Sandisk (SNDK) targets 80% gross margins and 100% cash return through FY2030

Sandisk Corporation (Nasdaq: SNDK) laid out its multi-year growth strategy and a new long-term financial model at its 2026 Investor Day, held in New York City on Wednesday under the “Sandisk In Focus” banner. Chairman and Chief Executive Officer David Goeckeler and Chief Financial Officer Luis Visoso used the event to translate the record fiscal fourth-quarter performance reported just over a week earlier into a durable framework covering fiscal 2028 through fiscal 2030, anchored by mid-to-high teens revenue growth, non-GAAP gross margins of approximately 80 percent, non-GAAP operating margins of approximately 75 percent, an adjusted free-cash-flow margin near 50 percent, and a commitment to return 100 percent of excess cash to shareholders after reinvestment in the business. The company also unveiled a new BiCS9 QLC node built on a two-dimensional scaling strategy, positioned High Bandwidth Flash as a maturing category with a broader ecosystem forming around it, and reiterated that its New Business Model agreements now cover approximately half of fiscal 2027 bits and roughly two-thirds of fiscal 2028 bits. The unresolved question sitting under every slide is whether NAND, historically the most cyclical corner of the semiconductor complex, has genuinely been re-engineered into a business capable of sustaining margins that would not have been credible even eighteen months ago.

What does the FY2028 to FY2030 financial model actually commit Sandisk to delivering?

The multi-year framework introduced at Sandisk In Focus 2026 is the first attempt by management to formalise the operating shape of the company beyond a single guidance quarter. Between fiscal 2028 and fiscal 2030, Sandisk expects revenue to grow at a mid-to-high teens percentage rate, described as consistent with bit growth. Non-GAAP gross margin is expected to sustain at approximately 80 percent. Non-GAAP operating margin is expected to sit at approximately 75 percent, implying operating expenses of roughly 5 percent of revenue with no meaningful contribution or drag from other income and expense. Adjusted free cash flow margin, after taxes, capital expenditure and working capital, is targeted at approximately 50 percent.

Management framed the model as optimised for growth, sustainability and returns, with 100 percent of excess cash to be returned to shareholders after the business is fully funded. Visoso attributed the confidence in the model’s durability to the New Business Model contracts, describing them as intimate multi-year relationships grounded in innovation and collaboration. The model itself is presented as forward-looking non-GAAP guidance and is not reconciled to comparable GAAP measures on the grounds that the timing and amount of certain charges cannot be reliably estimated.

Read against the fiscal 2026 exit run rate reported on August 5, the target model looks less like an ambition and more like a commitment to preserve current unit economics. Sandisk closed fiscal 2026 with revenue of $20.25 billion, up 175 percent year on year, and reported a fourth-quarter non-GAAP gross margin of 84.6 percent, non-GAAP operating margin near 79 percent and non-GAAP diluted earnings per share of $39.25. The FY2028 to FY2030 model therefore describes a slightly softer, but structurally similar, business rather than a step-change beyond where the company already sits.

Why does an 80 percent non-GAAP gross margin target represent a break from NAND’s historical cyclicality?

NAND flash has spent most of its history as a commodity business defined by aggressive supply investment, sharp average selling price swings and gross margins that troughed in single digits, or turned negative, in every serious downcycle. Non-GAAP gross margins above 40 percent were previously considered strong for the category. A structural anchor at approximately 80 percent, held across a three-year window that spans a full traditional cycle, has no direct precedent in the memory industry.

The company’s position is that this is not simply a repricing at the top of a cycle. Management argues that AI inference has changed the fundamental role of NAND from a cost-optimised storage substrate into a strategic component of the AI infrastructure stack, that KV cache and other inference-era workloads are pulling data closer to compute and thereby expanding the addressable data-centre flash market to approximately 1.2 zettabytes by 2030, and that the shift from spot allocation to multi-year contracted supply through the New Business Model framework has fundamentally rewired customer behaviour.

The counter-reading is that gross margins above 80 percent reflect the tightest supply and highest ASP environment NAND has ever seen, and that even structurally improved contracting will not fully insulate the industry from bit-supply catch-up as competitors invest into the current price signal. Management has already indicated that NBM-specific gross margins run at around 80 percent, meaning the 84.6 percent recently reported includes cyclical uplift above the contracted baseline. The FY2028 to FY2030 guide is essentially that NBM baseline. Whether it holds through a supply response elsewhere in the industry, particularly from Samsung Electronics, SK hynix and Kioxia Holdings Corporation, is the central unresolved question for the model.

How do the New Business Model agreements underwrite the visibility that anchors the long-term guide?

The New Business Model contracts are built on committed volumes, enforceable contractual frameworks with minimum financial guarantees, and structured pricing mechanisms that blend fixed and variable elements. The variable portion is subject to floors and ceilings that management has described as producing attractive margins even at floor pricing. Supply and demand commitments are defined by year and by quarter, and customer financial guarantees release toward the end of each agreement, meaning coverage relative to remaining obligations rises as the contract matures.

