Micron’s AI Memory Machine Keeps Humming—But the Tape Wants Even More

Micron (MU) delivered the kind of “beat/raise/reassure” quarter bulls were hoping for—and then some. Fiscal Q4 revenue rose 22% QoQ and 46% YoY to $11.3B, topping the Street’s ~$11.1–$11.2B bogey and above August’s raised guide. Gross margin expanded to 45.7% (from 39% in Q3) and EPS printed $3.03, comfortably ahead of the ~$2.75–$2.80 consensus. For FY25, revenue hit a record $37.4B (+49% YoY) with 41% gross margin (+17 pts YoY). The balance sheet remains sturdy ($11.9B cash; $14.6B debt), and operating cash flow was $5.7B in Q4.

The guide pushes the story forward: Q1 FY26 revenue $12.5B ±$300M, GM ~51.5%, and EPS $3.75 ±$0.15—a clear step toward (and through) the 50% GM threshold bulls wanted mapped out. Capex will rise from $13.8B in FY25 as Micron adds DRAM/HBM capacity, yet management still expects stronger free cash flow in FY26.

The engine: DRAM, skewed to AI

The revenue mix shows where the power is. DRAM was $9.0B (79% of Q4 sales), up 27% QoQ; ASPs climbed low double-digits with bit shipments up low teens. NAND contributed $2.3B (20%), up 5% QoQ as better pricing offset lower bits.

Nowhere is that AI pull clearer than the data center, which accounted for 56% of company revenue in FY25 with 52% GM. HBM revenue reached ~$2B in Q4—an ~$8B run-rate—as HBM3E ramps into leading AI accelerators. Micron says HBM share is on track to match overall DRAM share by this calendar quarter. The roadmap is intact: HBM4 samples are in customer hands, posting >2.8 TB/s bandwidth and >11 Gbps pin speeds. For HBM4E, Micron is partnering with TSMC on customizable base logic dies—a lever for margin uplift. The customer roster is now six, and most of 2026 HBM3E supply is already priced and allocated; the rest is being negotiated, with confidence in selling out 2026 HBM supply.

Beyond HBM, the company is leaning into LPDDR5 for servers (revenue up >50% QoQ; Micron is the sole supplier to Nvidia’s GB platform) and positioning GDDR7 (>40 Gbps) for future AI systems. On the storage side, Micron launched G9 NAND and PCIe Gen6 SSDs aimed at AI inference use cases (KV cache tiering, vector databases) and high-capacity needs.

End-market color: PCs firm, mobile steady, auto/embedded accelerating

• PCs: Outlook for 2025 units nudged up to mid-single-digit growth (Windows 10 sunset + AI PC). Micron posted record client SSD revenue and began shipping 16Gb 1γ DDR5 and G9 SSDs.
• Mobile: Units remain low-single-digit growth, but DRAM content is rising—about a third of Q2 flagships shipped with ≥12GB memory, a mix expected to increase with new launches. Micron qualified 1β LP5X (16/24GB, 10.7 Gbps). It ceased new mobile-managed NAND development to focus capital on higher-ROI areas.
• Auto & Embedded: Q4 revenue +27% QoQ, 31% GM, powered by ADAS, in-cabin AI, and emerging “physical AI” (drones/robotics/AR). Older-node DRAM (D4/LP4) remains tight; Micron is investing in Virginia to support long-lifecycle demand.

Supply, nodes, and the cost curve

On technology, 1γ DRAM reached mature yields 50% faster than the prior node and has already shipped to a hyperscaler; 1β will increasingly support the HBM ramp in 2026. G9 NAND is scaling in both TLC/QLC, with enterprise QLC now qualified. Management highlighted low industry inventory, constrained node migration (owing to extended D4/LP4 EOL support), and longer lead times/costs for new wafer capacity as reasons DRAM supply should remain tight into 2026, with NAND conditions also strengthening.

How the print stacked up to the bar

After a ~30% rally since Sept. 4 (roughly $120 → ~$164), the market wanted more than an August pre-announce. It got it. MU’s $11.3B / 45.7% / $3.03 beat both the raised guide and the Street; the Q1 outlook ($12.5B / 51.5% / $3.75) sits well above the working $11.8B / ~$3.00 bogey. Critically, management reaffirmed tightening DRAM and gave the HBM4/HBM4E cadence and 2026 allocation clarity investors wanted, helping to address 2026 ASP/yield worries and the three-supplier HBM dynamic.

Risks & watch items

• HBM execution: yields, packaging throughput, and power/permits for capacity adds remain key; any broad 2026 HBM ASP compression would dent the 50%+ GM narrative.
• NAND volatility: consumer-led NAND demand is still uneven even as data center mix improves.
• Macro/tariffs: the guide excludes potential new tariff impacts.
• Capex creep: FY26 spend is rising; investors will watch capex/sales discipline vs. FCF trajectory.

The tape’s verdict—so far

The shares spiked to ~$175 on the print and guidance before cooling as fast money digested the run-up and rotated out of winners; MU recently traded around $169, still handily above pre-earnings levels. Given the past month’s re-rate, some fade is unsurprising: the quarter largely validated the bull case (tight DRAM, HBM sold-through, margins marching past 50%), but the bar keeps moving. From here, progress on locking HBM4 volumes/pricing, continued data center mix gains, and node cost execution will likely determine whether MU’s next leg higher is a sprint or a grind.

Bottom line: Micron showed real operating leverage to AI memory—delivering a clean beat, a convincing raise, and credible line-of-sight to structurally higher margins. With DRAM tightness, HBM momentum, and PC/auto tailwinds, FY26 starts on strong footing. The stock’s post-print wobble looks more like positioning and profit-taking than disappointment with the fundamentals.