This is investment research, not personal financial advice.

Sandisk (NASDAQ:SNDK) gained 7.39% on Friday, from $1,528.11 to $1,641.11, after its Investor Day set out an FY2028-FY2030 model built around an 80% non-GAAP gross margin and a 50% adjusted free-cash-flow margin. The move added about $16.8 billion of market value in one session. It followed a 13.67% rise on Thursday, taking the two-day increase to 22.08%, or roughly $44.2 billion (Yahoo Finance 2026; StockAnalysis 2026).

The presentation offered more than an AI storage slogan. Eight customers have signed about ten long-term supply agreements with $91.1 billion of remaining performance obligations at floor pricing. Those agreements cover half of fiscal 2027 bits and roughly two-thirds of fiscal 2028 bits. Yet the explicit financial guarantees total $16.5 billion, equal to 18% of RPO, and the contracts have never met a NAND downturn. Friday's second move looks proportionate to the added detail and an analyst rerating. The full two-day response is more demanding: it already capitalizes a large part of management's claim that customer contracts can turn peak margins into durable owner cash.

The rerating arrived in two installments

Investor Day began during Thursday's session. Sandisk closed 13.67% higher on 22.1 million shares, then added another 7.39% on 20.4 million shares on Friday. The Philadelphia Semiconductor Index fell 0.31% on Friday, while Western Digital rose 4.41% and Micron rose 2.30%. Memory sentiment helped, but most of Sandisk's move was company-specific. Friday also brought a favourable JPMorgan research action after a restricted period, giving the long-term model a second distribution channel beyond the presentation itself (CNBC 2026; Yahoo Finance 2026).

The two sessions changed the market's time horizon. Fiscal Q4 had already shown what scarcity can do: revenue reached $8.97 billion, GAAP gross margin hit 84.6%, and adjusted free cash flow was $5.04 billion. Management said about two-thirds of the sequential revenue increase came from price and one-third from volume. Investor Day asked the market to treat that margin structure as more than a quarterly peak (Sandisk 2026c).

That distinction explains the size of the reaction. A 22% rerating is not payment for one quarter's earnings. It is payment for lower perceived cyclicality. At the Friday close, Sandisk's market capitalization was $244,525 million ($244.53 billion). That figure is 12 times FY2026 revenue and 28 times company-adjusted FY2026 free cash flow. The valuation becomes intelligible only if a large portion of the recent cash margin survives lower spot prices.

Eight customers sit behind ten agreements

The contract count needs care. Sandisk's 5 August results said it had signed five additional NAND business model, or NBM, agreements after announcing five in April. Three were with new customers and two expanded existing arrangements. The Investor Day deck therefore refers to eight individual customers across about ten agreements, rather than eight contracts (Quartr 2026; Sandisk 2026c).

At floor pricing, total expected contract revenue is $93.9 billion and remaining performance obligations are $91.1 billion. Contract terms range from one to five years, with a weighted average above four years. The agreements specify monthly product mix, volume and price, using fixed and variable terms inside floors and ceilings. Their reach is material: 50% of fiscal 2027 supply and about 67% of fiscal 2028 supply (Sandisk 2026a; Quartr 2026).

This changes the operating unit. Traditional NAND suppliers decide capital years ahead, then negotiate price against customers with short order books. NBM customers commit to future bits while supply is tight. Sandisk gains visibility on volume and a lower boundary on price; customers gain priority access. The arrangement can reduce the amplitude of both shortages and gluts on covered supply.

It does not remove operating risk. Sandisk still has to manufacture qualified bits at the promised mix and schedule. Customers may dispute specifications, invoke contract clauses or face their own demand shocks. Public materials omit names, credit profiles, termination rights, force-majeure terms, remedies and guarantee-draw mechanics. The purchase obligations may be enforceable beyond the explicit guarantees, but an outside reader cannot calculate that protection from the deck.

$91.1 billion is an obligation, not a cash pile

RPO is the minimum revenue that Sandisk associates with unfulfilled commitments at floor pricing. It is not cash on the balance sheet, and it is not profit. At fiscal year-end, $2.5 billion of customer prepayments and deposits sat in cash. Most of the remaining guarantee package is held by or supplied through third-party financial institutions (Quartr 2026; Sandisk 2026c).