As of today, Sandisk has signed NBMs with eight customers. Those agreements cover approximately 50 percent of fiscal 2027 bits and approximately two-thirds of fiscal 2028 bits. When the company reported fourth-quarter results on August 5, it disclosed that customer prepayments and deposits of $1.94 billion had already flowed into cash from operations during the quarter, and that $16.5 billion in aggregate financial guarantees underpinned the NBM book. The August 5 disclosure also noted five additional agreements had been signed since the April earnings call, three with new customers and two as expansions of previously signed NBMs, taking the total to ten agreements across the eight-customer set.

For an executive audience, the mechanical implication is that Sandisk has converted between one-half and two-thirds of its planned bit output for the next two fiscal years into a book of business closer in shape to a long-cycle contract manufacturer than a commodity memory supplier. The trade-off is customer concentration. Eight customers accounting for two-thirds of fiscal 2028 bits creates a small addressable set for any single defection, delay or contractual renegotiation. Sandisk has not disclosed the identity of individual NBM counterparties, which limits external assessment of concentration risk beyond what the aggregate numbers suggest.

What does the BiCS9 QLC hybrid architecture tell investors about capital-efficient NAND scaling?

The most concrete technology disclosure of the day was BiCS9 QLC, a new node that Sandisk framed as the first production example of a two-dimensional scaling strategy resting on its CMOS directly Bonded to Array, or CBA, foundation. BiCS9 combines a proven BiCS8 NAND array wafer with a BiCS10-generation CMOS wafer, producing a node calibrated for AI-driven workloads without requiring a full new array generation. The claim is capital-efficient production of custom derivatives that can be tuned to specific customer workloads at a pace faster than traditional single-axis scaling would allow.

Sandisk also disclosed that its BiCS10 QLC node achieves a 60 percent increase in bit density compared to BiCS8, which it presented as an industry benchmark for the combination of density, performance and power efficiency. The strategic significance for readers accustomed to memory-node economics is straightforward. Two-dimensional scaling separates array progress from CMOS progress, allowing the more expensive and slower-moving of the two layers to be reused where appropriate. In a business where every additional bit of density has historically required a fresh capital cycle, the ability to derive multiple product variants from a shared array is an argument for lower incremental capital intensity and shorter development timelines. Whether that shows up as capex discipline in the actual FY2028 to FY2030 window, alongside the 100 percent excess cash return commitment, is the metric worth tracking.

Where does High Bandwidth Flash fit in the AI memory stack, and does the ecosystem look real?

High Bandwidth Flash was framed at the Investor Day as gaining momentum, with an ecosystem forming around adoption. The claim rests on concrete progress made in the ten days before the event. On August 3, Sandisk and SK hynix released the first HBF technical specification through the Open Compute Project, following an initial standardisation partnership announced in August 2025 and the formation of the HBF workstream under OCP in February 2026. The specification supports capacities of up to 512 gigabytes using eight-high and sixteen-high NAND die stacks, with three bandwidth grades ranging from approximately 0.4 terabytes per second to 3.0 terabytes per second, and adopts the Universal Chiplet Interconnect Express standard for integration with CPUs, GPUs and other accelerators. Google and Tenstorrent joined the consortium during the standardisation process and are named as validation contributors.

The strategic thesis is that HBF sits between High Bandwidth Memory and conventional SSD storage, offering HBM-like bandwidth at NAND capacity economics, and that this new memory tier will be pulled into inference-era system designs where KV cache and long-context workloads make HBM capacity a binding constraint. The counter-signal is that no commercial HBF products or independent benchmarks are yet in the market, that competing memory technologies including HBM3E, HBM4 and CXL-attached DRAM are targeting overlapping workloads, and that the pace of design-win conversion will determine whether HBF becomes a category or a specification. For Sandisk, the ecosystem developments over the past fortnight are directionally supportive, but the FY2028 to FY2030 financial model does not appear to rest on HBF as a numerical contributor. Any HBF revenue that materialises inside the window would be additive to the guide rather than embedded in it.

What does returning 100 percent of excess cash mean for Sandisk’s capex intensity through 2030?

The commitment to return 100 percent of excess cash after business investment is the most consequential capital-allocation shift the company has communicated since spinning out of Western Digital Corporation in February 2025. In the fourth quarter of fiscal 2026, gross capital expenditure was $562 million, or 6.3 percent of revenue, and adjusted free cash flow was $5.04 billion at a 56 percent margin, excluding $1.94 billion in NBM prepayments and deposits. The Board of Directors expanded the share repurchase authorisation on August 5 by an additional $14 billion, bringing total remaining authorisation to $15.5 billion after $4.5 billion in fourth-quarter repurchases.

The FY2028 to FY2030 adjusted free-cash-flow margin target of approximately 50 percent, applied against revenue rising at mid-to-high teens from a fiscal 2026 base of $20.25 billion, implies a very large annual cash-return capacity if the model is delivered. The tension inside that commitment is whether the two-dimensional scaling strategy, the BiCS10 node, ongoing BiCS8 and BiCS9 production, and future capacity for HBF can genuinely be funded at capital-intensity levels consistent with the guide. NAND has historically absorbed a much larger share of revenue in capex during upgrade cycles. If two-dimensional scaling delivers as advertised, the capex line stays disciplined and the cash-return commitment holds. If competitors force a supply response that requires Sandisk to accelerate its own investment, the 100 percent excess cash return moves from durable policy to conditional policy.

