Sandisk SNDK has gone from forgotten storage spin-off to one of the loudest memory chip names AI investors are watching. AI training and inference need huge SSD arrays, and Sandisk is now a pure-play on that demand.
The 2026 question is not whether the tailwind is real, but whether SNDK is the right way to own it.
This piece walks through the spin-off setup, the enterprise SSD cycle, margin mechanics, and how SNDK stacks against Micron and Western Digital.
Sandisk Post Spin-Off From Western Digital
Sandisk began trading on Nasdaq under SNDK in February 2025. The split left Western Digital with HDDs and gave Sandisk the full NAND and SSD business.
For investors, that matters more than the plumbing. SNDK is now a single-product story, easier to model but more cyclical.
What the new SNDK actually sells
The product mix splits into three buckets. Client SSDs for PCs, consumer flash for cameras and cards, and enterprise SSDs for data centers.
Enterprise is the smallest bucket today but the fastest growing. That is where the AI capex story shows up in the income statement.
Why pure-play status changes the multiple
Inside Western Digital, NAND was buried in an HDD-heavy P and L, valued on a blended cycle that rarely worked in NAND's favor.
As a standalone, SNDK gets a cleaner read. Up-cycles compress less, but down-cycles also hit harder.
Enterprise SSD Tailwind From AI Hyperscaler Capex
AI accelerators from NVIDIA and Broadcom need somewhere to park training data and inference checkpoints. That somewhere is high-capacity enterprise SSDs.
Hyperscaler capex guidance for 2026 sits above 500 billion dollars combined, mostly earmarked for AI infrastructure. NAND is a direct beneficiary, not an adjacent one.
Why SSDs are eating into HDD share
Inference workloads need low-latency reads, not just bulk capacity. SSDs win on performance per watt and rack density, which matters when data centers are power-constrained.
Hyperscalers are quietly shifting more of their storage tier from HDD to QLC SSD, and Sandisk's high-density flash is built for that swap.
Order book signals from recent prints
Management has flagged long-term agreements that stretch into 2026 with multiple hyperscalers. That is a different pattern from the usual quarterly spot-price exposure.
Lock-in pricing dampens cycle volatility on the way down, which is the part of the story SNDK bulls underweight.
Want exposure to the AI memory cycle without picking the exact winner? You can hold SNDK and MU side by side on Gotrade in fractional sizes.
Margin Trajectory: NAND Cycle Recovery Mechanics
NAND is a brutal commodity in the down-cycle and a fat-margin business at the top. SNDK is climbing out of a multi-quarter trough that started in 2023.
Why NAND gross margins swing 30 points
Fixed costs for fabs are enormous, so utilization drives everything. When demand returns, pricing power follows within one or two quarters.
Sandisk gross margin troughed in the teens during the 2023 glut. The current cycle has it tracking toward the 30s, with bulls modeling 40-plus at peak.
What could go wrong
Two risks matter. First, hyperscalers could pause AI capex if inference monetization disappoints. Second, Samsung and SK Hynix could over-supply NAND to defend share.
Both are real, but neither has shown up in the order book yet.
SNDK vs MU vs WDC: Which Pure-Play to Own
The three-way comparison comes down to what kind of AI memory exposure you want.
SNDK: cleanest NAND beta
SNDK is the purest NAND play. No HDD, no DRAM, no HBM. If enterprise SSD demand is your AI capex thesis, SNDK is the highest-conviction ticker.
MU: HBM-led, more diversified
Micron sells DRAM, NAND, and HBM. HBM is where the structural AI growth lives, and MU is the only US-listed pure HBM play. We covered this in our Micron HBM analysis.
MU carries less cycle risk than SNDK but also less upside if NAND prices rip.
WDC: HDD with optionality
Post-split, WDC is an HDD pure-play. AI still pulls nearline HDD demand for cold-storage tiers, but the multiple is structurally lower than NAND.
Catalysts To Watch In Q2 2026 Earnings
SNDK's next print is the most important data point for the thesis. Three lines matter most.
Enterprise SSD revenue mix
Watch for the enterprise share of total revenue. A move from low-teens to mid-twenties percent in two quarters would confirm the AI capex pull-through.
Gross margin trajectory and capex guide
A gross margin guide above 35 percent next quarter would signal pricing power is back. Capex guidance also matters, since over-build is how NAND cycles end early.
Conclusion
SNDK is a clean way to own the AI memory cycle, but not the only way. Cleaner exposure comes with higher cycle risk.
For investors who want NAND directly, SNDK is the highest-conviction ticker on US exchanges right now. For more diversified AI memory exposure, MU is the safer pair trade.
If you are sizing AI exposure for 2026, check what you already own via ETFs, then add or trim on Gotrade. You can buy SNDK and pair it with MU in fractional sizes.
FAQ
Is SNDK still a spin-off trade in 2026?
The spin-off completed in February 2025, so the technical setup is over, but the fundamental pure-play thesis is still playing out.
How is SNDK different from MU?
SNDK is pure NAND and SSD, while MU sells DRAM, NAND, and HBM, with HBM being its main AI growth engine.
What is the biggest risk to SNDK?
A pause in AI hyperscaler capex or a NAND over-supply move from Samsung or SK Hynix would compress margins fast.
Does Sandisk benefit from NVIDIA capex?
Yes, indirectly, because NVIDIA accelerators sit inside hyperscaler builds that pair GPUs with enterprise SSDs from Sandisk and peers.