That separation matters for owner cash. FY2026 operating cash flow was $11.67 billion. Subtracting $177 million of reported property and equipment purchases gives $11.49 billion of mechanically computed free cash flow. Sandisk's own adjustment then removes $275 million of Flash Ventures activity and $2.48 billion of NBM prepayments and deposits, producing adjusted free cash flow of $8.74 billion. Subtracting $232 million of stock compensation leaves an owner-cash proxy of about $8.51 billion, or 42% of revenue (Sandisk 2026c).

A stricter bridge also recognizes tax timing. FY2026 tax expense was $1.58 billion, cash taxes were only $146 million, and income-tax payables increased by $1.37 billion. Removing that payable build reduces the owner-cash proxy to about $7.14 billion, or 35% of revenue. This lower figure may be too severe if part of the payable becomes a recurring timing float. Ignoring it would be equally generous in a year when cash taxes lagged accrued taxes by more than $1.4 billion.

The usable FY2026 owner-cash range is therefore $7.1 billion to $8.5 billion. It produces a 2.9% to 3.5% cash yield on Friday's equity value. That is a strong result, but it comes from a year in which gross margin rose to 71.5%. The burden at $1,641.11 is durability, not a shortage of current cash.

Four years show how violent the NAND cycle remains

Sandisk's history before separation from Western Digital was prepared on a carve-out basis. Corporate costs were allocated, treasury was centrally managed, and WDC debt was omitted unless Sandisk was the legal obligor. FY2025 also crosses two accounting bases: carve-out results through 21 February 2025 and standalone results thereafter. Those limitations make the direction more reliable than any single decimal (Sandisk 2024; Sandisk 2025a).

USD millions except ratios FY2023 FY2024 FY2025 FY2026
Revenue 6,086 6,663 7,355 20,248
Net income / (loss) (2,143) (672) (1,641) 11,433
Operating cash flow (713) (309) 84 11,671
Cash capex 219 166 204 177
Computed OCF less capex (932) (475) (120) 11,494
Gross margin 7.1% 16.1% 30.1% 71.5%
Computed period-end ROIC (18.1%) (5.4%) (15.9%) 99.2%
Computed debt / equity 8.0% 7.3% 20.1% 0.0%

Revenue, net income, operating cash flow, capex and gross margin are source-reported. OCF less capex, ROIC and gearing are author calculations. FY2026 is unaudited because Sandisk had filed its annual results release but not the fiscal 2026 Form 10-K by 15 August (Sandisk 2025a; Sandisk 2026c).

The ROIC calculation uses GAAP operating income after the reported tax rate, divided by period-end invested capital. Invested capital is debt plus equity less cash. Period-end capital is used throughout because the filed carve-out history does not provide the exact opening debt balance needed for a consistent FY2023 average. NOPAT was negative $2.18 billion on $12.07 billion of capital in FY2023. FY2026 NOPAT was $10.88 billion on $10.97 billion of capital. The resulting 99.2% is arithmetically correct and economically abnormal.

Two accounting facts increase the noise. FY2023 and FY2025 include goodwill impairments of $671 million and $1.83 billion, respectively. FY2025 also includes the separation financing and a $1.5 billion distribution to WDC. Even after those caveats, three years of negative cash after capex followed by $11.5 billion in one year is the defining evidence. The memory cycle can overwhelm ordinary incremental-return analysis.

The manufacturing moat is shared with Kioxia

Sandisk obtains substantially all its NAND wafers through Flash Ventures with Kioxia. The joint ventures account for roughly one-third of global NAND wafer output, giving both partners scale, process-learning density and access to large fabrication campuses. Sandisk generally bears about half of joint-venture capital requirements when venture cash is insufficient. Reported capex of $177 million therefore understates the manufacturing economics that sit behind its wafers (Sandisk 2025a; Quartr 2026).

This is a strong position with a dependency attached. Sandisk cannot freely replace Kioxia output, and the Flash Partners and Flash Alliance arrangements reach scheduled expiries in 2029 unless extended. Flash Forward runs to 2034. Yields, process transitions and allocation decisions inside the ventures determine whether contracted bits become delivered products.

Product capability adds a second layer. BiCS10 uses a two-terabit QLC design that Sandisk says provides 60% more density and 75% better read/write power efficiency than BiCS8. Enterprise SSD qualification, controller firmware and customer integration can lengthen relationships beyond a wafer contract. Yet peer results show that current margins are not unique. Micron reported an 84.6% GAAP gross margin in fiscal Q3, while Kioxia's June-quarter gross margin was about 78%. Scarcity and mix are lifting the field (Micron 2026; Kioxia 2026).

