Sandisk's $42 billion contract floor has bought the bull case three times trailing revenue; the bear trades it for the multiple

Sandisk Corp (SNDK $1,279.96 +25.99%) re-rated from a post-spin name into the cheapest pure-play NAND name in the peer set on EV/Revenue and EV/EBIT, below Western Digital and Seagate on both metrics and below Arista on P/E, on the back of one quarter: fiscal Q3 2026 datacenter revenue of $1,467 million against $440 million the prior quarter (the Q3 2026 results), driven by five multi-year contracts, three signed in the quarter alone, with the three Q3-signed deals carrying roughly $42 billion of minimum contractual revenue and all five deals backed by more than $11 billion of customer financial guarantees. The thesis is that this contract structure, not a spot-price cycle, is what underwrites the 78.4% non-GAAP gross margin and 70.9% non-GAAP operating margin Sandisk printed in the same quarter (the Q3 2026 call), and that the supply discipline now visible at Samsung, SK Hynix and Micron (Astute Group, 2026-03-04; eeNews Europe, 2026) makes those contractual floors stickier than the historical NAND cycle would suggest. The bear case is that the deal economics depend on AI capex holding, and that the 45.3% drawdown from the 52-week high of $2,335 already prices in a meaningful slip on a 14.1× EV/Revenue and 44.4× P/E multiple.

The deal structure is being read as a cycle, but the contracts look more like a floor

Price performance vs peers (Computer Hardware and cross-industry comps)

The market is treating Sandisk as a leveraged NAND call option: the share price is sitting 45.3% below the $2,335 52-week high, on a +438.5% year-to-date return, after a quarter in which datacenter revenue went from $440 million to $1,467 million sequential (the Q3 2026 results). That shape, parabolic, then half off, reads like the front end of a memory cycle that has already peaked, and the bear frame is straightforward: spot pricing rolls over, contract minimums have less floor value than they look, and the multiple compresses.

The read-through cuts the other way at the adjacent listed names. Western Digital Corp (WDC $544.84 +2.21%) is the closest structural comparable — the other half of the split, riding the same datacenter demand from the HDD side — and printed non-GAAP gross margin of 50.5% on $3.337 billion in the same April window, up from 46.1% the prior quarter, with a guided FQ4 FY2026 of $3.65 billion at 51–52% gross margin (FY2026 outlook). Dell Technologies Inc. (DELL $404.81 +9.51%) raised its FY27 AI-Optimized Servers guide from roughly $50 billion to roughly $60 billion in a single quarter, on $43,842 million of Q1 FY27 revenue and a 214% jump in non-GAAP EPS (the Q1 2026 call). The pricing environment is not fading at the customers writing the checks.

The tension inside the tape is that Sandisk has converted a spot-cycle profile into a contracted one. Management has signed five multi-year partnerships, three in the fiscal third quarter alone, with fixed-and-variable pricing, quarterly volume commitments that step up over contract life, and the longest contract extending to five years (the Q3 2026 call). The drawdown is the market's way of saying the contracts might not be as binding as they look; the evidence at the listed names most exposed to the same AI capex says the demand is still binding.

What management teams are saying

Sandisk's CEO frames the shift bluntly: "[w]e're finally, for the first time in decades in this business, getting to the point where the value of our technology is getting recognized" (the Q3 2026 call). The reading is that NAND's pricing reset isn't a spot quirk but a contracted floor: five multi-year deals with $11 billion in financial guarantees, variable pricing that lets Sandisk capture upside, and the longest tenor at five years. The adjacent names tell the same supply story from the other end.

At the lithography layer, ASML's Christophe Fouquet says memory customers are "sold out for 2026 and their supply constraint will last beyond 2026," with the 2026 revenue band lifted to €43–€45 billion in July from €36–€40 billion in April (the Q2 2026 6-K). SK Hynix is funding its P&T7 advanced-packaging plant for AI memory through 2032, and Samsung, per Citi via Reuters, saw NAND contract prices up roughly 53% Q/Q in Q2 2026. The constraint is being absorbed by every memory peer, not just Sandisk.

