In the U.S. equity market, cyclical stocks—especially memory names—have historically topped out 6–8 months before the actual earnings peak. In other words, share prices often start declining well before profits reach their high. Historical examples: - 2008 cycle: ~4 months early - 2012 cycle: ~4 months early - 2018 cycle: ~8 months early ChangXin Memory (CXMT)’s Friday earnings report may be a warning sign: revenue grew, but profits did not. Right now, revenue growth is slowing and gross-margin expansion is moderating. Upside to sell-side earnings estimates looks limited, and Street forecasts have become lagging indicators rather than leading ones. At previous cycle peaks, valuations always looked “cheap,” but cheap valuations alone have never been enough to prevent forced selling. Key risks include: supply catching up with demand, digestion of existing orders, incremental Chinese capacity, changes in AI architecture, HBM yield issues, and downstream hardware overbuild. Company management is typically optimistic at these points. Relying on management commentary or earnings calls to call the cycle inflection has historically been ineffective. This is not a call that memory fundamentals are broken. It is a call that the easy-money phase for memory stocks may be ending—the assumption that these names can still generate easy returns is what is being questioned. Stock prices are disconnected from earnings and fundamentals about half the time, because the optimistic expectations are already gone.