The headline from SanDisk’s fiscal fourth quarter was a $14 billion repurchase authorization, lifting the total remaining limit to $15.5 billion. The number that carries more information is $4.5 billion, which is what the company actually spent buying its own stock during the quarter. That figure represents 89% of the $5.035 billion in adjusted free cash flow generated in the same three months. SanDisk did not announce an intention. It returned nearly all of a record quarter’s cash to shareholders inside the quarter itself, and then asked the board for permission to do it again at triple the scale.
The execution window is what makes this interesting rather than routine. The first authorization was announced on April 30, when the stock closed at $1,096.51. The fiscal quarter ended in late June. Between those two dates the stock printed a record close of $2,335 on June 25. Unless the buying was heavily front-loaded into the first week of May, the average price SanDisk paid for that $4.5 billion sits meaningfully above the roughly $1,260 where the shares trade after the print. The company’s own capital, deployed at scale, is currently underwater. The 10-K contains a monthly repurchase table with average price paid, and that table is the single most useful disclosure still outstanding on this name.
Management’s response to the round trip was to add $14 billion. Goeckeler described the intent as consistent execution rather than opportunistic timing, and Flores framed buybacks as the third capital priority behind business investment and a cash balance the company now considers complete, chosen over a dividend on tax-efficiency grounds. Consistency is the right posture for a company that cannot predict NAND pricing, but it also means the authorization is not a valuation statement. Nobody at SanDisk is claiming the stock is cheap at $1,260. They are claiming the cash has nowhere better to go.
The scale needs a check against the market cap rather than against the press release. At roughly $200 billion, $15.5 billion is about 7.8% of the company. At $1,260 per share it retires around 12.3 million shares against roughly 157 million diluted. That is a real number and a meaningful one over eighteen months. It is not the number that will move the stock. Guidance of $44.00 to $46.00 in non-GAAP diluted EPS for the first quarter of fiscal 2027, on that share count, annualizes to something near $28 billion of earnings power. The entire remaining authorization is roughly six months of earnings at the guided run rate. Buybacks are a rounding adjustment against a pricing cycle that moved revenue 372% year over year. Anyone building a thesis on share count reduction has picked the smaller variable.
Where the buyback does matter to the share price is mechanically. A company committed to steady execution against $8.74 billion of fiscal 2026 adjusted free cash flow, with fiscal 2027 tracking far higher, becomes a standing bid underneath the stock every trading day. In a name that has moved 36-fold since the February 2025 IPO, that carries a lot of leveraged and momentum ownership, and that just sold off on a beat, the presence of a large price-insensitive buyer changes the shape of drawdowns. It compresses the downside tail and thins the float that the next upside move has to clear. Volatility does not go down. It goes asymmetric.
The competitive point sitting behind all of this is asset structure, not capital return. SanDisk carries $674 million in net property, plant and equipment against $20.2 billion of annual revenue because the fabs live inside the Kioxia joint venture. Western Digital is working through a $4 billion program approved in February. Micron has been constrained by repurchase limits attached to its CHIPS grant, which expire in December, and UBS has argued it could retire more than 40% of its shares afterward. Of the three, SanDisk converts a margin dollar into a retired share with the least capital drag, because it is not funding its own wafer capacity out of the same cash flow. BiCS8 has ramped to a majority of bit production and BiCS10 is announced, so the technology position is being maintained without the balance sheet that normally accompanies it. That is the moat worth underwriting. The buyback is a consequence of it.
The New Business Model agreements complicate the read in both directions. Five more were signed in the quarter, three with new customers. Floor pricing on multi-year contracts converts a portion of future cash flow from spot NAND into something closer to a contracted annuity, which makes a large buyback defensible in a way it never was for a memory company selling entirely into a commodity market. It also caps participation in fiscal 2027 pricing, which is precisely what the market punished. Buying back stock against contracted cash flow is prudent. Buying back stock against contracted cash flow while the multiple is set by uncontracted pricing expectations is a mismatch, and the June to August drawdown is that mismatch resolving.
On the trajectory, the shares sit near $1,258 after closing at $1,350.50 into the print and slipping again after hours, up more than 500% year to date, capitalized around $200 billion, and trading at roughly 44 times trailing earnings before the report. Street targets span an unusually wide band: Susquehanna at $3,250, Bernstein at $3,000, Bank of America maintaining $2,500, Goldman at $2,200, Citi cutting to $2,100 from $2,500, Morgan Stanley at $1,750, with the low of the published range near $1,620 and an average close to $2,400. Every one of those targets sits above the current price, which is itself a warning about how much positioning is already committed.
Base case is $1,600 to $2,000 over twelve months, requiring only that the NAND market reaches the $300 billion management expects for calendar 2026, that bits stay on allocation, and that the buyback runs at a steady quarterly pace. Bull case is $2,500 to $3,250 and requires the calendar 2027 approach toward $500 billion to hold, datacenter to continue compounding off the $2.98 billion quarterly base, and the full authorization to be executed inside four quarters at prices below $1,500, which would retire close to 10% of the company at a discount to the first tranche. Bear case is $800 to $1,100 and does not require earnings to disappoint at all. It requires only cohort derating. Memory has always been priced off the second derivative of pricing rather than the level, and the first quarter gross margin guide of 83% to 85% against 84.6% delivered, plus consumer revenue down 32% sequentially, is enough of a decelerating tell for a market that has already begun selling the group. In that scenario the buyback stops being a support and becomes the exhibit, a management team that spent $4.5 billion of shareholder money near the top of its own chart and was handed $14 billion more.
Which is why the 10-K table is the thing to read rather than the press release. If the average price paid in fiscal Q4 comes in near $1,300, management timed a violent tape well and the incremental $14 billion deserves the benefit of the doubt on execution. If it comes in near $1,900, the board has just tripled the mandate of a team that bought its own high, and every future tranche should be discounted for the same judgment.