Market Structure View on $SKHY The key point is simple: A 7x oversubscribed book proves that the primary market wants allocation. It does not prove that the secondary market has to keep bidding the stock higher. The real issue here is that the primary market and the secondary market are trading two completely different things. SK Hynix’s U.S. listing is huge. The basic facts are clear: 17.79 million new common shares. 177.9 million ADSs. 10 ADSs represent 1 common share. Demand reportedly exceeded available supply by more than 7x. Large U.S. institutional orders started around $200 million. Baillie Gifford, Coatue, Situational Awareness and other major funds have shown interest. That demand is real. But primary demand and secondary price action are not the same thing. The primary market is

If I had to summarize my first half of 2026 in one sentence, I would say: It was a first half filled with both pain and reward. The reward was that I saw the opportunity in the memory sector early. The pain was that I got the direction right, but I did not truly hold on. The best decision I made in the first half of the year was starting to build a position in the memory sector in January, mainly through Micron. My thinking at the time was simple: As HBM prices continued to rise, DRAM and NAND were also entering a new pricing upcycle. The memory industry was likely moving back into a strong cycle. Historically, memory has always been a classic cyclical industry. Every few years, it enters a powerful upcycle. But this time is different. This cycle is not only driven by a normal supply-deman

AI infrastructure is entering a new phase. In the first phase, the market was trading one question: Who has the most GPUs? Then the focus shifted to data centers. Who can secure enough power, land, cooling, servers, and deployment capacity? But now, the question is changing again. As AI Mega Clusters move from hundreds of thousands of GPUs toward millions of GPUs, the bottleneck is no longer just GPUs. It is no longer just power. The next bottleneck is whether multiple data centers can be connected into one unified AI training cluster. That is Scale-Across. SemiAnalysis recently published a deep dive on this topic. Their core view is clear: future AI clusters cannot rely forever on the expansion of a single campus. More cloud providers will have to connect multiple data centers, multiple c

Samsung’s profit surged 19x, yet memory stocks sold off hard. So what exactly is the market pricing in? Samsung delivered an almost flawless Q2 earnings guidance. Revenue is expected to reach 171 trillion KRW, while operating profit is expected to hit 89.4 trillion KRW, up nearly 19x year over year. This is not a normal cyclical recovery. This is a profit explosion driven by AI servers, HBM demand, and DRAM price increases. In theory, this kind of guidance should have ignited the entire memory sector. But the market’s reaction was the opposite: A sharp selloff. And that is exactly what makes today important. Memory stocks did not fall because Samsung’s earnings were weak. They did not fall because HBM demand suddenly disappeared. They fell because the market is shifting from pricing in ear

Bernstein just released what I think is one of the most important research notes on the memory industry this year. $纳指100ETF(QQQ)$ $闪迪(SNDK)$ Most headlines focused on one number: A Bull Case valuation of $4,400 for SanDisk. But in my opinion, that's not the real takeaway. The report spends far more time explaining why this memory cycle may be fundamentally different from every cycle before it. A year ago, when SanDisk was trading a fraction of today's valuation and most investors still viewed NAND as a deeply cyclical commodity business, Bernstein was one of the very first firms on Wall Street to publish a $1,000 price target. Back then, many thought it was far too aggressive. Looking back, they were simp

🧵 There's a new "stock guru" all over Twitter today hyping up $CBRS (Cerebras). It IPO'd barely over a month ago, billed as "the next NVDA" — and the stock has already been cut in half from its highs. So should we buy the dip? $Cerebras Systems(CBRS)$ 1️⃣ How strong are the fundamentals, really? Cerebras builds wafer-scale AI chips (WSE) — massive single-wafer chips that fuse compute and huge amounts of memory directly onto one die. Its specialty is high-speed, low-latency inference, which is the core architectural difference from NVDA's GPU approach. In January 2026, OpenAI signed a multi-year deal worth over $10 billion with Cerebras, committing to deploy 750MW of Cerebras systems, rolling out in phases through 2028. The latest big update: OpenA