Where the peers diverge is on contracting. Seagate's Mosley says nearline capacity is "almost fully allocated through calendar 2027," with build-to-order contracts being finalized through fiscal 2027 (the Q3 FY2026 call). Seagate's tool is long-dated contracts. Arista, by contrast, is supply-rationed by 52-week lead times and $8.9 billion in purchase commitments, and CEO Ullal concedes the constraint is "a one or two year phenomena" (the Q1 2026 call). Arista is the demand-led name; Sandisk and Seagate are the contract-led names.

Dell is the demand ledger. Clarke booked $24.4 billion of AI orders in a single quarter, lifting the FY27 AI-Optimized Servers guide to roughly $60 billion (the Q1 2027 8-K): the buyer's checkbook behind the contracts Sandisk and Seagate are sitting on.

The Q3 FY2026 print moved the scoreboard on every line

Valuation vs peers (Computer Hardware and cross-industry comps)

Performance vs peers (Computer Hardware and cross-industry comps)

Sandisk management guidance, from the 8-K filed 2026-04-30

Datacenter revenue goes from $440 million in the prior quarter to $1,467 million in Q3 FY2026, a roughly 233% sequential lift off a base that had stepped down across three prior stated readings (the Q3 2026 results). Same-quarter a year ago, the line was $197 million; the new print is several multiples of that base. Q4 FY2026 guidance of $7,750–$8,250 million (the Q3 2026 8-K) implies management expects the contracted volume to carry the next quarter, not spot re-acceleration.

Edge revenue prints $3,663 million for Q3 FY2026 against $1,678 million the prior quarter, a larger absolute step than datacenter, and a $927 million base a year ago (the Q3 2026 results). The segment is the largest single line in the quarter and explains why the headline 97% sequential revenue jump lands even as datacenter takes the loudest share of the commentary.

Consumer contracts 10% sequential to $820 million from $907 million, against $571 million a year ago. Outside of datacenter, CEO Goeckeler said on the call, "we're seeing some contraction in the market just because of unit decline" with a 2027 bounce-back expected (the Q3 2026 call). Consumer is the line the bull case has to defend least, and the line the bear case can most plausibly blame on mix erosion if pricing softens.

On the product split, FY2025 product-line revenue of $7.4 billion breaks into Client Devices $4.1 billion (56%), Consumer $2.3 billion (31%), and Cloud $960 million (13%), versus FY2024 shares of $4.1 billion / $2.3 billion / $325 million (FY2025 results). Cloud's share roughly tripled year over year inside a flat-to-up total: that is where the contracted volume first surfaced in the issuer's own tagging. Read against the operating lines above, this is the same revenue cut a second way — Client Devices is still 56% of the FY2025 total, the anchor that keeps Sandisk from being a pure AI-momentum name even as the datacenter line prints the quarter's headline.

At 14.1× EV/Revenue and 34.6× EV/EBIT, Sandisk trades below Western Digital at 14.5× and 47.9× and below Seagate at 15.4× and 54.6×. On P/E, Sandisk's 44.4× sits above Dell at 34.0× and Western Digital at 29.7×, but below Arista at 58.7× and Seagate at 70.5×. Dell's 2.2× EV/Revenue sits in a different gear because of scale. The comparison carries a caveat the bull should press and the bear should answer: the 14.1× is struck on trailing twelve-month revenue of $13.2 billion, a base Q3's $5,950 million and the $7,750–$8,250 million Q4 guide leave far behind, so the multiple the market is arguing about is already stale. The Q4 FY2026 guide of non-GAAP gross margin 79–81% at non-GAAP diluted EPS $30.00–$33.00 against ~158 million shares (the Q3 2026 8-K) is the benchmark the multiple has to defend next.

Adjacent reporters point the same way Sandisk does, only louder

Arista reported more recently than Sandisk: $2,709 million in Q1 2026, up 35.1% on the same quarter a year earlier, with the deferred-revenue balance at $7.7 billion (the Q1 2026 10-Q): meaning the contracted AI networking pipeline Sandisk's datacenter line rides on is being booked by the switch supplier, not assumed. Ullal frames the supply constraint as a one or two year phenomena on the call, with 52-week lead times the operating norm. Sandisk's premium pricing sits downstream of that bottleneck; if Arista's lead times extend, Sandisk's contractual floor stays in place. If they compress, Sandisk's pricing reset meets real supply.