Counterpoint estimated Sandisk's NAND revenue share at roughly 12% to 13% for five consecutive quarters. YMTC had reached about 14% shipment share. Sandisk's recent growth came mainly from price and product mix, not a clear share gain. HBF and 3D Matrix may improve the technology position later, but no commercial HBF product or independent benchmark belongs in the FY2028-FY2030 cash model today (Counterpoint 2026).

The first downturn becomes the contract audit

The near-term NAND backdrop supports Sandisk. AI inference and large data-centre deployments are increasing enterprise SSD demand, while suppliers added little capacity in 2026. Counterpoint put enterprise SSDs at 48% of global NAND bit shipments in the June quarter, up from 26% a year earlier. Both Micron and Sandisk have moved customers toward multi-year commitments with binding volume terms and pricing floors or bands (Counterpoint 2026; Micron 2026).

The harder evidence arrives later. TrendForce projected a 4% to 5% NAND deficit during 2026, followed by a positive supply-demand balance in the second half of 2027 as process migration and Chinese capacity add bits. It put smartphones and notebooks near 40% of demand even as those end markets contracted, and estimated Chinese suppliers could approach 19% of global bit output (TrendForce 2026).

That timing intersects with the contract ramp. Half of FY2027 bits are under NBMs; roughly two-thirds of FY2028 bits are covered. If spot NAND falls below contracted economics in late 2027, customer behaviour will reveal the structure's quality. Timely deliveries at stated floors would support management's claim. Renegotiations, slower takes, disputes or unexplained RPO changes would show that legal commitments are not identical to cash certainty.

The strongest anti-thesis to cyclical caution is commercial, not technological. Customers are accepting multi-year terms across more than one supplier because AI infrastructure requires dependable storage supply. If those customers treat availability as more valuable than the lowest spot price, the old commodity analogy becomes less useful. Sandisk can then sustain higher average margins even while spot pricing moves.

Capital returns depend on separating cash from timing

Sandisk repaid $1.9 billion of term debt during FY2026 and finished July with no debt, $4.76 billion of cash and $1.78 billion of marketable equity securities. It also repurchased $4.52 billion of shares during the year. The board expanded authorization by $14 billion, leaving about $15.5 billion available (Sandisk 2026c; Sandisk 2026d).

A debt-free balance sheet matters in a cyclical industry. It allows the company to fund joint-venture commitments and product transitions without refinancing into a weak market. It also makes net cash a meaningful scenario input rather than a covenant buffer. The capital-allocation framework puts reinvestment first, maintenance of a debt-free balance sheet second and return of excess cash third (Quartr 2026).

The sequence is sensible; the definition of excess is still being tested. Customer prepayments add liquidity before revenue is earned. Tax payables delay cash taxes. Share repurchases reduce the denominator, but stock awards push the other way, and the fiscal Q1 2027 guide uses about 155 million diluted shares against roughly 149 million issued shares in the market-cap snapshot. A clean capital-allocation record requires buybacks to reduce diluted shares after compensation, not merely absorb issuance.

Management's FY2028-FY2030 framework also relies on mid-single-digit capital intensity. The joint-venture structure keeps conventional property purchases low on Sandisk's cash-flow statement, so capital intensity needs to be read together with Flash Ventures contributions. A business can look asset-light in statutory capex while still carrying half of a fabrication system's economic investment.

The valuation prices a fast path to durable cash

A commodity producer with unstable margins is poorly described by one year's earnings multiple. The primary valuation here capitalizes normalized owner cash, then cross-checks that result with a discounted cash-flow replication of the market price. The scenario formula is normalized revenue multiplied by owner-cash margin and an appropriate cash-flow multiple, plus net cash, divided by 148 million reference shares. The market-cap snapshot uses 149 million issued shares; the lower valuation denominator is a conservative approximation of the latest exact common-share count.

