Hello everyone, due to health issues this week, this weekend's special analysis will be posted in the group in the form of a short article. The core ideas and logic remain unchanged, but limited by energy and time, it won't be as polished as before. Considering my life goals and the market environment over the next six months, I think I might adopt a format where I won't spend so much time writing long articles unless necessary.

If you joined the group mainly to read my long articles or plan to do a lot of intraday trading and copy trades, this decision might disappoint you. You can privately message me about a kind refund. However, our discussions within the group about the U.S. stock market, semiconductors, the storage industry, investment philosophy, and experience will not stop. They will still be updated 24/7 as before, and I will discuss and manage the group with everyone at any time. After all, the group already has enough members, mainly focusing on market information discussions and Q&A. For myself, I will continue holding my SNDK position long-term, aiming for the 2028 target unchanged. But in the next six months, I plan to calm down and wait for new opportunities.
Back to today's topic. Actually, there are two main topics today: first, why I suddenly turned bearish in the first week of July; second, how to understand the sudden outbreak of so many "bearish" news last week.
Honestly, the main reason for turning bearish is technical. Last Thursday, after the Korean stock market deleveraged and Hynix dropped 15%, SNDK and MU each broke below their long-term support levels on the 20-day moving average during the upward trend, even breaking below the 30-day moving average, which was very alarming. Although the Korean stock market rebounded strongly on Friday due to Samsung's DRAM price increase of 20% and Meta executives helping Zuckerberg "cover up," this was also due to the relatively large scale of the Korean 2x leveraged ETF. A real rebound wouldn't be so fast, and there may still be further pullback next week.
Looking back at MU's earnings report two weeks ago, many things were already hinted at. The decline before the earnings report was thought to be profit-taking due to uncertainty about the earnings results, but it turned out to be a rehearsal for institutions starting to short. However, MU's earnings were indeed impeccable, and short sellers temporarily stopped losses, leading to a rebound. But if you look closely at this so-called rebound, you'll see that within just four days, SNDK experienced multiple sharp rises and falls, repeatedly being pushed back at the 2200 previous high — once last Friday and once on Tuesday. This wide range of volatility not only shows that bulls cannot maintain high levels but also reveals the shorts' target. The decline is not simply profit-taking and exiting; it's more likely that institutions, after detecting certain signals, began systematic shorting. Therefore, after multiple pullbacks, they still haven't retreated — of course, there's also the possibility of "pump and dump."
Personally, I feel this pullback is different from previous ones. This time, it may last longer — at least three months (a quarter) or even up to six months. Of course, even if the decline is not a straight line downward, after breaking below the moving averages, the previous support levels become resistance levels, with repeated volatility and shorts staying long-term, causing the stock price to decline amid fluctuations.
Given this judgment, my conclusions are as follows:
1. Long-term investors should learn to "play dead" and wait until 2027 to see if there are better opportunities to add positions or enter. My cost basis is relatively low, so I can endure.
2. For short-term positions, first, don't go long; don't go against the trend. Second, if there is a big rebound on Monday that brings me back to breakeven, I will choose to clear short-term positions directly, including all leveraged ETFs and positions where you cannot tolerate a 30% drawdown, such as those entered at high levels around 2000.

There are basically three chances to deleverage next week: first, if the market opens higher on Monday, consider clearing leverage; second, on Tuesday during Samsung's earnings report, which should theoretically exceed expectations — if it pumps, you can clear leverage again; third, on Friday, during Hynix's IPO. You can clear everything on Monday, or split it into three times, or even wait until Friday if you're bold. But regardless, in an environment where short positions are clearly exposed, I will not hold leveraged long positions. I will finish clearing by Friday at the latest.
During the upcoming pullback and consolidation, after clearing leverage, the best strategy for retail investors is actually to do nothing. In the short term, the stock market is just a big casino. The reason people lose money in the casino is sometimes not that they misjudged, but that the leverage is unequal — you always lose big and win small, and over time, you lose. Only by holding long-term and "playing dead" can you accumulate small gains into big gains.
If the first half of the year saw a one-sided upward trend that allowed gambling retail investors to make money by bottom-fishing, doing T+0 trades, and buying calls, the second half may not be so simple. Since it's impossible to predict daily stock prices or even weekly fluctuation ranges — especially when the 30-day moving average support can fail at any time — I think doing nothing is actually the best. Preserving principal and reducing risk are the most valuable qualities for retail investors in this kind of market.
