In the era of AI reasoning, NAND explodes, SanDisk (SNDK.US) Investor Day may unveil a new growth curve! Storage bull market awaits next fundamental catalysis

The Zhitong Finance App learned that SanDisk Investor Day on August 13 may become an important catalyst for memory chips and even the global AI computing power industry chain. The second quarter (the company's fourth finance) performance previously announced by the US NAND memory chip giant SanDisk (SNDK.US) exceeded expectations, but the revenue and profit outlook for the next quarter was slightly lower than the market's escalating expectations. This also highlights that demand for enterprise-grade NAND memory chips used in artificial intelligence data centers continues to expand, but the market's expectations for future growth of SanDisk are becoming more strict, and may be triggered by less than expected Stock prices fluctuated sharply. Any signals of strengthened growth prospects or NAND demand and technology iteration signals released by management on “SanDisk Investor Day” may become a “fundamental confirmation” for the continued rebound of global storage stocks.

The performance outlook announced by SanDisk management during the performance conference call last week did not significantly exceed expectations, and management also gave guidance that gross margin would trend towards the platform period and remain around 83% to 85%, corresponding to annual earnings per share (EPS) of approximately $180. If calculated at a price-earnings ratio of 10 times, the corresponding share price is approximately $1,800. Currently, SanDisk's stock price is hovering around $1,350. So essentially, the market doesn't believe that 80% or more gross margin is sustainable, but is pricing a lower future EPS.

The market has acquiesced that the core performance data for at least the next two quarters exceeded expectations. Now, the focus has turned: once NAND flash prices stop rising and supply gradually catches up with demand, can management send a powerful signal to the market about what level the normalized gross margin will actually be and the growth prospects for 2027, or whether it will reveal any technical iteration signals about HBF, one of NAND's largest structural demand increments in the next round.

More specifically, investors will focus on this signal: whether the company's new business model contract based on LTA (i.e. long-term NAND supply agreement) can continue to support higher gross margins and the current increasingly strong enterprise data center NAND business revenue structure. Only then will investors believe that capital will substantially revalue this stock once again. Before this signal appeared, perhaps any increase at the weekly level was simply a fluctuation clothed in the guise of an increase. If fundamentals don't change, some analysts don't think the stock price can rise again to an all-time high.

Kioxia, Sandisk (Sandisk), SK Hynix, and Samsung clearly define high-bandwidth flash memory (HBF) as a new form of NAND for AI “memory walls”. The goal is to provide greater capacity in AI inference, and claim that HBF can achieve comprehensive transmission performance close to “unlimited capacity HBM” in related inference tests, while significantly increasing available memory capacity. SanDisk and SK Hynix have released the first OCP technical specification to try to place high-capacity, long-lasting NAND closer to AI accelerators to ease the “memory wall” in the inference phase, increase bandwidth and reduce the overall cost of token services; it is not a short-term replacement for HBM, but may create a new near-computing storage layer between HBM and traditional SSDs.

NAND ushered in the “age of inference” key verification! Facing HBF commercialization on SanDisk Investor Day?

SanDisk (or “SanDisk”) Investor Day is indeed likely to become a key catalyst for global memory chip transactions and even AI computing power infrastructure deals, but what the market really wants to hear is not “how much more can the price increase next season”, but three things — NAND normalizes gross profit margin, profit stability of long-term supply agreements (LTA/NBM), and the timeline for HBF (High Bandwidth Flash, or High Bandwidth Flash) to commercialization.

SanDisk and SK Hynix just released the first OCP HBF technical specification on August 3. AI technology leaders such as Google (Google) and Tenstorrent participated in the verification, clearly positioning HBF as a new memory layer in AI inference systems that is close to xPU and takes into account high capacity and high bandwidth; on August 12, it also released a next-generation 2TB QLC 3D NAND for AI infrastructure.

HBF (high-bandwidth flash memory) is likely to become one of the largest structural demand increases for NAND in the next round — even some analysts emphasize that the latest HBF technology plans and prospects mean “NAND is trying to replicate the historical moment of HBM”. SanDisk and SK Hynix just released the first open HBF technical specification in August: it is based on 3D NAND, but it is no longer just remote SSD storage, but is close to the CPU/GPU/xPU through advanced packaging and UCie interfaces. The maximum capacity of a single package is 512 GB, 0.4 to 3.0 TB/ s Bandwidth; Sandisk (Sandisk) plans to provide the first batch of HBF samples in the second half of 2026, and the first batch of AI inference devices equipped with HBF is expected to enter the sample stage in early 2027.