What are the main forward risks that could weaken the multi-year thesis Sandisk laid out at Investor Day 2026?

Several forward variables sit outside management’s direct control. A meaningful NAND supply response from Samsung Electronics, SK hynix, Kioxia Holdings Corporation, Micron Technology and Yangtze Memory Technologies Co. over the FY2028 to FY2030 window would compress the pricing environment that the 80 percent gross margin assumption implicitly requires. Slower-than-expected AI inference deployment, whether from data-centre power constraints, cost-of-inference economics or shifts in model architecture that reduce the pull on flash capacity, would soften the demand curve that anchors the 1.2 zettabyte 2030 TAM. Concentration in the NBM book means that any single major customer renegotiation would be visible in the reported numbers.

Execution risks around the two-dimensional scaling strategy, HBF ecosystem adoption and BiCS10 ramp are internal-facing but material. Sandisk continues to operate its manufacturing base in joint arrangements historically involving Kioxia, and any change in that relationship would affect the shape of the capex commitment underlying the model. Regulatory or trade-policy actions affecting flash memory supply chains between the United States, Japan, South Korea and China sit above the company’s operational control, and forward-looking statements in the Investor Day disclosure explicitly cite evolving trade policies and tariff regimes as material risks.

How is the market pricing Sandisk shares heading into the FY27 execution window?

SNDK opened the Investor Day trading session on the front foot. The stock traded at $1,504.41 by late morning on August 13, up 11.91 percent from Tuesday’s close of $1,344.29, with the 52-week range spanning $41.00 to $2,354.39 and one-year performance sitting at more than 2,700 percent. The Investor Day reaction reversed a portion of the selloff that followed the August 5 fiscal fourth-quarter release, when shares fell in the low double digits after hours despite the revenue and earnings beat, on concerns that first-quarter fiscal 2027 guidance of $10.30 billion to $10.80 billion in revenue and non-GAAP earnings per share of $44.00 to $46.00 had not cleared the highest published buy-side expectations.

Sell-side action into the event was mixed. Argus Research upgraded SNDK from Hold to Buy on August 10 with a $1,600 price target. Wells Fargo trimmed its price target on the same name to $1,400 from $1,620 while retaining an Equal Weight rating, citing expectations for shares to digest slowing near-term momentum. Business News Today did not identify a widely published current broker consensus that resolves the range between valuation-cautious equal-weight views and outright buy calls carrying targets above $2,000. The Investor Day intraday move suggests the market read the FY2028 to FY2030 framework as reinforcing the durability case that fiscal fourth-quarter numbers had already made, rather than pricing a step-change beyond what was already visible.

Key takeaways from Sandisk’s 2026 Investor Day and new FY2028-2030 financial model

• Sandisk Corporation (Nasdaq: SNDK) unveiled a multi-year financial model at its 2026 Investor Day covering fiscal 2028 through fiscal 2030, targeting mid-to-high teens revenue growth, approximately 80 percent non-GAAP gross margins, approximately 75 percent non-GAAP operating margins and approximately 50 percent adjusted free cash flow margins.
• The company committed to returning 100 percent of excess cash to shareholders after reinvestment in the business, following the August 5 expansion of the share buyback authorisation by an additional $14 billion to a total remaining $15.5 billion.
• The model essentially annualises Sandisk’s fourth-quarter fiscal 2026 profitability profile, in which non-GAAP gross margin reached 84.6 percent and non-GAAP earnings per share reached $39.25 on revenue of $8.97 billion.
• New Business Model contracts with eight customers now cover approximately half of fiscal 2027 bits and roughly two-thirds of fiscal 2028 bits, underpinned by $16.5 billion in aggregate financial guarantees disclosed on August 5.
• The BiCS9 QLC node combining a BiCS8 array wafer with a BiCS10 CMOS wafer was presented as the first example of a two-dimensional scaling strategy targeting capital-efficient NAND derivatives.
• The BiCS10 QLC node was disclosed as achieving a 60 percent bit-density gain over BiCS8, framed as the industry benchmark for combined density, performance and power efficiency.
• High Bandwidth Flash was framed as gaining momentum, with the first OCP technical specification published on August 3 in partnership with SK hynix and consortium contributions from Google and Tenstorrent.
• Management expects the total addressable market for enterprise data-centre flash to grow to approximately 1.2 zettabytes by 2030, driven by AI inference workloads and KV cache reshaping the memory hierarchy.
• Central execution tests through FY2030 include supply-response behaviour from Samsung, SK hynix, Kioxia, Micron and YMTC, AI inference deployment pace, NBM customer concentration, and whether two-dimensional scaling actually delivers the capital-intensity discipline the 100 percent excess-cash-return policy requires.
• SNDK traded at $1,504.41 late Wednesday morning, up 11.91 percent from Tuesday’s close of $1,344.29, with a 52-week range of $41.00 to $2,354.39 and roughly $199 billion in market capitalisation as of August 12.