The DRAM storage sector has suffered severe emotional setbacks in the past few days, with storage stocks including Micron, Hynix, and SanDisk experiencing significant pullbacks. There is actually no need to panic, for two reasons: 1) In addition to the collective pressure on the US stock market caused by the expectation of rising inflation and interest rate hikes after the non farm payroll data, storage giants such as Micron and Hynix have clearly seized the opportunity to aggressively suppress their valuations. On the surface, the trigger was a report from well-known semiconductor research firm SemiAnalysis, which pointed out that Nvidia's Rubin shelf CPAmericaOCAMM DRAM capacity was reduced from the planned about 55TB to about 28TB. The market interpretation is that upstream Nvidia's procurement demand will weaken, so as a supply chain service provider, the company's valuation will naturally decline. However, in reality, this is only due to the severe shortage of PDDR5X globally. Nvidia has adopted an active downgrade and multi shipment strategy to accelerate the large-scale deployment of Rubin, and the core storage module HBM4 has not been moved. Therefore, it does not affect everyone's logical judgment on the future strong demand for the storage sector; 2) The fundamental reason is that the short-term storage sector has been excessively FOMO, because although the strong demand for storage in the AI super cycle has not changed, the current market valuation of storage giants relies entirely on Nvidia's valuation and its corresponding purchase order demand to support it. Once Nvidia has a "adjustment" plan on the purchasing side, it will be instantly interpreted as a peak in demand and trigger a stampede style pullback; But to be honest, relying solely on upstream purchase orders and storage modules to link the valuation expectations of the target in the AI super cycle is inherently irrational. Storage itself has a strong hardware replacement cycle attribute. In the short term, the valuation expectations amplified by order expectations will naturally brake and stabilize rationally after reaching a certain level. Rather than saying that the decline in storage this time is due to procurement demand, it is more reasonable and necessary to say that the irrational valuation model supported by a single order in the past has been distorted. In the long run, storage remains the hardest seller of AI infrastructure, and short-term emotional setbacks do not affect long-term fundamentals.