Memory’s 30-Year Cycle: Will This AI Boom Also Burst?SK hynix is posting record profits, yet its share price has plunged 58% in just four months from the June peak. Samsung Electronics also reported a strong quarterly operating profit but saw its stock drop more than 20%. The natural question arises: “Is this another bubble?” Revisiting 30 years of memory semiconductor history. Memory has repeated the same boom-and-bust pattern five times: an 80% drop in DRAM prices in 1996, Hynix under creditor management in 2001, Qimonda’s bankruptcy in 2009, Elpida’s bankruptcy in 2012, and Hynix’s 7.7 trillion won operating loss in 2023. Not once has the industry broken this cycle.It Resembles Price-Driven Booms, Not the Dot-Com BubbleMany compare today’s AI surge to the dot-com bubble, but looking purely at memory, the parallel is different.
Hynix and Samsung’s profits exploded not from higher volume but from higher prices. DRAM contract prices have risen two- to threefold this year alone. Selling the same chip now brings in two or three times the revenue.This kind of pure price boom has already been seen in shipping, shipbuilding, polysilicon, and lithium. All four industries shared one key trait with memory: supply could not be increased quickly, so prices soared.
- Shipping (HMM): Freight rates drove massive profits in 2021–2022, but the stock price peaked a full year before earnings did. Hynix’s stock is behaving the same way—falling sharply while profits are still rising. In major price booms, the market often exits before profits actually break.
- Polysilicon: Spot prices jumped from about $40/kg in 2004 to $475 in March 2008, then collapsed to $6.75 by June 2020 as capacity expanded more than sixfold. Long-term fixed-price contracts signed at the peak became toxic, leading to lawsuits and bankruptcies.
Current memory long-term contracts (Hynix has locked in roughly 10 customers for typical five-year terms, often with advance payments) look similar on the surface. However, three critical differences exist:
① Prices are market-linked, not fixed.
② The buyers are financially strong Big Tech firms such as Nvidia and Microsoft.
③ China cannot easily catch up in advanced DRAM and HBM because of EUV equipment restrictions (roughly a three-year technology gap).
Thanks to these factors, the chance of profits plunging below zero into deep losses has decreased, and the floor has been lifted by the contracted volumes. Still, if market prices fall, profits will shrink. - Lithium and Shipbuilding: Sudden demand drops (lithium) or multi-year construction lead times for new capacity (shipyards) closely mirror memory. Building and ramping a memory fab also takes years. The supply is unlikely to catch up with demand before 2028.
The Real Risk May Come from the Money PipelineEven before supply catches up, demand could stall. A large portion of Big Tech’s AI data-center spending is funded by corporate bonds. A credit-rating downgrade (for example, Oracle) or a sharp rise in spreads on the lowest investment-grade bonds (the video uses a 1.9 percentage-point threshold) could constrain investment. The video also outlines three paths that could lead Big Tech to slow construction.How Many Years of Boom Are Already Priced In?At the time of the video, Samsung Electronics’ share price already appeared to discount roughly 2.8 years of boom conditions. Who holds the shares—foreign investors versus retail—and overall supply-demand dynamics will also help determine where the bottom forms.Conclusion and Signals to WatchThe video takes a balanced view: this cycle has important differences (long-term contracts, China’s technology gap, solid Big Tech demand) that have raised the floor, but the fundamental boom-and-bust pattern itself has not disappeared. The end could arrive if capacity expands sharply before 2028, if inventories build and lead times shorten, or if the corporate-bond market for Big Tech wobbles.If you remember only one thing: For 30 years, memory has risen and crashed without exception. This boom feels special, yet history has repeatedly followed a similar script. Keep a close eye on three signals—fab expansion plans for 2028 volumes, inventory and delivery-lead-time trends, and Big Tech corporate-bond spreads.
(This post s not investment advice. Investment decisions and their outcomes are solely the responsibility of the individual investor.)