Sandisk’s NAND upcycle is being repriced too slowly: the surprise is margin, not revenue

The actionable read from this print is that the market likely priced in a NAND recovery, but not this degree of flow-through. What was priced in was a revenue beat off estimates of $2,687.6 million, because the prior quarter already showed $2,308.0 million of revenue, +21.4% QoQ, and the company had guided above the depressed FY2025 trough. What actually surprised was the magnitude and quality of the beat: revenue of $3,025.0 million was +12.6% versus the street, EPS of $6.20 was +71.3% versus the street, and gross margin stepped to 50.9% from 29.8% in one quarter. The stock debate after a memory print usually centers on whether revenue is being pulled forward; here the cleaner question is whether gross margin has reset to a materially higher level before consensus has had time to rebuild normalized earnings. The company’s own words make that reset explicit without relying on sell-side extrapolation: Luis Visoso said, “Our forecast for non-GAAP gross margin for the third quarter is between 65-67%.” That wording matters because it commits to another step-change immediately after a quarter in which the reported gross margin was already 50.9%, rather than describing a distant-cycle aspiration.

The revenue path supports that margin thesis because the acceleration is not coming off one isolated customer bucket. On the company’s own reported basis, revenue for the second quarter was $3.025 billion, and Visoso framed it as “up 31% quarter over quarter and 61% year over year”; on the street-comparison basis, the same event was $3,025.0 million against $2,687.6 million expected. Those are different lenses, but both point in the same direction: demand exceeded the model while the P&L absorbed that demand at much higher profitability. The segment numbers are important because they reduce the risk that the quarter was a single-end-market squeeze. Edge revenue was $1.678 billion, up 21% sequentially; Consumer was $907 million, up 39% quarter over quarter; and data centers were $440 million, up 64% sequentially. Data centers were still the smallest of the three disclosed buckets, but the 64% sequential growth rate is the one that matters for second-order earnings because enterprise and cloud storage demand typically carries mix implications when supply is tight. Consumer’s $907 million at 39% quarter-over-quarter growth says pricing is not confined to one high-end channel, while Edge at $1.678 billion gives Sandisk scale in the portion of the business that can absorb volume without leaving the quarter dependent on one pocket of AI-adjacent demand.

That breadth explains why the gross margin surprise should be treated as the center of the thesis rather than a footnote. The reported history shows the company moving from -12.3% gross margin in Q1 FY2024 to 9.7% in Q2 FY2024, 27.2% in Q3 FY2024, 36.1% in Q4 FY2024, then back down to 22.5% in Q3 FY2025 before recovering to 26.2% in Q4 FY2025 and 29.8% in Q1 FY2026. This quarter’s 50.9% is therefore not just a new point on a smooth line; it is the first data point that says the recovery has crossed from utilization repair into pricing and mix leverage. The call basis reinforces that conclusion: non-GAAP gross margin was 51.1%, up from 29.9% in the prior quarter, and that compared with guidance of 41 to 43%. A guide miss by expenses can be transitory; a gross margin outcome roughly ten points above the high end of guidance, followed by a 65% to 67% non-GAAP gross margin guide, is the part of the print that should force memory investors to revisit earnings power.

The EPS bridge is just as important because the revenue beat alone cannot explain $6.20 of EPS versus $3.62 expected. Sandisk’s quarterly EPS history had been erratic, with diluted EPS of -$13.33 in Q3 FY2025, -$0.16 in Q4 FY2025, and $0.75 in Q1 FY2026 before this quarter’s $5.15 in the quarterly history and $6.20 on the call’s non-GAAP basis. The print’s street-comparison EPS is the clean surprise metric, and it was +71.3%. The call’s non-GAAP EPS comment is the clean operating metric, and Visoso said non-GAAP EPS was $6.20, up from $1.22 in the prior quarter and above the $3 to $3.40 guidance range. The distinction matters because mixing the company’s non-GAAP EPS and the table’s diluted EPS would overstate precision; keeping them separate shows the same directional conclusion from both bases. Operating expenses helped, but they were not the main event. Non-GAAP operating expenses were $413 million, 13.7% of revenue, below the $450 million to $475 million guidance range due to a non-recurring benefit from changing how new product introductions are managed. Because management itself labeled the benefit non-recurring, investors should not capitalize the full expense underspend. They should capitalize the margin structure if Q3 comes in near the guide.

