Deck
SanDisk designs and manufactures NAND flash memory — the chips inside phones, SSDs, and AI servers — produced through a 49.9% Japanese fab joint venture with Kioxia and sold to hyperscalers, device OEMs, and retail consumers.
Five hyperscaler contracts either anchor a contracted floor or follow spot down — none has run a downcycle.
- What's signed. $42B of minimum contractual revenue across five multi-year hyperscaler agreements, $41.6B of remaining performance obligations, and over $11B of third-party financial guarantees. Customer prepayments grew from $25M (June 2025) to $511M (April 2026) — covering roughly a third of FY27 bit shipments.
- The catch. Management's own language describes the pricing as 'fixed and variable components' — not pure take-or-pay. The variable leg is an explicit concession that part of the contract follows spot down, and zero of the five have run through a downturn.
- Where the gap lives. If the contracted floor enforces through one cycle, the evidence supports $90–120B of EV. If the variable leg collapses with spot, EV reverts toward $40–60B. Today's $293B market cap is consistent with a utility weight north of 60% — a posture the NAND track record cannot yet support.
Q3 FY26 was the most extreme single-quarter print in NAND history.
Quarterly operating margin moved roughly 110 percentage points in four quarters; Micron's record swing was 63 points over two years. Three forces stack: AI scarcity took Q1 2026 NAND contract prices up 85–90% QoQ, mix shifted to higher-priced datacenter SSDs, and depreciation fell 69% from $525M (FY22) to $163M (FY25) as the Kioxia JV fab base ran off prior cost layers. Management is 9-for-9 on quantitative guides since spin; Q4 FY26 is guided to $7.75–8.25B revenue and $30–33 non-GAAP EPS.
Management is trying to convert NAND from spot commodity to contracted utility.
Before: SanDisk sold most NAND on quarter-to-quarter spot pricing. FY23 ran gross margin negative for three straight quarters; the carve-out posted ~$3.9B of cumulative operating losses across FY23–FY25, and the franchise's 'good year' FY22 ROIC was just 7%.
Pivot: Across Q3 and Q4 FY26, management signed five multi-year hyperscaler agreements with fixed-and-variable pricing, customer prepayments, and third-party financial guarantees — covering roughly a third of FY27 bits. Separately, $1.9B of separation debt was retired in twelve months and S&P upgraded the rating to BB+ on May 12, 2026.
Today: Whether this is a regime change in NAND or a partial hedge depends on what the variable leg does when spot rolls. The first cycle test is the Q1 FY27 print in November 2026.
A 55x re-rating, 97% realized vol, and the people inside the building selling — not buying.
- The move. Up roughly 55x from a $36 spin-day open in February 2025 to $1,980 by June 12, 2026; 30-day realized volatility sits at 97% and the 200-day moving average is $573 — fully 245% below spot. Sixteen months of public history is not a multi-cycle base; every level on the chart is tactical.
- Insiders aren't buying. Since the spin, the CTO, CAO, CLO and three directors disposed roughly $6.1M of stock into the $1,400+ price spike — including a $3.5M CAO sale on May 12 and a $3.5M CTO sale on June 1. Zero open-market insider purchases. Stanley Druckenmiller fully exited in Q4 2025.
- The Street is below spot. Mean of 22 sell-side price targets is $1,751 — about 12% under last close, with the bull tail at $2,900–3,250 and a public short call out since February 25, 2026. Reported short interest is 6.2% of float at 0.8 days to cover — decision-relevant but not crowded.
Half the fab joint venture expires on December 31, 2029 — the first renewal between two listed companies.
- What it is. Flash Partners and Flash Alliance — two of the four Kioxia JV arms that produce SanDisk's wafers — terminate on Dec 31, 2029 unless extended. The Yokkaichi piece was already extended in January 2026 for $1.165B in installments to Kioxia, defusing part of the risk but not all of it.
- Why it matters. Kioxia IPO'd on the Tokyo exchange in December 2024 and now negotiates with an independent capital base. By the Yokkaichi precedent, a comparable 2029 renewal could cost SanDisk a mid-single-digit-billion up-front payment no sell-side model carries today, and even a modest split shift takes 50–100bp off through-cycle gross margin.
- Why it caps the bull case. Combined with Samsung's permanent DRAM/HBM/NAND bundle at hyperscalers — Samsung holds 29% of NAND and the full memory stack — the 2029 renewal defines the through-cycle return ceiling no execution can lift.
The structural case is real; the cycle test arrives in late 2026.
- What supports the regime change. $41.6B of contracted revenue and $11B of third-party guarantees that no NAND maker has carried before; net cash of $3.7B against a $6B buyback authorization; 9-for-9 on guides since the spin; hyperscaler AI capex is a demand level no prior NAND cycle had; Kioxia JV co-control prevents the unilateral over-build that ended every prior cycle.
- What cuts against. A 78% gross margin has no historical analog — the franchise's through-cycle median is closer to 25% and FY23 trough was 7%; NBM contracts explicitly contain a variable leg that follows spot; the depreciation reset and customer prepayments produced roughly $1.13B of non-repeating cash tailwinds; six senior officers and directors sold into the spike with zero open-market buys; the 22-analyst consensus target sits 12% below last close.
- What decides it. Q4 FY26 (August 2026) and Q1 FY27 (November 2026) are the first prints likely to face a sequential ASP roll. Gross margin holding above 60% with spot down 20% supports the contracted-floor read; a step-down of more than 15 points in two quarters refutes it.
- The range of outcomes. A bear sum-of-parts compressing peer multiples to 12–14x on a $32–36 normalized EPS implies $450/share; a bull contracted-utility re-rate at 22x $110 normalized EPS clears $2,800; the Street midpoint sits between, at $1,751.
Watchlist to re-rate: Three things to track over the next six months: NBM contract count (does it cross eight by mid-2027?), consolidated gross margin behavior after the first sequential ASP decline, and the contract-liabilities line — a sequential decline before new signings would be the bear's smoking gun.