According to CoinMeta, analyst mr.p stated in a podcast on September 26 that the claim that so-called hyperscale cloud providers have “10 years of demand visibility” is not credible. It is difficult for cloud providers to predict demand changes two years ahead. AI Infrastructure investments will ultimately be constrained by cyclical expenditures. In the short term, AI Inferred demand will continue to drive the growth of storage products such as HBM, DRAM, and NAND. It is estimated that next year, hyperscale cloud providers' capital expenditures will reach $1.1 to $1.2 trillion, of which storage expenditures may account for 50% to 60%, or approximately $500 billion to $700 billion. UBS even predicts that this figure could reach $900 billion. mr.p believes that the current AI Computing power bottleneck has gradually shifted from merely GPU Quantity to aspects such as electricity, advanced packaging, storage, and ABF Circuit boards.