Everyone thinks SanDisk's 84.6% gross margin is a testament to its technological superiority. The data says otherwise.
I've spent the last three weeks dissecting the on-chain footprints of the NAND flash market—tracking shipment volumes, contract terms, and capacity utilization across the major players. The headline number is seductive: a five-quarter swing from a 22.5% gross margin to 84.6% implies a demand shock of historic proportions. But the real story is not about demand. It's about a supply chain that has been deliberately starved, and a client base that is now locked into a Faustian bargain.
Context: The NAND Flash Market in 2026
SanDisk, the storage arm of Western Digital spun out in 2023, is a pure-play NAND IDM. It designs, manufactures, and sells flash memory chips and SSDs, with a heavy focus on enterprise-grade drives for AI data centers. The company's joint development agreement with Kioxia gives it access to BiCS 3D NAND technology, currently at 218-300+ layers. The product is commoditized at the wafer level, but differentiation comes from controller firmware, power efficiency, and reliability.
The market was in a deep freeze in 2023. NAND revenue collapsed by 40% as hyperscalers slashed capital expenditure and consumer demand evaporated. Every major player—Samsung, SK Hynix, Micron, Kioxia, SanDisk—cut production and bled cash. Then came the AI feeding frenzy. In 2024, training clusters demanded petabytes of high-speed storage for model checkpoints and dataset loading. By 2025, inference workloads exploded, requiring far more storage per server than training. The result: NAND supply, which had been crimped by the 2023 cuts, faced a sudden, insatiable demand.
SanDisk's revenue tripled from $6.7 billion in fiscal 2024 to an estimated $20.2 billion in fiscal 2026. The gross margin spike is real. But the composition of that growth is what should concern every investor.
Core: The On-Chain Evidence Chain
Let me walk you through the data I've been tracking. I built a Python script to scrape public shipment data from SanDisk's quarterly filings, Kioxia's investor presentations, and third-party teardown reports. I also cross-referenced contract announcements from the eight largest hyperscalers and server OEMs. Here's what I found:
- Price contribution vs. volume contribution: SanDisk CFO explicitly stated that two-thirds of the revenue growth came from price increases, not unit shipments. This is a red flag. In a healthy market, volume growth should lead. When price dominates, it signals that the supply chain is constrained, not that demand is structurally expanding.
- Contract lock-in dynamics: Eight clients signed multi-year agreements covering 50% of fiscal 2027 shipments and 66% of fiscal 2028 shipments. These contracts include price floors—a mechanism that protects SanDisk if spot prices fall, but caps upside if spot prices rise further. The contracts are a double-edged sword. They guarantee revenue visibility but also lock in a price ceiling. The data shows that the eight clients are primarily AI hyperscalers (AWS, Azure, GCP, Meta, etc.) and server OEMs (Dell, HPE, Supermicro). The concentration is extreme: the top five clients likely account for over 60% of revenue.
- Capacity utilization: With 84.6% gross margin, SanDisk is operating at effectively 100% capacity utilization. The company has not announced any major greenfield fab expansion. The only new capacity coming online is from YMTC's Wuhan Phase 3, which is expected to add roughly 10% of global NAND output by 2027. SanDisk is not expanding its own wafer capacity. This means the supply squeeze is self-imposed, not a result of technical limitations.
- Depreciation shadow: The current gross margin is inflated because older fabs are fully depreciated. New capacity would carry a 5-7 year depreciation schedule, crushing margins. Management's guidance of 80% long-term gross margin is a coded admission that they expect to invest in new capacity—and that the margin will normalize downward.
Contrarian: Correlation ≠ Causation
The bull case is simple: AI demand is secular, storage-per-server is rising, and SanDisk is the leading supplier of high-capacity enterprise SSDs. The data supports this, but only up to a point.
Here's the contrarian angle: The 84.6% gross margin is not a sign of tech moat—it's a sign of supply chain fragility. SanDisk is benefiting from a temporary supply deficit caused by the 2023 industry-wide production cuts. Those cuts were a response to a demand collapse, but they were synchronized across all major players. Now, as demand recovers, the same players are hesitant to build new capacity for fear of repeating the 2023 glut. This is a classic prisoner's dilemma. If everyone holds back, margins stay high. But if one player (like Samsung or YMTC) breaks ranks and adds capacity, the price war begins.
YMTC is the wildcard. Despite US export controls, it is building its Wuhan Phase 3 with domestic equipment. The Chinese government is subsidizing the capex. If YMTC reaches 10% of global capacity by 2027, the supply-demand balance shifts. The eight clients currently locked into multi-year contracts may have gotten favorable terms, but those contracts are not unlimited. The remaining 34% of shipments in 2028 are exposed to spot market dynamics. If YMTC floods the spot market, prices could crash.
Another blind spot: the HBM factor. SanDisk does not produce HBM (High Bandwidth Memory), which is used directly alongside AI accelerators. Samsung and SK Hynix offer integrated HBM+SSD solutions. Hyperscalers may prefer a single vendor for both memory and storage to reduce complexity. SanDisk's client relationships are strong, but they are not ecosystem-bundled. If the hyperscalers shift to a one-stop-shop model, SanDisk could lose share.
The Next-Week Signal
The next important data point is the quarterly shipment report from SanDisk, expected in late April. I will be watching for two things: first, whether the volume contribution to revenue growth ticks up from the current 33% to above 40%. Second, whether any of the eight clients disclose a reduction in purchase commitments. If volume growth remains price-driven, the supply constraint is not improving. But if clients start reducing their commitments, it signals that the market is topping.
Volume without intent is just digital noise. SanDisk's current performance is a brilliant recovery from the 2023 abyss. But the structural reality is that NAND remains a cyclical commodity business, and the current margin peak is unsustainable. The contracts provide a floor, but they also cap the ceiling. The real question is whether SanDisk can convert its temporary supply advantage into long-term market share gains—or whether it will be caught in the next downturn with an overleveraged balance sheet.
Follow the gas, not the gossip. The on-chain data of NAND shipments tells a story of a market that is artificially tight. When the supply spigot opens, the margin will drain faster than the hype.