11.6.25
Company executives predicted that the weakness in the NAND market seen earlier this year would prove to be a “mid-cycle pause.” Skeptical is not a strong enough word to describe my view of this prediction. Alas, SanDisk’s management appears to be in the right and I am in the wrong. The company guided for revenue between $2.1B and $2.2B. Actuals were above the high end of the range at $2.3B. That is $150M above the midpoint of guidance, following the $100M beat last quarter. Management’s optimism for the recovery in the NAND market has still been short of the actual strength seen so far. Non-GAAP gross margin of 29.9% was above the guided range of 28.3% and 29.2%. GAAP gross margin was 29.8% for the quarter. That is 340 bps higher than last quarter. Company GAAP gross margin for the previous nine quarters, from oldest to most recent have been (10.3%), 7.9%, 27.4%, 36.5%, 38.9%, 32.5%, 22.7%, 26.4%, and now 29.8%. Now the market is really accelerating. GAAP guidance for the fiscal second quarter of 2026 is for revenue between $2.55B and $2.65B with gross margin between 40.8% and 42.8%. That jump would completely skip the 30s and would be a stunning rise. The NAND market is back, baby!
From last quarter to this quarter, the company changed the names of their business units. The changes were->are: Cloud->Datacenter, Client->Edge, Consumer->Consumer (the same). Revenue surged across all three business units. Datacenter was up 26% sequentially. That BU was up only 8% in the prior quarter. Client saw the largest increase last quarter (19%), and this quarter it’s sales were up 26% sequentially. Consumer revenue was up 11%, following a 2% sequential rise last quarter. We have been seeing low investment in NAND capacity for almost two years. I think that reduction in supply has finally caught up with demand across the industry. Sure, the boom in AI data center investment is the top draw, and is taking bits out of other segments, but the big increase in sales to PCs, mobile phones, and consumer applications is more a function of pricing and customers buying ahead of further rises than of major change in demand. The company says global datacenter and AI investments are accelerating. I’ve heard this from other memory companies as well. Management gave a long-term total as the only forecast in this segment, $1T in total datacenters and AI infrastructure by 2030. That is a base-on-balls. Management cites the PC refresh cycle and higher NAND content per phone as drivers of demand in these areas. They say something like that every quarter.
Last quarter, SanDisk reported a mid-single digits percent rise in NAND ASPs sequentially. They saw the same rise this quarter. Unlike last quarter, the other three NAND companies I follow (Samsung, Hynix, and Micron) also all showed increases, indicating a broad-based recovery in the health of the NAND market. Pricing rose some, but the company’s bit shipments rose even more. They were up mid-teens percent quarter-over-quarter. Over the last two years, SanDisk’s NAND bit shipments have risen a total of 15%, so less than a 7% CAGR. The NAND upturn started in the fall of 2023. Pricing was down in the third quarter and then recovered sharply in the fourth quarter. Bit shipments turned upward two quarters earlier. Going back to that inflection point (the bottom being the first calendar quarter of 2023), SanDisk’s total NAND bit shipments have risen around 67%. That is over a period of ten quarters. That is similar to both Hynix’s and Samsung’s bit growth. It is Micron that has been the outlier over this period. Hence Micron has gained market share while those other three companies have maintained, not including the growth of YMTC. Unlike the other three companies, SanDisk did not see a massive rise in NAND bit shipments last quarter. I don’t know why this is. SanDisk is the only one of the four companies that didn’t absorb a price decline last quarter. They may have held back some potential shipments in order to preserve pricing, anticipating even better pricing in the future.
This section is commentary from management’s conference call with analyst.
· The CEO, in his opening comments, reiterated his prediction that demand for their NAND products will outpace supply through the end of calendar 2026 and beyond. I continue to think making a prediction like this, going out more than thirteen months, is nuts.
· The CFO said the same thing in his prepared remarks, that the market will remain undersupplied through the end of calendar 2026. They are on product allocation across all of their end markets. This means the NAND market is in a serious undersupply situation.
· They have data center customers coming to them seeking multi-quarter supply and price agreements. They are seeking certainty of supply through into 2027. I bet this is why they have the surprising level of confidence they have in NAND being undersupplied all the way through 2026. This sounds similar to what the DRAM companies are saying that they have customers seeking supply agreements going out several quarters. I think the NAND companies have even less leverage than the DRAM companies do. There are more of them.
· They are investing in capacity to meet a mid-to-high teens percent bit demand growth rate over the long term. They don’t want to invest so much that they disrupt the market, so they are targeting to keep their market share.
· Three months ago they thought data center bit demand would be up mid-20% level in 2026. They’ve now upped that to mid-40%. That is a large change in just three months. The market is moving quickly. SanDisk aims to get their fair share of that market.
· 2026 will be the first year that the data center market is the largest market segment in NAND.
· Their fabs are all running at 100% capacity now.
· They saw supply growth of 8% in 2025 and expect this to be 17% in 2026.
· They have “very little volume and price commitments that are beyond” one quarter. Said another way, they have customers that are interested in longer arrangements now that the market looks really undersupplied. I am doubtful that they will actually get to take-or-pay agreements.
· Multiple questions and answers talked about how the strength of demand from data centers is “structurally” changing the NAND market. The four most dangerous words in business are “this time it’s different.”
The enthusiasm for building data centers to supply AI compute has exploded in the last two to three months. That demand has spread to the NAND market in the form of demand for enterprise SSDs. I think that has collided with underinvestment in NAND capacity over the last two years, causing the current shortage of bits in the market and resulting rise in prices. This quarter, the company’s second fiscal 2026 quarters, should show a large rise in prices. I think what is happening is hyperscalers have raised their investment plans so much over the last quarter or so that the memory suppliers are hearing from their customers that they want supply certainty several quarters out. The NAND market has now been drawn up into the AI investment cycle. I think it is crazy that SanDisk’s executives, people who have been around memory for a long time, are believing that this time it’s different. They clearly are. They believe NAND will be undersupplied all the way through 2026 and beyond. I think their customers are so urgent for supply that they believe severe undersupply will continue for the foreseeable future. Maybe they are right. Maybe this time really is different. I’ve been around memory for too many cycles to believe that is the case. But maybe this is the upturn that proves people like me wrong.
– S. Hughes (short MU)