That's my reason for turning "bearish," summarized as follows:
First, a comprehensive collapse of technical support indicates that this pullback is not simple;
Second, from before the earnings report to today, over two weeks, I've seen a systematic and organized attack by shorts, different from the previous profit-taking at highs.
Unfortunately, the upward channel for storage may have indeed ended last Thursday. That triple top (at the 2200 level) marks the short-term top, and the Nasdaq also showed a diamond top pattern. Of course, this involves my subjective judgment, and I hope the market proves me wrong. But I think, with such continued volatility expansion and the candlestick structure clearly at the top, being cautious is never a bad thing.
If you are very confident in your operations, feel free to keep going long. Or trade long on a simulated account. This way, you can test how much loss "not operating" can save everyone.
Now, let's talk about the so-called bearish news from last week. Ironically, my decision to turn bearish and conclude a "top" had nothing to do with last week's news.
Last week's bearish news mainly came from Meta, two in total. First, Meta inadvertently revealed that its infrastructure CapEx, based solely on developing its own model, is unlikely to recoup costs. As we said before, there may be no one on this planet who wants to use Meta's large language model for work. So management, in a snap decision, decided to follow SpaceX's old path and compete directly with Neocloud by selling computing power. This behavior was interpreted by the market as "excess computing power," resulting in a semiconductor avalanche.
The second piece of bad news came from Zuckerberg's surprising remarks in an interview the next day, saying that the progress of developing Agents was not as expected, implying that AI Agents are not useful and that the benefits generated by AI are far less than those of the programmers he just fired.
Neither of these two pieces of bad news is truly bad. On one hand, Meta claims "excess computing power" and considers launching "Meta Compute" to rent out computing power; on the other hand, it has signed new computing power lease agreements with multiple companies — a $27 billion deal with Nebius in March for AI computing capacity, an expanded $21 billion long-term AI cloud capacity agreement with CoreWeave in April, and in June, it was reported to have purchased about 1.6GW of computing capacity from Crusoe's two data centers. Meanwhile, Korean media reported that Meta is negotiating with Samsung Foundry for a foundry contract worth about 10 trillion won (approximately $6.5 billion) for MTIA AI chips. If computing power is truly excess, why would it buy others' computing power and make new chips? The talk of selling computing power is likely just a characteristic of this company's short-sightedness. After six months of working on large language models and finding they indeed couldn't beat competitors, the stock price plummeted, putting pressure on management, so they decided to try this as a last resort.
Furthermore, after Google's earnings last quarter, it was rumored that Meta would also use bond issuance or ATM stock sales to finance its CapEx. Monetizing computing power to boost the stock price is likely a preparation for better conditions for the next financing round. This is not a negative for global computing demand but a major positive.
Finally, regarding the so-called failure of Agent development, it's entirely because Zuckerberg himself is not an AI engineer, completely different from OpenAI and Anthropic management with technical experience. He evaluated AI Agents from an ROI perspective: Agents are not only more expensive but also not yet capable of replacing the programmers he just fired. As soon as Zuckerberg finished speaking, his subordinate Alexandr came out to clean up, saying that Meta's "Waterlemon" is about to be released. The contradictions in this company's management's words and actions left everyone speechless.
In summary, Meta's behavior and remarks are entirely caused by the company's own lack of technical strength, product-market mismatch, and short-sighted corporate culture. In today's rapidly evolving AI technology, its behavior is almost like a clown. I'm surprised that it can cause such market volatility. It seems many Wall Street institutions still place great importance on Meta's role.
But from another perspective, this also increases my bearish sentiment for the second half of the year. Clearly, a group of institutions has started shorting, and Meta is just an excuse for them to start exerting force.
The real reason for the pullback is actually the inevitable "reflexivity" of the system after storage prices rise to unsustainable levels. After all, DRAM prices are so high that the market can no longer bear them; and continuous price increase positives make bulls fatigued, giving shorts an opportunity. The so-called "good news is bad news" — when storage prices continue to rise in the second half of the year (Trendforce expects about an 18% increase) and gross margins reach as high as 90% (this is the market's expectation for Hynix's gross margin in the second half of 2026), instead of boosting stock prices, they cause them to fall — that's when short-term investors must exit. This also ties into some of our discussions.
Alright, that's all for today's discussion. Thank you all for your support. Depending on my health recovery and major market developments, I will continue to post weekly summaries in the group. If something is particularly worth writing an article about, assuming I have time, I will summarize it in a better format and post it on the website.
Thank you again for your support. We can continue discussing the storage situation in the second half of the year within the group.