The balance sheet and cash conversion make the upcycle more investable than the prior loss-making phase because Sandisk is reducing financial risk while margins expand. The company closed with $1.539 billion in cash and cash equivalents and $603 million in debt, leaving a net cash position of $936 million after paying an additional $750 million of debt during the quarter. Adjusted free cash flow was $843 million, representing a 27.9% free cash flow margin, with $1.019 billion from operations partially offset by $176 million from net cash capital spending. Those numbers matter because memory upcycles often tempt investors into paying for earnings before cash appears; here the cash is appearing while the guide still points higher. The one confusing disclosure is capital intensity: Visoso said gross capital spending totaled $5.255 billion and represented 8.4% of revenue, which cannot be reconciled arithmetically with second-quarter revenue of $3.025 billion in the same sentence. The prudent interpretation is not to build a model off that phrase; the defensible conclusion is that the actual cash-flow line, $843 million in adjusted free cash flow, and the balance-sheet line, $936 million of net cash, are the cleaner evidence for reduced downside risk.

That balance-sheet shift also changes how to read the Q3 guide, because management is not merely harvesting a one-quarter pricing squeeze to repair leverage. The guide calls for revenue between $4.4 billion and $4.8 billion, non-GAAP gross margin between 65% and 67%, non-GAAP operating expenses between $450 million and $470 million, non-GAAP interest and other expenses between $25 million and $30 million, non-GAAP tax expenses between $325 million and $375 million, and non-GAAP EPS between $12 and $14 assuming 157 million fully diluted shares. The variant perception is that the market may treat those Q3 figures as a cyclical peak because memory companies rarely sustain supernormal margins, but the guide suggests the next quarter is still in the steepening part of the earnings curve. The important contrast is that operating expenses are guided up from $413 million to $450 million to $470 million, yet EPS is still guided to $12 to $14. That is the clearest signal of operating leverage: the company is not relying on repeat expense favorability to move earnings higher.

The tone of the call was consistent with that financial message, but not unambiguously euphoric, which is useful for PMs trying to separate management confidence from promotional cadence. The tone history shows overall sentiment rising to 0.34 in Q2 FY2026 from 0.29 in Q1 FY2026, while tone_confidence increased to 0.42 from 0.30. Prepared sentiment rose to 0.76 from 0.72, and QA sentiment rose to 0.19 from 0.14. The part that tempers the read is guidance_tone, which fell to 0.23 from 0.43, while uncertainty rose to 42.2 from 33.7. That mix is not contradictory if read alongside the numbers: management delivered a much better quarter and a much higher guide, but the language model still picked up more uncertainty around forward conditions. For the subsequent Q3 FY2026 call in the tone table, sentiment rose again to 0.44, guidance_tone rose to 0.51, and tone_confidence rose to 0.61, but uncertainty also rose to 50.0 and qa_evasiveness moved to 2.5. The call-over-call delta from Q2 FY2026 to Q3 FY2026 was sentiment +0.11, guidance_tone +0.28, tone_confidence +0.19, uncertainty +7.8, and qa_evasiveness +14.0. The delivery read is therefore not “management sounded bullish”; it is more specific: confidence and guidance tone improved after the Q2 setup, while uncertainty and evasiveness also increased, which is exactly the pattern one would expect when the P&L is accelerating faster than supply/demand visibility.

