SNDK · Forward model · Bull case

The Bull case, 20 quarters out

Each segment is projected from its own operating driver, rolled up into consolidated cash flow, and discounted back to a fair value per share. The assumptions below are editable — change them and every number on this page moves with them.

WHAT IS REPORTED. Sandisk runs a single reportable operating segment, so the only thing split three ways is revenue: the Datacenter/Edge/Consumer End Market Summary published in every earnings release. All ten quarters here are as reported and sum to consolidated revenue to the dollar in each one - $1,705M, $1,760M, $1,883M, $1,876M, $1,695M, $1,901M, $2,308M, $3,025M, $5,950M and $8,965M. Fiscal 2024 Q1 and Q2 are absent from every vertical because Sandisk did not file separately before the February 2025 separation and that split has never been published; nothing is back-filled from Western Digital, whose Cloud/Client/Consumer lines mixed HDD with flash. The FY2025 Cloud/Client/Consumer labels and the FY2026 Datacenter/Edge/Consumer labels are the same three lines renamed - the FY2025 totals are identical under both. UNITS ARE AN INDEX. Sandisk discloses exabyte GROWTH by end market and has never disclosed an exabyte LEVEL, so every unit driver here starts at 100 in the basis quarter and price per index point is that quarter's revenue divided by 100. Nothing in the traces is a dollars-per-gigabyte figure the company published. MARGINS ARE COMPANY-LEVEL. No profitability is reported below the company line - not in the release, not in the deck, not in the 10-K. The 84.6% basis margin and the 80.0% terminal margin on each vertical are Sandisk's own non-GAAP GROSS margins, applied uniformly and tagged assumed; operating expense is charged once at the corporate line at 5.0% of revenue, which is where the FQ1 2027 guide of $520-540M on a $10.55bn midpoint lands and what the Investor Day calls 'around five percent'. Gross less opex gives a 78.9% operating margin in the first projected quarter against the 79.2% just reported, and 75.2% across FY2031 against the ~75% guided for FY2028-FY2030. No Datacenter margin premium is invented; the one margin-relevant fact that is derivable points the other way, since Datacenter earns about 0.87x the company average revenue per bit. CAPEX IS ON A GROSS BASIS. 6.3% falling to 6.0% of revenue is Sandisk's GROSS investment including its share of Flash Ventures capex, matching the $562M shown for the basis quarter and the 'approximately 6% of revenue' guided for FY2027. It is NOT the $153M of cash capex (1.7%) that Sandisk's own reported free cash flow deducts. The 10-K's $6,559M of Flash Ventures commitments is deliberately not added as a separate corporate programme: its FY2027 tranche of $2,627M is 5.8% of a $48.96bn consensus year, so it IS the 6% guide, and carrying both would fund the joint venture twice. FREE CASH FLOW IS BEFORE WORKING CAPITAL. The engine takes EBITDA less gross capex less cash tax at the guided 15% non-GAAP rate. It prints a 61.8% margin in the first projected quarter, above the 56.2% adjusted free cash flow margin of the basis quarter and above the ~50% Sandisk guides for FY2028-FY2030, because management's 50% is stated 'after accounting for taxes, capital expenses, and working capital to support growth' and this model does not carry a working-capital line. Read the FCF margins as pre-working-capital. NO SEASONALITY ANYWHERE. Ten quarters of per-vertical history exist, spanning only two complete fiscal-year windows, one of which was a 53-week year with a 14-week first quarter. The window-to-window spread exceeds the candidate signal in all three lines - 1.423 for Datacenter, 0.718 for Edge, 0.274 for Consumer - and the three verticals compete for one capped bit pool, so Consumer's fiscal-Q4 trough is the mirror of Datacenter's peak rather than a season. Fitting factors would have projected a 2.3x Datacenter quarter every fiscal Q4 forever. Consumer is worth re-testing after FY2027 closes. WHERE THIS SPEC DEPARTS FROM THE RESEARCH BRIEF, AND WHY. (1) The brief's assumed volume rates blended to roughly 10% a year, below the mid-teens sellable bit growth Sandisk guides for FY2027; Datacenter is set to 6.5% a quarter instead of 4.0% so the bit-weighted blend comes to between 16% and 18% a year across the horizon whichever way the undisclosed Edge/Consumer bit split falls, consistent with the guide and with the disclosed allocation shift from 12% to 38% of bits. (2) The brief set the first quarter's price drift to zero, which would have printed $9.26bn against a guided $10.30-10.80bn; +14.0% in all three lines lands the quarter on the guided midpoint and implies about $45.7 of non-GAAP EPS on 155M guided diluted shares, inside the guided $44.00-46.00. The drift then decays at 0.35 a quarter to a negative terminal, so the guided quarter is a step and not a trend. (3) The exit is 6.0x terminal EBITDA rather than 9.0x, because the market itself prices Sandisk at 5.6x this model's FY2027 EBITDA today and memory comparables do not hold premium multiples on peak-cycle earnings. SHARES AND NET CASH. 146,419,001 shares outstanding at 7 August 2026 per the 10-K cover, the same count behind the $219.5bn market capitalisation. The guided FQ1 2027 diluted count is about 155 million, so a fully diluted fair value per share would be roughly 5.5% lower. Net cash of $6,539M is $4,762M of cash plus $1,777M of marketable equity securities against zero debt; a cash-only definition gives $4,762M and the equity securities are mark-to-market, which is what inverted GAAP EPS above non-GAAP this quarter. ONE DATA CAVEAT. data/companies/sndk/series.json carries $2,310M and $3,030M for 2026 Q1 and Q2 where the End Market Summary and income statement say $2,308M and $3,025M, and it mixes GAAP and non-GAAP EPS across 2026 Q3 and Q4. This model uses the released figures throughout.

Where each case comes from

From cash flow to fair value

The projected path

Every row is projected. A year-over-year change is shown only where the quarter it compares against exists — an em dash means there is no comparable quarter, not a flat year.

Model revisions

Assumptions are marked to reality as each quarter prints. Every change is appended here, with the fair value the model produced at the time, so the model's own history stays visible.