Guessing the top is very dangerous, so I ran two days in advance. The impact of Mu last night was mainly due to a misinterpretation of an article, but when you ask why GPT fell, GPT will interpret it as poor performance guidance from AVGO, leading to a decline in semiconductors. The real market reason is: an article about NVDA, which is from an authoritative semiconductor research institution, claims that the Rubin NVL72 rack and DRAM memory market previously thought it was 192GB of SOCAMM, but Rubin used 96GB, which is equivalent to the expected 55TB of a machine now actually using 28TB. Then Mu started to decline. But the actual situation is that this article states that it will not be fully inserted, which is the default factory configuration. You just need to plug it in and buy more yourself. However, media outlets with lower intelligence have started to spread the rumor that "the memory is half over estimated", which has pushed the big picture and poured mud on it. In fact, this kind of operation is very common. When a stock continues to rise to its limit due to demand, the main funds will sell at a high point in advance, because the market will naturally shield unfavorable noise during the rise, known as the "three bullish lines changing views". However, after the shipment brings a decline at the high point, negative news that is very unfavorable and completely shakes the market's view begins to appear. After a period of decline, it enters a stable period, and then demand is recognized by the market again, taking off again, and repeating the cycle.