The entire next-quarter guide is price. Bits are guided lower while revenue is guided up 46% sequentially.
Thesis: NAND pricing power is real and cash-confirmed: $843M FCF on a 29.9% to 51.1% GM swing with bits roughly flat, JV supply locked to 2034, and no capacity additions beyond mid-to-high-teens bit growth. The market is pricing this as a commodity cycle while data center becomes the largest NAND market and hyperscalers compete for allocation. The bet is that supply discipline holds longer than the street underwrites, and each signed LTA de-risks the model further.
Verdict: LONG — Conviction: MEDIUM
Catalyst: Second and third LTA announcements with prepayment components, plus Q3 delivery of $4.6B at 66% GM proving the ASP-driven guide was not sandbagged.
Key Risk: Q3 guide is 100% ASP on mid-single-digit lower bits, and data center is only 15% of revenue. Consumer and edge, which grew 39% and 21% sequentially, absorb the same allocation pricing that reverses first if hyperscaler CapEx pauses.
The Tell: Goeckeler volunteered, unprompted, that NVIDIA key value cache demand of 75-100 exabytes in 2027 is in none of the current numbers and doubles the year after. He volunteered upside beyond guidance, which reads as confidence but also as narrative loading ahead of LTA conversion.
Friction Level: MODERATE_FRICTION — Both sides agree on the facts: demand exceeds supply, no capacity additions, cash is real. Disagreement is whether a 15-point margin gain delivered entirely on flat bits is a structural reset or a cycle peak.
2026Q2 · Jan 29, 2026
4 of 5 readings agreed4 LONG · 1 HOLDRead the full call
Revenue Guide Up 46%, Bits Guide Down Mid-Single Digits
NAND pricing power is real and cash-confirmed: $843M FCF on a 29.9% to 51.1% GM swing with bits roughly flat, JV supply locked to 2034, and no capacity additions beyond mid-to-high-teens bit growth. The market is pricing this as a commodity cycle while data center becomes the largest NAND market and hyperscalers compete for allocation. The bet is that supply discipline holds longer than the street underwrites, and each signed LTA de-risks the model further.
Goeckeler volunteered, unprompted, that NVIDIA key value cache demand of 75-100 exabytes in 2027 is in none of the current numbers and doubles the year after. He volunteered upside beyond guidance, which reads as confidence but also as narrative loading ahead of LTA conversion.
+72
CONFIDENT"Direct answers with specifics: quantified the one-time OpEx benefit at $35M, disclosed one signed LTA with prepayment terms, gave tax rate guidance of 14-15%, volunteered that no underutilization charges exist in actuals or guide."
Full P&L, balance sheet, FCF, segment splits, bit growth plans, JV terms with dollar figures, and named qualifications at a second hyperscaler. Gaps only on LTA terms and ESSD revenue share.
Second and third LTA announcements with prepayment components, plus Q3 delivery of $4.6B at 66% GM proving the ASP-driven guide was not sandbagged.
Q3 guide is 100% ASP on mid-single-digit lower bits, and data center is only 15% of revenue. Consumer and edge, which grew 39% and 21% sequentially, absorb the same allocation pricing that reverses first if hyperscaler CapEx pauses.
“Revenue for the second quarter was $3.025 billion, up 31% quarter over quarter and 61% year over year.”
“For the third quarter, we expect revenue between $4.4 billion and $4.8 billion.”
Figures are extracted from the transcripts and checked in code: every quote appears word for word. Descriptive only: none of this feeds the verdict.
Beat-and-raise on every line: revenue 15% over high end, EPS double the guide, GM 800 bps over guide. Cash confirms it, $843M FCF and $1.4B of debt retired in a quarter. Supply is locked: fabs at full utilization, no capex change, JV extended to 2034, no spec build anywhere. Demand is not speculative either, enterprise SSD went 29% to 64% sequential with a second hyperscaler qualified and a third coming. Pricing power is company-wide, consumer +39% because bits flow to the highest-value bid. The street models 30-35% through-cycle GM; management just printed 51% and guided 66%. Each signed LTA with prepayment locks the price. High conviction long.
The Q3 guide is 46% sequential revenue growth on bits guided down mid-single digits. That is a 15-point GM expansion delivered entirely by price on flat supply, in a market where consumer and edge absorb the same allocation pricing as data center. Capacity is a hard ceiling presented as discipline: full utilization, no capex change, expansion limited to Kitakami K2 space. The structural story that justifies the re-rate is one signed LTA with undisclosed terms, several more 'in the queue.' Data center exabyte forecasts moved 2.5x in nine quarters with zero commitments behind them. KV cache and HBF are AI Label Without Substance, die in design, no revenue. Data center is 15% of revenue, so 85% of this print is cyclical consumer and edge pricing. The cash is real, the cycle is not.
Unsustainable Trend Confidence
Say/Do Gap
Backlog Quality
AI Label Without Substance
LONG, medium conviction. Pricing power is proven in cash, not narrative: $843M FCF, bits roughly flat, no capacity additions, supply locked to 2034. The asymmetry sits with supply discipline holding through a hyperscaler upcycle while the street models 35% cycle GM. Size for the guide to print, but do not add after $13 EPS prints. The tell to exit is the second LTA not closing or a hyperscaler cutting CapEx, because 85% of revenue is consumer and edge pricing that reverses first.