Selling Fei SNDK I am not sad

Ask AI · How does the author’s position strategy reduce the regret of missing the sell?

Starting from over 600, I gradually sold SNDK, and up to now, it has risen to $1,562 per share, while I have already sold nearly 60% of my SNDK position. It’s a bit regrettable, but I’m not upset.

I didn’t start building my position in SNDK early; it was at the end of January this year, just one day after it released its earnings report, that I began gradually accumulating. Through batch buying and buying on dips, I executed my position-building process quite well, with an average cost around $560. At that time, it was a high price because it had already surged nearly 20 times in just a year.

Fear of heights + insufficient cognition determined that I couldn’t calmly face the fluctuations at that time, and during the 2-3 months after I completed my position, the US stock market was in a continuous decline, so I started reducing my holdings around $650.

Fortunately, I have the habit of reducing positions in small batches, usually selling about 5% each time, and sometimes doing some rebuys when opportunities arise. But even so, I had already sold nearly 60% of my SNDK position.

I think all the missed opportunities can be explained with one sentence: you can never earn more than your cognition allows.

But today, I want to explain in detail why I don’t feel regretful, and it’s not just because of the above paragraph.

In short, my calm acceptance of this “mistake” is related to my own understanding of storage and investment strategies, which are also connected to SNDK’s position in the storage industry.

By the way, let’s first briefly look at SNDK’s financial performance.

Note that SNDK’s fiscal year begins on July 1 and ends on June 30 of the following year. Its fiscal Q3 2026 actually covers the period from January 1 to March 31, 2026.

As of the March quarter (fiscal Q3 2026), SNDK’s revenue was $5.95 billion, up 97% quarter-over-quarter and 251% year-over-year. Non-GAAP gross margin reached a record high of 78.4%, operating profit was $4.22 billion, with an operating margin of 71%, even higher than Nvidia. Adjusted earnings per share reached $23.4, far exceeding expectations.

Gross margin increased by 55.7 percentage points year-over-year and 25.3 percentage points quarter-over-quarter—absolutely explosive. In this quarter, NAND prices increased by about 90% quarter-over-quarter, and SNDK’s revenue grew by 97% quarter-over-quarter, meaning the absolute driver of revenue growth was the soaring ASP of NAND flash memory. Almost all of this revenue increase was converted into profit, with revenue up by $2.92 billion quarter-over-quarter and operating profit up by $3.08 billion—more than the revenue increase itself. Year-over-year, revenue increased by $4.25 billion, and operating profit by $4.22 billion—almost matching the revenue growth.

Due to product price increases, but factory capacity expansion being delayed, operating costs hardly needed to increase, so the revenue increase naturally translated into profit.

Using a simple annualized estimate, assuming this level of earnings continues for the next four quarters, the next fiscal year’s EPS would reach about $94. At the current stock price, with a PE ratio of 16.6, but you know that just a month ago, its stock price was only half of the current level, meaning the implied PE at that time was only 8.3.

In reality, SNDK’s EPS for the next fiscal year will definitely be far more than $94.

According to the company’s guidance, SNDK’s revenue for the next quarter is expected to reach $8 billion (based on the median of the guidance), with a gross margin of 80%, and EPS of $31.5.

In the second quarter of this year (Q4 of SNDK 2026), the trend of price increases in the storage industry continued, with NAND expected to rise about 40% quarter-over-quarter. Management’s guidance for revenue growth is 34% quarter-over-quarter, which is probably still a conservative estimate; at the upper limit, it could be 39%.

Now, the whole world knows that storage shortages will last at least until the end of 2027. Very conservatively predicting, SNDK’s EPS for fiscal 2027 could reach $135. At the current stock price of $1,562, that’s a PE of 11.6, still not expensive.

In fact, in the next fiscal year, SNDK’s EPS will surely be well above $135.

Based on the company’s guidance, SNDK’s revenue for the next quarter is expected to reach $8 billion (median estimate), with an 80% gross margin, and EPS of $31.5.

This year’s Q2 (Q4 of SNDK 2026) continues the trend of rising prices in the storage industry, with NAND expected to increase about 40% quarter-over-quarter. The management’s guidance for revenue growth is 34%, possibly still conservative; at the upper limit, it could be 39%.

Now, everyone knows that storage shortages will last at least until the end of 2027. Very conservative estimates suggest that SNDK’s EPS for fiscal 2027 could reach $135, and at the current price of $1,562, that’s a PE of 11.6, still quite reasonable.

In the context of the US stock market re-pricing the storage industry, gradually abandoning the valuation model of cyclical stocks and shifting to infrastructure growth stocks, if SNDK’s PE ratio reaches 25—which has a 70% probability—this corresponds to a stock price of $3,366, with an upside of 115%.

If earnings further beat expectations, with EPS for 2027 rising to $150–180, plus some market bubble factors, reaching $5,000 per share is not impossible.

But more importantly, we must consider not only investment returns but also risks.