Case Normalized revenue Owner-cash margin Cash-flow multiple Observed value per share
Severe downside $16bn-$20bn 10%-20% 7x-10x $96-$304
Bear $22bn-$27bn 25%-35% 10x-13x $399-$871
Base $30bn-$38bn 35%-42% 14x-17x $1,034-$1,887
Bull $38bn-$48bn 45%-52% 17x-21x $2,011-$3,609

The severe case assumes contract stress and a return toward commodity margins. The bear case grants Sandisk structurally better economics than the pre-separation years but assumes 70% to 85% gross margins fade. The base case requires NBM floors to preserve a 35% to 42% owner-cash margin through a moderate reset. The bull case requires AI storage intensity, contract enforcement and capital discipline to make a 45% to 52% owner-cash margin durable.

Friday's $1,641.11 close sits inside the base range. It also admits a precise reverse valuation. Starting with FY2026 revenue of $20.25 billion, applying 17% annual growth through FY2030 and a 50% free-cash-flow margin from FY2027, then discounting at 10% with 3% terminal growth, produces about $1,640 per share after adding cash. The market price is almost an exact replication of the midpoint growth ambition and the full 50% cash-margin model.

That result does not prove the price is irrational. It identifies what the price contains. A 40% cash margin, 11% discount rate and 2.5% terminal growth reduce the same framework to about $1,097 per share. A 50% margin with 19% growth, a 9% discount rate and 3.5% terminal growth produces about $2,203. Margin persistence and discount rate matter more than the next quarter's revenue beat.

A second sensitivity fixes normalized revenue at $35 billion and net cash at $4.76 billion. At a 15-times cash multiple, a 35% margin gives roughly $1,274 per share and a 45% margin gives $1,628. At 18 times, the same margins give about $1,522 and $1,948. Around Friday's price, the market needs roughly a 45% margin at 15 times, 38% at 18 times or 34% at 20 times. Every combination is above Sandisk's pre-FY2026 cash record.

Friday was proportionate; the two-day price contains proof not yet supplied

Friday's 7.39% rise had two identifiable inputs: the Investor Day model was being absorbed, and JPMorgan attached a longer-duration interpretation to it. Volume remained 69% above Wednesday's level while the semiconductor index fell. A second day of price discovery was reasonable.

The full 22.08% rise is harder to separate from execution. It added about $44.2 billion of equity value. Most contract coverage and guarantee figures had already appeared with the 5 August results. Investor Day provided the 80% gross-margin and 50% adjusted cash-margin framework, plus a clearer account of contract duration, technology and capital returns. The presentation increased coherence more than it increased signed obligations.

The market reaction is therefore proportionate with a front-loaded valuation response. Sandisk has created a credible mechanism for reducing volatility on covered bits. Peer contracts support the mechanism, current peer margins validate the operating peak, and the balance sheet can absorb investment. Counter-evidence is equally concrete: guarantees cover 18% of RPO, one-third of FY2028 bits remains outside NBMs, customer terms are private, and industry supply may loosen in the second half of 2027.

Five reported numbers will decide the contract experiment

Fiscal Q1 2027 is the first checkpoint. Sandisk guided to $10.3 billion to $10.8 billion of revenue, an 83% to 85% non-GAAP gross margin and roughly 155 million diluted shares. Results outside those ranges would show whether price, mix or dilution is already departing from the Investor Day path (Sandisk 2026c).

The fiscal 2027 disclosures need to preserve four other links. NBM coverage must remain near 50% of bits before rising toward two-thirds in FY2028. RPO must run down in line with recognized revenue rather than through quiet contract changes. The guarantee-to-RPO ratio needs to hold or rise as obligations shrink. Adjusted owner cash needs to stay above 35% after removing customer prepayments, stock compensation and tax timing.

Then the industry supplies the decisive test. TrendForce's expected shift to surplus in the second half of 2027 puts cheaper spot NAND beside contracted floors. Customer performance during that period will reveal whether $91.1 billion is durable revenue protection or a negotiating position formed at the top of a shortage.

Confidence and source notes

Confidence is high in the filed financial history, FY2026 results, market move and published Investor Day deck. Confidence is lower in legal protection because Sandisk has not filed the customer agreements and does not identify counterparties. FY2026 annual figures are unaudited pending the 10-K, and the 99.2% computed ROIC is distorted by peak margins, separation accounting and a period-end capital denominator.

Sandisk has divided its cycle exposure rather than eliminated it. Half of FY2027 bits and two-thirds of FY2028 bits have committed economics; the rest still meets the market. At $1,641.11, the market is already close to capitalizing 17% growth and a 50% cash margin into perpetuity. The next downturn will determine whether eight customers can carry that assumption when spot NAND becomes cheaper.

References