Dell is the buyer's checkbook. Q1 FY27 AI orders of $24.4 billion against $16,132 million of AI server revenue, and a FY27 guide lifted to ~$60 billion from ~$50 billion (the Q1 FY27 8-K), confirm the hyperscaler-spend base Sandisk's $42 billion of minimum contract revenue is being written against. The transmission is direct: Dell's AI servers consume Sandisk's enterprise SSDs, and Dell's guide is now twice the rate implied at the start of the fiscal year.

The implication for Sandisk is conditional: the demand signal is corroborated, but the gross-margin premium is specific to the contract structure, not the cycle.

Seagate is the cross-check on contract-led demand. Mosley says near-line capacity is almost fully allocated through calendar 2027, with build-to-order contracts being finalized through fiscal 2027 (the Q3 FY2026 call). The HDD peer has the same long-dated contract posture.

Cash on hand, zero debt, and a $6 billion buyback reset the return profile

Sandisk balance sheet, as of Apr 3, 2026 (10-Q filed May 1, 2026)

The Q3 FY2026 10-Q shows the cleanest capitalization in the NAND set: $3.7B cash, $0 short-term debt, $0 long-term debt, $13.8B of stockholders' equity against $17.1B of total assets (the Q3 2026 10-Q). Inventory of $2.2B and accounts receivable of $2.7B are the only meaningful working-capital asks, and both are modest against quarterly revenue in the $5,950 million range. There is nothing to refinance and no covenant to defend.

The Q3 FY2026 8-K then authorized a $6 billion share repurchase program, funded out of operating cash flows (the Q3 2026 8-K). CFO Visoso framed it as the third leg of a stated priority sequence, invest in the business, reach a net cash position, then return cash, and confirmed on the call that the buyback will be paced to cash generation (the Q3 2026 call). Adjusted free cash flow of $2,955 million on Q3 FY2026 alone covers roughly half the authorization in a single quarter; nine-month operating cash flow of $4,545 million from the 10-Q does the rest in two (the Q3 2026 10-Q).

Capital deployment this fiscal year has been disciplined on the supply side first: a $972 million equity investment in Nanya, settled April 8, 2026 (the Q3 2026 10-Q), and the Kioxia JV extended through December 2034 (the Q3 2026 call). Those commitments preceded the buyback and were sized against the contractual demand the five multi-year partnerships are converting into revenue.

Five conditions hold, or the gross-margin floor walks

The contract structure underwriting Sandisk's gross-margin step has to clear five tests at once, and each one is something the bear case can pull on.

Datacenter demand at hyperscalers has to hold its pace. Management raised its calendar 2026 datacenter growth view to the mid-70s percent range, after guiding progressively lower views three months earlier, with demand described as outpacing supply through the end of calendar 2026 and beyond. The corroborating buy-side read is real: Dell booked $24.4 billion of AI orders in a single quarter and lifted its FY27 AI-Optimized Servers guide to roughly $60 billion from roughly $50 billion at the start of the fiscal year. If that order book softens, the contracted volume underneath Sandisk's $42 billion of minimum contractual revenue starts to look like the optimistic end of a range rather than a floor.

NAND pricing discipline has to persist across the four major suppliers. Astute Group reported contract NAND prices jumping into Q1 2026, and Citi via Reuters has NAND up roughly 53% Q/Q in Q2 2026 at Samsung, the global pricing setter. eeNews Europe carried TrendForce's projection of 33–38% NAND contract-price gains. Omdia/Briocean flagged joint production cuts at Samsung, SK Hynix, Kioxia and Micron in the second half of 2025. If any of the four breaks discipline, Samsung in particular, the variable-pricing leg of Sandisk's five multi-year deals compresses, even with the fixed minimums in place.

Contract economics have to behave as written. CFO Visoso said the three Q3-signed contracts carry approximately $42 billion of minimum contractual revenue, with the five deals backed by more than $11 billion of financial guarantees and roughly $0.4 billion of prepayments on the Q3 balance sheet. CEO Goeckeler said customers are putting up billions in collateral that survives the contract life, with non-performance triggering immediate compensation. If prepayments unwind or contracts get renegotiated, the prepay-funded supply advantage disappears. Management has also said these NBMs account for over a third of FY2027 bits, so any contract cancellation directly hits fiscal 2027 mix.

