricing power could erode. Bears argue that today’s tightness is a temporary supply-demand imbalance rather than a permanent structural shift, and that the market is correctly discounting a normalization of margins.
The Valuation Math
Where the debate gets interesting is valuation. If Micron’s earnings are genuinely higher-quality and more durable than in past cycles, the current multiple looks too low. If they are simply peak-cycle earnings destined to mean-revert, the multiple is appropriate — or even generous. The entire investment case hinges on which of those two worlds we are in.
The bulls’ answer is that HBM is not a commodity in the traditional sense. It requires advanced packaging, tight integration with customers’ designs, and years of qualification. Switching costs are high, and the technical barriers are steep. That argues for more stable pricing and stickier relationships than legacy DRAM ever enjoyed.
None of this guarantees a favorable outcome. Competition could intensify, yields could improve industry-wide and loosen supply, or AI spending could disappoint. But the market’s refusal to award Micron a premium multiple despite record demand and sold-out capacity is precisely the inefficiency the bulls are betting against. Whether that bet pays off depends on whether HBM truly rewrites the rules of memory — or merely delays the next downturn.
Comments are closed.