Sandisk's Three-Year Model Guides Gross Margin Below Where It Already Is. The Stock Closed Up 13.67% Anyway

by Clara Borresen · August 14, 2026

Senior Earnings Analyst, Earnings Dispatch

Sandisk (Nasdaq: SNDK) spent Thursday in New York walking investors through a multi-year financial framework, and the most striking thing about it is that the headline margin target sits below the margin the company is already earning. The model, covering fiscal 2028 through fiscal 2030, calls for revenue growth in the mid-to-high teens broadly in line with bit growth, non-GAAP gross margins of approximately 80%, non-GAAP operating margins of approximately 75%, operating expenses of roughly 5% of revenue and an adjusted free cash flow margin of approximately 50%. Shares closed Thursday at $1,528.11, up 13.67%, according to TradingKey's session wrap; Invezz reported the same close as a 13.7% gain.

The arithmetic inside the model is internally consistent, which is worth noting because multi-year frameworks often are not: 80% gross margin less 5% of revenue in operating expense produces the 75% operating margin the company is guiding to. What it is not consistent with is the recent past. In fiscal fourth-quarter results released after the close on Wednesday, Aug. 5, Sandisk reported revenue of $8.965 billion, up 51% sequentially, with GAAP and non-GAAP gross margin both at 84.6%, GAAP diluted EPS of $43.97 and non-GAAP diluted EPS of $39.25. Non-GAAP operating income was $7.104 billion. Full-year fiscal 2026 revenue was $20.248 billion, up 175% year over year, with datacenter revenue up 437% to $5.153 billion. Guidance for the current fiscal first quarter is revenue of $10.30 billion to $10.80 billion, non-GAAP gross margin of 83.0% to 85.0% and non-GAAP diluted EPS of $44.00 to $46.00.

Set the two side by side — both on a non-GAAP basis, which is the only way this comparison is legitimate — and the three-year model asks investors to accept roughly four points of gross margin below the midpoint of the guidance the company issued nine days ago. In a commodity memory business that is not obviously a downgrade; it is a statement about where a cyclical peak sits relative to a through-cycle floor. But it is a guide down from present conditions, and it was published in a press release rather than buried in a slide footnote.

What the market appears to have paid for instead is the contract structure. Sandisk said eight customers have signed agreements under what it calls the New Business Model, which the release describes as built on committed volumes, enforceable contractual frameworks with minimum financial guarantees, and structured pricing mechanisms, and which the company said covers approximately 50% of bits in fiscal 2027 and approximately two-thirds of bits in fiscal 2028. The dollar size of that book does not appear in the release text available to this newsroom. Crypto Briefing put it at $93.9 billion of minimum contracted revenue at floor pricing across 10 New Business Model agreements with a weighted-average duration of more than four years, backed by $16.5 billion of aggregate financial guarantees. J.P. Morgan's Harlan Sur, in a note summarized by Investing.com, described roughly $94 billion of total contract value at floor pricing, a weighted-average duration above four years and the same $16.5 billion of financial guarantees, and put gross margin near 80% even at the floor pricing tier. The point worth stating plainly, because it is easy to get backwards: these are not volume-only minimums. Every account of the structure pairs the committed volumes with a price floor.

David Goeckeler, Sandisk's chairman and chief executive, said in the release: "Our strong performance today is the direct result of disciplined execution against the strategy we outlined 18 months ago." He went on to attribute that position to decades of NAND flash innovation, systems-level expertise, a diversified portfolio and capital-efficient operations. Chief financial officer Luis Visoso set out the capital allocation side, saying the company expects to return 100 percent of excess cash to shareholders after investing in the business, and grounding his confidence in the model's sustainability in the multi-year New Business Model agreements.

The reaction is more interesting read against what happened just over a week earlier. When Sandisk reported those fiscal fourth-quarter numbers on Aug. 5, the stock fell about 5% the following session. Non-GAAP EPS of $39.25 beat a Zacks consensus of $34.24 and revenue beat by about 8%, but the fiscal first-quarter revenue guide of $10.30 billion to $10.80 billion sat below a consensus around $11.16 billion, and the guide is what traded. Same company, same conservative posture, opposite outcome. The difference is the horizon: a single quarter's revenue guide is a bar the company can miss in ninety days, while a three-year margin floor attached to signed volume commitments is a statement about cyclicality itself.

Analysts moved accordingly. J.P. Morgan's Harlan Sur upgraded the stock to Overweight from Not Rated, writing that the New Business Model framework has structurally reset the company's margin profile higher and materially reduced cyclicality. Argus Research's Jim Kelleher upgraded to Buy. Evercore ISI's Amit Daryanani reiterated Outperform. Jefferies' Blayne Curtis kept a Buy rating while cutting his price target. The disagreement is not really about the reported numbers, which are the same for everyone; it is about whether a NAND supplier can be underwritten as a contracted business rather than a spot one.

The technology case the company put behind the bit-growth assumption rests on a two-dimensional scaling strategy built around its CMOS directly Bonded to Array architecture, applied across its BiCS9 and BiCS10 QLC nodes, plus High Bandwidth Flash aimed at AI inference workloads. Separately, and a day before the investor day, Sandisk and Kioxia unveiled a ninth-generation, 2-terabit QLC 3D flash device on that CBA architecture with a roughly 33% faster interface reaching 4.8 gigabits per second. At the investor day itself the company sized the enterprise datacenter flash opportunity at 1.2 zettabytes by 2030. Whether that TAM materializes is the load-bearing assumption in the entire framework; the margin structure only works at the volumes the AI buildout is assumed to demand.

Two risks are worth stating plainly. The first is that approximately 80% gross margin is an extraordinary level for a NAND business historically characterized by violent price cycles, and the company is modeling it as a multi-year plateau rather than a peak. The second is concentration: covering half of fiscal 2027 bits and two-thirds of fiscal 2028 bits with eight customers converts price risk into counterparty and renewal risk, which is why the financial guarantees described in the third-party accounts above matter to the structure. Sandisk shares were up 541% year to date as of Thursday's close, according to Invezz, which means the model is being received by a shareholder base that has already been paid for the cycle.

The move was not isolated. Memory led Thursday's tape, with Micron up 4.23%, Western Digital up 7.31% and Seagate up 4.91%, per TradingKey's close report, which put the Philadelphia Semiconductor Index up 0.46% on the day — a gap that shows how narrow the bid was. The S&P 500 finished at 7,798.99, up 0.65%, and the Nasdaq Composite at 26,803.03, up 0.81%.