Micron Technology projected first-quarter revenue of $61.5 billion on Wednesday, plus or minus $1.5 billion, clearing analyst estimates of $57.02 billion by roughly 8% and signaling that the AI-linked memory cycle still has room to run.

The Boise, Idaho-based chipmaker guided adjusted earnings of $38.15 per share for the quarter, plus or minus $1, against a consensus of $35.40 derived from LSEG-compiled data.

What did Micron just print for the fourth quarter?

Fourth-quarter revenue more than quadrupled year-on-year to $54.23 billion, comfortably ahead of the $51.07 billion analysts had modeled. Adjusted profit for the period came in at $33.42 per share, versus a $31.61 consensus.

The results land against an industry backdrop in which high-bandwidth memory (HBM) — the stacked DRAM assemblies used to feed Nvidia's accelerators — has shifted from a niche specialty to a strategic chokepoint of AI infrastructure.

Why is HBM the swing variable?

Nvidia's graphics processors dominate advanced AI training and inference workloads. Those processors need HBM to keep their tensor cores supplied with weights and activations. Micron counts Nvidia — the dominant supplier of advanced AI processors — among its key HBM customers, and sits as one of three merchant suppliers of the technology, and the only major US-based producer.

Order intake for current-generation HBM has run well ahead of installed capacity, a dynamic the company has flagged repeatedly in prior quarters. That backlog overhang now extends into next-generation HBM4, where Micron, SK Hynix, and Samsung all compete for qualification slots at Nvidia and AMD.

The tight supply picture has spilled into adjacent product lines. Conventional DRAM and NAND contract pricing firmed through the quarter, lifting average selling prices across Micron's broader portfolio.

How do the segment economics stack up?

• HBM share of mix: rising mechanically as wafer capacity reallocates from standard DRAM
• Conventional DRAM and NAND: smaller unit volumes but materially higher ASPs
• NAND specifically: pricing has tracked AI-server storage demand alongside hyperscaler capex

What does the Q1 outlook imply?

Stripping the guidance:

• Sequential revenue growth of roughly 13% at the midpoint, from Q4's $54.23 billion
• Adjusted EPS up about 14% from Q4's $33.42 figure
• Implied gross margin trajectory consistent with continued HBM weighting

The $4.50 differential between guided EPS and consensus — versus a typical one-quarter print of under $1 — points to unusually high conviction from buy-side analysts feeding into the LSEG aggregation.

What are the downstream read-throughs?

Two Korean competitors will test the read-through in coming weeks. SK Hynix, widely seen as the HBM volume leader for Nvidia's current-generation parts, has guided to continued tightness through 2027. Samsung's memory division has trailed on HBM qualification but is pouring capital into new capacity.

Server OEMs and hyperscalers — the next customer tier downstream — face the prospect that memory pricing now anchors a larger share of total system bill-of-materials cost than at any point in the prior DRAM cycle. AI server designs appear to lock that structural shift in for at least the next product generation.