The supply-chain read-through is concentrated because the data pack names one strategic counterparty on both sides: Kioxia. Sandisk’s customers include Kioxia through JV NAND fabs at Yokkaichi and Kitakami, and its suppliers include Kioxia through the same JV NAND fabs. Sandisk’s Q2 revenue of $3,025.0 million, +61.2% YoY in the quarterly history, and gross margin of 50.9% imply that the JV fab ecosystem is participating in a NAND price and mix recovery rather than simply absorbing volume. The second-order implication for Kioxia is not vague positive beta; it is that Sandisk’s reported margin moved from 29.8% in Q1 FY2026 to 50.9% in Q2 FY2026 while Q2 revenue grew +31.1% QoQ, giving Kioxia exposure to both higher wafer economics and tighter downstream availability through Yokkaichi and Kitakami. The peer table supports that read: Kioxia’s latest reported gross margin is 64.1% with revenue YoY of +188.9%, while Sandisk’s latest peer-table gross margin is 78.4% with revenue YoY of +251.0%. That pairing says Sandisk’s margin expansion is not idiosyncratic accounting noise; it is part of a NAND supply chain repricing, with Sandisk currently showing higher latest gross margin than Kioxia by the peer table’s reported figures.

The competitive context makes the Sandisk print harder to fade on the usual “memory is already peaking” objection. In the peer table, MU shows 84.6% gross margin and +345.7% revenue YoY, while Sandisk shows 78.4% gross margin and +251.0% revenue YoY, and WDC shows 50.2% gross margin with +45.5% revenue YoY. Sandisk is not the absolute margin leader versus MU, but it is materially above WDC on both gross margin and revenue growth in that latest reported quarter. That matters for relative positioning because the market may instinctively rotate to the broadest memory winner, yet Sandisk’s print offers a sharper revision setup: a Q2 street EPS surprise of +71.3%, a Q3 non-GAAP EPS guide of $12 to $14, and a balance sheet already at $936 million net cash. The best long thesis is not that Sandisk is the only beneficiary of memory tightness. It is that Sandisk’s earnings model is being revised from a lower base while its reported and guided margins have already moved into a peer-competitive range.

The bear case deserves respect because the data contain two flags that should keep valuation discipline in the position. First, Q2 operating expense favorability included a non-recurring benefit, so the $413 million non-GAAP operating expense level should not be treated as the run rate when Q3 is guided to $450 million to $470 million. Second, tone uncertainty increased from 33.7 in Q1 FY2026 to 42.2 in Q2 FY2026, then to 50.0 in Q3 FY2026 in the tone history, even as sentiment and confidence improved. Those figures argue against extrapolating a straight-line supercycle without checking supply response. But neither flag breaks the thesis because the company’s Q3 EPS guide of $12 to $14 already embeds higher operating expenses, and the gross margin guide of 65% to 67% is the operative variable. If gross margin rolls over before revenue reaches the guided range, the upcycle is more fragile than the print implies. If gross margin lands inside the guide while revenue lands inside $4.4 billion to $4.8 billion, the market’s prior EPS anchor is obsolete.

What to watch next is therefore narrow and numeric. For Q3, the thesis is confirmed if revenue lands within or above $4.4 billion to $4.8 billion, non-GAAP gross margin lands within or above 65% to 67%, and non-GAAP EPS lands within or above $12 to $14 on 157 million fully diluted shares. It is weakened if operating expenses exceed $450 million to $470 million without revenue exceeding the guide, because Q2’s $413 million included a non-recurring benefit. It is broken if gross margin fails to hold the guided 65% to 67% range while uncertainty keeps rising from the Q2 FY2026 level of 42.2, because that would mean the call’s caution was a leading signal rather than ordinary cycle visibility. The next dated checkpoint is the Q3 FY2026 period in the company’s sequence after the 2026-01-29 call; investors should anchor the debate on whether Sandisk can convert the guide into the kind of latest-quarter profile shown in the history, $5,950.0 million of revenue, 78.4% gross margin, +96.7% QoQ revenue growth, and $23.03 of diluted EPS, because that is the level at which this stops being a recovery trade and becomes a normalized earnings reset.