Therefore, on Investor Day, management further explained the HBF roadmap, the pace of mass production/customer introduction, eSSD and QLC expansion, and how the new business model is likely to maintain high gross profit. However, investors need to be wary that the “HBF Technology Route and Major Commercialization Notice” cannot be regarded as a definitive disclosure event until the conference is officially disclosed.

SanDisk's total revenue for FYQ4 (the fourth fiscal quarter) has reached 8.97 billion US dollars, a sharp increase of 51% over the previous year. About two-thirds of the increase came from prices, while pointing out that total revenue related to its data center business increased sharply by about 400% year over year; if SanDisk management can further upgrade “high ASP” to “long-term contract+AI storage structural increment+new HBF product cycle”, it may further boost the recent strong rise in SanDisk and global memory chip stocks.

More importantly, the current “AI storage bull market” already has hard evidence of the cross-company and cross-medium industrial chain, not just SanDisk's single-point boom. Samsung clearly anticipates that demand for server DRAM, eSSD, and HBM will accelerate further in the second half of 2026, and even if demand for mobile phones/PCs slows down, the industry will remain in short supply; SK Hynix claims that Agentic AI is expanding the entire DRAM/NAND memory chip demand base. Both NAND and DRAM prices rose significantly in the second quarter, and will continue to shift NAND production capacity to 321 layer, high-capacity and high-performance products.

Western Digital's fourth fiscal quarter revenue increased sharply by 44% year over year to 3.75 billion US dollars, and Seagate's total revenue increased 48% year over year to 3.63 billion US dollars, and free cash flow was 1.1 billion US dollars. Both companies emphasized the need for large-scale storage brought about by AI cloud data centers; what needs to be strictly distinguished is that Western Digital and Seagate now mainly verify HDD/mass-capacity storage (HDD/mass-capacity storage) requirements rather than directly prove NAND requirements, but they all together show that the AI reasoning era is creating a larger data life cycle — Training data, model checkpoints, RAG databases, logs, inference context, and long-term data retention will push requirements all the way from HBM to DRAM, eSSD/NAND, and HDD. The South Korean stock market, where the two major storage giants hold 50% of the weight, rose sharply by more than 20% from the July 30 low and entered a technology bullish region. Together, it can be said that the HBM/DRAM/NAND storage supercycle driven by the AI infrastructure frenzy is still going strong.

Damo bears switch to multiple verifying that the storage super cycle is yet to be completed

As far as the bullish sentiment in memory chip stocks is concerned, Shawn Kim, a senior analyst at Morgan Stanley, went from being a “bearish spokesperson” to believing that the adjustment is nearing its end, which is an important market signal, but the more accurate understanding is not “he suddenly thinks that storage prices will always rise”, but rather that the market has overtraded the “second-order price derivative decline.” According to information, Shawn Kim believes that the most drastic adjustment of memory chips in the short to medium term has come to an end. The current valuation provides a tactical re-entry opportunity and raised SK Hynix's 2026 EPS again.

According to the Morgan Stanley research report, the Q3 NAND contract price is expected to rise sharply by 20% month-on-month against the backdrop of a high base base in the previous quarter, and the industry is continuing to shift production capacity from the consumer side to ESSD; at the same time, Kim is shifting the next phase of the stock price driver from simple ASP to “LTA long-term supply agreement+free cash flow growth trajectory (FCF) + strong capital return.” The analyst's latest opinion also revealed that the next stage of NAND investment logic is shifting from “NAND price increase beta” to “AI inference demand+increase in ESSD content volume+long-term contract price lock+new HBF architecture alpha.”

From the perspective of AI system engineering, the biggest difference in this round of storage supercycles is the physical bottleneck of HBM/server DRAM/enterprise SSDs being upgraded from ordinary cycle products to AI computing power systems. Musk rarely commented on the storage market during the SpaceX earnings call for the second quarter of 2026. In the conference call, he said that storage supply is growing by about 20% every year, but the demand growth rate is 200% or even higher. There is a serious imbalance between supply and demand, and price increases are a basic law of economics. Guo Luzheng, CEO of SK Hynix, said earlier in a July conference call that 2027 may be the tightest supply year in the history of the global storage industry, and that the state where customer demand exceeds the company's supply capacity may continue beyond 2030.