Memory Update

by Brad Gastwirth Global Head of Research and Market Intelligence

Included Below

• Micron’s FQ4’25: what we think they’ll say about end-markets (Data center, PC, Mobile, Auto/Industrial) and product lines (HBM/DRAM, DDR5/DDR4/DDR3, NAND/SSD)

• Our base-case outlook into 1H’26, plus an industry update: pricing trends, what’s driving it, when new capacity likely lands, “what to watch,” and non-consensus 2026 risks/opportunities.

Happy to chat if interested. Best, Brad

Micron FQ4’25: What they’ll likely emphasize

End-markets

Data center / AI: Expect reiteration that AI demand is the core engine; HBM capacity for CY2025 is sold out, with 2026 demand under LTAs; bit shipments and mix likely flagged as tight/optimized. Watch for color on HBM4 customer milestones and 2026 mix. 

PC / Client: Directionally better vs 1H’25 on AI-PC refresh and DDR5 attach; Micron has previously pointed to constructive pricing and tight inventories exiting FY25. We’d expect cautious optimism into holiday/1H’26. 

Mobile: Gradual recovery narrative (LPDDR5X content, flagship mix) rather than a surge; look for comments on China Android and premium iOS tiers driving density. (Tone inferred from prior calls + market mix.) 

Auto/Industrial: Likely “steady-to-up” on content growth (ADAS/IVI, LP/auto-grade NAND). Micron has highlighted leadership in low-power/data-center LP DRAM; they may echo disciplined supply to industrial. 

Product lines

HBM (HBM3E → HBM4): Expect a remind that HBM revenue grew sharply in FQ3 and that HBM3E is tight into 2026, with HBM4 sampling/shipments progressing for next-gen AI platforms. Any update on HBM4 bandwidth/throughput targets and customer validations will be key. 

DDR5 / DDR4 / DDR3: Messaging likely: DDR5 up-bias and mainstreaming; DDR4 structurally tight as capacity tilts to HBM/DDR5; DDR3 remains tail-supply. Expect Micron to lean into pricing discipline and mix. (Consistent with raised FQ4 GM outlook tied to stronger DRAM pricing.) 

NAND / SSD: Tone should be firmer: enterprise SSD demand + vendor discipline driving +mid-teens to ~20% contract pricing into Q4; Micron has cited reduced NAND inventories and better mix exiting FY25. Watch for comments on controller availability and enterprise vs client spread. 

Near-term what we expect

FQ4’25 guide already lifted to revenue ~$11.2B ±$0.1B, non-GAAP GM ~44.5% ±50 bps, EPS ~$2.85 ±$0.07; on the call, look for framing around “low inventories + constructive demand = continued pricing improvement,” especially in DRAM. 

Set-up into FY26: Management likely frames supply as tight where they want it (HBM/DDR5), and disciplined in NAND. Street is primed for a positive tone after a big run-up into earnings. 

Our Outlook cue cards (Micron + peers) through 1H’26

HBM3E: Stays firm; flattening risk only as incremental capacity lands and HBM4 mix rises in 2026. Key swing: speed of customer validations and packaging throughput. 

DDR5: Gradual up-bias near-term; normalizes as yields improve and platforms fully transition. 

DDR4: Elevated/choppy given structural undersupply and EoL management; “panic buy” spikes possible.

DDR3: Thin tail, prone to episodic spikes.

NAND/SSD: Up-bias into Q4’25 (enterprise > client) on AI storage builds and vendor discipline; stickiness into early ’26 depends on CSP ordering cadence. 

Industry update (pricing drivers, capacity timing, watch items)

What’s driving pricing now:

1. AI mix pull: Wafer/capex tilt to HBM/advanced DRAM, starving legacy nodes (DDR4/DDR3). 

2. Tight inventories by design: Micron flagged exiting FY25 with tight DRAM and reduced NAND inventories; peers echo discipline. 

3. Enterprise storage restock: CSP/AI builds absorbing high-density NAND; retail deals thinning. 

Capacity timing (high-level):

HBM/DRAM: Incremental supply phases in late ’25 → ’26 (SK hynix M15X pull-ins; Samsung 1c DRAM for HBM4 ramp; Micron HBM4 shipments/sampling already disclosed). Net: HBM tight near-term; relief is staged. 

NAND: Big step-ups are longer-dated (e.g., WD/Kioxia doubling by FY2029), so 2025–early-2026 relies more on discipline than new bits. 

What to watch on/after the call:

HBM4 customer milestones & speeds (any color above “>2 TB/s,” or commentary on 10Gbps asks from GPU customers). 

DDR4 EoL cadence (explicit extensions or exits).

NAND promo depth into holidays vs enterprise strength (how much of the +15–20% sticks). 

Inventory posture (any move away from “tight by design”). 

2026: likely non-consensus swing factors

CXL memory adoption curve: If early CXL pooling at hyperscalers scales faster, DRAM-per-server needs could ease by late ’26, capping commodity DRAM upside while HBM stays premium. (Inference; watch vendor roadmaps + hyperscaler disclosures.)

HBM4 yield/throughput surprise: A fast multi-supplier ramp could loosen DRAM balance sooner; conversely, TSV/base-die or thermal limits could extend HBM3E tightness. 

Packaging whiplash: If CoWoS/2.5D adds overshoot, pressure may first hit older HBM3E ASPs while HBM4 holds premium—creating an intra-HBM spread the street underestimates.

Wafer-bonded NAND ramps better-than-feared: Cost/bit improvements could soften enterprise SSD pricing in 2H’26 despite solid unit demand. 

Policy shocks: Export rules or regional incentives could shift HBM/advanced DRAM flows, creating regional price dislocations even if global supply is adequate.

by Brad Gastwirth Global Head of Research and Market Intelligence