Honestly, I believe that the oligopoly nature of HBM will be highly predictable to trade at around 25 PE in the future. NAND may not be as lucky, or even if it is, its high PE window will be shorter. How to avoid top risks or entering narratives that may never materialize within a relatively shorter timeframe is worth serious consideration.

When we talk about storage, what exactly are we talking about?

Storage mainly divides into DRAM and NAND products. Both are widely used in consumer electronics, enterprise, data centers, and AI accelerators. DRAM is further divided into regular DRAM memory and high-end HBM storage products. HBM is mainly used in AI accelerators, packaged together with GPUs for data center data storage.

NAND and DRAM inherently differ in industry structure and manufacturing barriers. The difference between NAND and HBM within DRAM is even greater:

First, NAND is essentially a standardized commodity with strong substitutability for customers choosing NAND. HBM, on the other hand, is already part of GPU performance; it works when packaged with GPUs. The process involves Nvidia’s design (architecture between chip and storage), storage manufacturers’ R&D and manufacturing, and TSMC’s packaging. Comparing the two is like car tires vs. engines + gearboxes—tires can be easily replaced, but engines and gearboxes are integral to the vehicle system.

The complexity of the supply chain makes expanding HBM capacity much more difficult. Even if storage manufacturers produce more HBM, TSMC’s packaging capacity is limited. Moreover, the technical difficulty of expanding HBM is higher than NAND.

Second, NAND has lower market concentration than HBM, and industry expansion motivation is much greater. HBM is almost a complete oligopoly, with Samsung, Micron, and SK Hynix dominating the global HBM market. There are no real competitors. Based on historical experience, the three giants are more likely to reach a consensus on shared interests, leading to convergence in capacity expansion speed. The NAND market, however, is more fragmented, with not only the three HBM giants but also Japan’s Kioxia, Western Digital’s spin-off SNDK, and China’s Yangtze Memory. Historically and currently, this market is more susceptible to expansion driven by China’s supply chain growth.

However, the current reality is that neither NAND nor HBM capacity can keep up with demand growth before 2027. That’s why SNDK, Micron, Hynix, and Samsung can enjoy a period of valuation expansion at this stage.

From an industry cycle perspective, NAND’s cyclicality will be stronger, and the cycle will end sooner.

Moreover, SNDK’s financial reports show that its performance trend actually reversed in fiscal Q4 2025 (Q2 2025), although the market’s reversal started several months earlier.

This is why, over the past year, the rise of SanDisk has led the entire storage industry: it is a combination of turnaround from adversity and the super cycle of storage AI. Meanwhile, companies like Micron already experienced a rally in 2023.

Please note, I am not saying that SNDK is currently a risk-reward disproportionate target. I tend to believe it still offers a high reward-to-risk ratio.

I just think that compared to Micron, Hynix, Samsung, it is more likely to retreat earlier and more sharply, making it a higher-risk asset.

Therefore, in my position allocation, I have allocated more to Micron, and in my reduction strategy, I prioritize it.

(I’ve sold about 2% of my Micron; if opportunities arise, I might buy back below my selling price, but if not, so be it.)

These ideas have always been like this, and I’ve shared them multiple times with my members.

From a purely psychological perspective, I also find that selling SNDK leaves me only slight regret but no sadness; after selling AMD, I felt very anxious, even though at that time, I sold AMD mainly to increase my position in Micron.

I had planned to sell large tech stocks when US stocks reached unreasonable levels, swapping for more flexible Micron. AMD was one of the stocks I planned to sell (along with Amazon, Meta, Nvidia, and Broadcom, which I sold at lows to buy Micron). Unexpectedly, AMD’s rebound was even stronger than Micron’s, outperforming significantly in April.

Later, I added to AMD in batches at 320/348/397/410/405, accumulating to over 2.5 times my original position.

I also believe that HBM might truly become de-cyclicized, turning into a long-term AI infrastructure bottleneck similar to GPUs (of course, from a longer-term perspective, even GPUs have cycles, which is another story). NAND, on the other hand, is more enjoying the current explosive demand in AI, with capacity possibly catching up within three years, leading to price wars. When major funds realize the risks (they usually trade performance 12–18 months ahead), they might shift investments to Micron, Hynix, Samsung, and other HBM-related stocks. This means that at certain stages, if you see Micron leading and SNDK not following or even declining, it could be a signal that the major funds are trading the top in advance.

If you want to understand SNDK’s trading cycle thoroughly, selling on the right side might be one of the references.

As for myself, if SNDK reaches the 2000–3000 range (if it gets there), I will still adhere to a high sell and low buy strategy. Above $3,000, I might only sell and not buy back easily.

For AMD and Micron, I will hold steadily with no immediate plans to sell.

Created a paid group chat, including a paid comprehensive US stock group, a paid Chinese concept stock group, and more than ten paid groups for cryptocurrencies, MSTR, Hims, NBIS, Nvidia, Tesla, Apple, Google, Meta, Amazon, Microsoft, TSMC, and other US stocks.