Short memory? The most dangerous words in investing are back

This week, Micron Technology achieved a remarkable milestone, surpassing a market valuation of $1 trillion. This achievement would have seemed outlandish to many investors just a few years back. The memory chip sector has a notorious reputation for extreme volatility, often characterized by rapid growth followed by severe downturns. In 2017, Bill Joy, co-founder of Sun Microsystems, posited that the DRAM market was undergoing a fundamental transformation. He suggested that a decreasing number of suppliers, coupled with increasing demand from sectors like cloud computing and artificial intelligence, would mitigate the industry's propensity for self-destruction through overproduction. Fast forward to today, and although Joy's predictions were ahead of their time, they were not incorrect; Micron's stock has surged nearly twentyfold since then. Samsung also recently crossed the $1 trillion mark in market capitalization, with SK Hynix following suit shortly thereafter. The financial metrics driving these valuations are compelling; for instance, Samsung reported over $30 billion in profits in the first quarter alone. This has led investors to ponder a critical question: Is this market cycle genuinely different from previous ones? One significant factor contributing to this potential shift is industry consolidation. In the early 1990s, the global DRAM market boasted over 20 key players. Today, it is predominately dominated by just three companies: Samsung, SK Hynix, and Micron. Historically, memory manufacturers would respond to increased demand by saturating the market with new supply, leading to price crashes and profit losses. However, with fewer competitors, there may be reduced motivation to repeat this pattern. Another pivotal change is the burgeoning demand from AI technologies. Modern AI applications require vast amounts of memory as they continually transfer and process large data sets within expansive data centers. Even the most advanced methods aimed at alleviating computing bottlenecks still face limitations due to memory constraints. For example, startup Lightmatter is innovating with photonics—utilizing light instead of copper—to enhance AI data center efficiency, yet CEO Nick Harris acknowledges that this does not resolve the memory bottleneck issue. This intersection of soaring AI demand and limited supply is noteworthy. Analysts from UBS have highlighted that memory manufacturers are entering into multi-year contracts with major cloud providers, which secure both volume commitments and partially fixed pricing. This strategy is expected to introduce an additional layer of market stability, with UBS projecting that these agreements could keep the DRAM market undersupplied until 2028. However, caution is still warranted. There remains the possibility of another significant market downturn, driven by potential market flooding from remaining players, new entrants disrupting the status quo, a cooling in AI demand, or advancements in technology that lessen memory requirements. Nevertheless, investor sentiment suggests that the memory chip market may be on the verge of a transformative evolution.