Capital allocation has to stay disciplined. The $6 billion repurchase authorization announced April 30, 2026 is funded from operating cash flows. The $972 million Nanya equity investment paid on April 8 and the Kioxia JV extension through December 2034 preceded the buyback and consume cash first. Q3 adjusted free cash flow of $2,955 million covers roughly half the authorization in one quarter; nine-month operating cash flow of $4,545 million covers it in two. If a large M&A bid or a Kioxia-related capital call pulls cash forward, the buyback pace slows and the return-of-capital story degrades.

Valuation has to tolerate any slippage. At 14.1× EV/Revenue and 34.6× EV/EBIT, the multiple sits below Western Digital's 14.5× and 47.9× and below Seagate's 15.4× and 54.6×, but the absolute level is high and the company is priced for the contracts to deliver. The Q4 FY2026 guide of non-GAAP gross margin 79–81% at non-GAAP diluted EPS $30.00–$33.00 on roughly 158 million shares is what the multiple is priced to. A two-point gross-margin miss or a $1 EPS shortfall reframes the story as expensive, and the share is already 45.3% off the 52-week high of $2,335.

The contracts pay for the bear to be wrong, but the multiple pays for the bull to be right

The bull case is anchored to a number management has put on the record: minimum contractual revenue of approximately $42 billion across the three Q3 FY2026 new business models, with more than $11 billion of financial guarantees and roughly $0.4 billion of prepayments on the Q3 balance sheet (the Q3 2026 call). The figure says the contract floor is roughly three times trailing twelve-month revenue of $13.2 billion, and the prepay-funded supply advantage is already on the cash-flow statement. The bear case is anchored to a different number: the share is 45.3% off the $2,335 52-week high on a 14.1× EV/Revenue and 34.6× EV/EBIT multiple, against the guided Q4 FY2026 non-GAAP gross margin band of 79–81% at non-GAAP diluted EPS of $30.00–$33.00 (the Q3 2026 8-K). Each side is holding the other's collateral: the bull leans on the contract floor, the bear on the multiple. The bear's strongest point, which management has not addressed, is that the $42 billion minimum is stated as a cumulative contract figure rather than an annualized run-rate, and no quarterly run-rate on record is high enough to validate the 79–81% gross margin band beyond management's Q4 guide.

The evidence read together favors the bull, with two caveats. The contracted demand base is corroborated at the buy side: Dell booked $24.4 billion of AI orders in a single quarter and lifted its FY27 AI-Optimized Servers guide to roughly $60 billion (the Q1 FY27 8-K). Arista extended $7.7 billion of deferred revenue across two-year horizons with 52-week lead times (the Q1 2026 10-Q). The supply discipline is corroborated at the supplier side: ASML raised its 2026 revenue band to €43–€45 billion in July, with memory customers sold out beyond 2026. ASML's raised band reflects sold-out memory demand broadly, including NAND customers, though the figure is not NAND-specific (the Q2 2026 6-K). The risk is that the 14.1× EV/Revenue and 44.4× P/E multiple is already priced for the contracts to deliver, and a single-quarter gross-margin miss reframes the trade as expensive on a 45.3%-off-high tape.

The scenarios bound cleanly. The bull walks away with the 14.1× EV/Revenue re-rated toward 15.4× at Seagate or held in line as the guided gross margin band prints over the next two quarters. The bear walks away with the multiple compressing materially if the guided band misses by 200 basis points or NBM contract economics soften on a renegotiation. The monitorable triggers are the Q4 FY2026 print and the FY2026 10-K filing for the first contracted revenue recognition, the next two earnings calls for fiscal 2027 contracted-bits disclosure, Samsung's Q2 2026 release for global NAND pricing discipline, and any disclosure of an NBM renegotiation or cancellation, which management has said would directly hit fiscal 2027 bits, as the decisive bear-case trigger.

Data comes from SEC filings, earnings-call transcripts and Equibles' deterministic datasets. This article is for information only and is not investment advice.