Since the last time I posted, I haven't posted the middle article yet On the one hand, I am too busy to have a whole block of time On the one hand, I also want the market to verify it Yesterday's major adjustment at least partially verified the correctness of the direction in the article: My core logic is divided into two parts Firstly, the main driving force for the growth of AI upstream and downstream comes from: -Directional injection of US dollar liquidity -The initiator of CAPEX in the upstream of AI supports valuation expectations through mutual aid groups -Transmitting liquidity and valuation downstream through CAPEX and sustained CAPEX expectations This is essentially credit creation, which is similar to the logic of obtaining a high proportion of mortgage loans from banks despite consistently low transaction prices and high prices in a certain region Secondly, the "bottleneck narrative cluster" downstream of AI in the secondary market does not follow the downward flow logic of the supply chain, but competes with each other for the same liquidity pool 2. Recent market trends have been verified 1. The Impact of CAPEX Upstream Expectations on Downstream Market Liquidity Three major events in the upstream recently: -META announces rental of redundant computing power -Apple and MU have a supply price dispute -Hynix's long-term supply contract changed to no price cap The latter two are direct contradictions of the CAPEX mutual aid group, and the dispute reflects that the initiator of CAPEX (such as Apple) is already considering the issue of "expensive". When you have the idea of considering "expensive", "urgent" will no longer have a premium This is a signal of a logical reassessment of the bottleneck in production capacity Renting redundant computing power from META does not mean that the market is not lacking in computing power, nor does it mean that it will reduce or stop CAPEX (recently announced the development of a new AI chip with Hynix). But the expectation of market changes revolves around the sustained impact of META's own demand changes on CAPEX expectations, and even the addition of some META businesses downstream of CAPEX. After all, META has a rich history in the metaverse, and he said that investing and actually getting the money in place are completely different things These are completely sufficient to change the expectation of injecting downward liquidity into the upstream of AI at least in the short term 2. Downstream "AI bottleneck narrative liquidity" is first stripped off with premium The recent decline is not due to the complete collapse of all AI related stocks, but rather to the fact that the impact of upstream mutual aid stocks is not significant, while downstream stocks have the most severe collapse, including: -Neocloud that relies 100% on CAPEX: NEBIUS, COREWEAVE -Downstream semiconductors, the more downstream they are, the more obvious the "bottleneck" collapse becomes, such as MRVL, GLW, etc ”The idea of "splitting the external liquidity relay to the top → downstream reassessment" is clearly correct Next, we need to look at the financial reports of major upstream companies in July to confirm whether there is a "halt in US dollar Ponzi input → capex stagnation → downward revision of forward revenue → semiconductor full cycle stocks"“ Also, for those who have experienced the market for 21 years, do you think yesterday and 519 could be so similar?