Sandisk Bets on AI and Enterprise SSDs: Can It Outpace MU & STX?

Sandisk is scaling BiCS 8 NAND, preparing BiCS 10 and investing in high-bandwidth flash to capture AI-storage demand, with datacenter rising to 38% of its bit portfolio from roughly 12% a year earlier. The company began QLC Stargate revenue in fiscal Q4 2026 and expects datacenter to represent about 50% of NAND TAM in 2026, versus about 30% in calendar 2025. SNDK shares are up 630.2% year to date, while fiscal 2027 consensus EPS of $213.31 implies 200.95% year-over-year growth, though Micron and Seagate remain significant competitors in enterprise SSD and AI-storage markets.

Analysis

SNDK is now a high-duration NAND-cycle expression rather than a clean AI infrastructure proxy. Its rerating leaves little room for a normal NAND pricing reset: incremental enterprise mix can lift blended ASPs and utilization, but hyperscaler SSD qualification cycles are long and customer concentration gives buyers substantial pricing leverage. The key near-term question is whether enterprise SSD gross-margin expansion exceeds the market’s already aggressive expectations, not whether datacenter storage demand grows.

MU is competitively better insulated because its AI revenue stack spans memory categories beyond NAND; this makes it the preferable long leg if storage demand remains strong but enterprise-SSD share gains prove contested. STX is a non-obvious beneficiary of AI data growth because inference, checkpoints, and cold datasets migrate to lower-cost capacity tiers; however, it is more exposed to cloud-capex timing and HAMR execution. Over 6-18 months, SSD adoption may expand total storage workloads rather than fully displace nearline HDDs, supporting a barbell of MU plus STX over a single-name SNDK allocation.

The contrarian issue is valuation and estimate quality. A 630% YTD move paired with minimally changing near-term consensus suggests the share price is discounting a multiyear margin and share outcome before independent evidence of durable contracts, realized QLC pricing, and yield performance is available. A NAND spot-price downturn, higher-than-expected industry bit supply, or one hyperscaler qualification delay could trigger sharp multiple compression within days; conversely, evidence of multi-quarter enterprise backlog and gross-margin upside would invalidate the cautious view over the next 1-3 months.

Key Decisions for Investors

• Initiate a 3-6 month pair trade: long MU / short SNDK in equal dollar amounts. The thesis is broader AI-memory exposure and lower execution concentration at MU versus elevated SNDK expectations; target 15-20% relative outperformance, with a 10% adverse relative-move stop.
• Do not chase SNDK outright after the momentum move. Reassess only after the next earnings release confirms enterprise SSD revenue conversion, gross-margin expansion, and customer commitments; a guidance increase driven solely by NAND pricing rather than mix should not qualify as a bullish catalyst.
• Maintain a small 6-12 month STX long as a capacity-storage hedge against AI data-lake growth, preferably funded against a storage-equipment or high-beta NAND basket rather than MU. Exit if cloud customers defer nearline deployments or HAMR qualification/ramp timing slips.
• Set alerts on NAND contract and spot pricing, SNDK gross-margin guidance, and hyperscaler capex revisions. A sequential NAND-price decline or missed margin guide would favor increasing the SNDK short; sustained contract-price increases plus verified enterprise backlog would falsify that stance.