BlockBeats News, July 18th, Citrini analyst Jukan posted an article stating that a recent market analysis believes that the recent pullback in storage chip stocks is not only influenced by deleveraging factors, but the market is also beginning to price in the 2028 supply expansion anticipation in advance. Although the industry generally expects the tight supply-demand situation of high-end storage such as HBM to continue until 2027, research institutions and sell-side analysts generally believe that with large-scale production expansion by companies like Samsung Electronics and SK Hynix, the supply-demand gap is expected to start easing in 2028. The traditional experience suggests that storage stocks usually peak about two quarters ahead of storage prices, but this analysis proposes that the market may not necessarily be limited to this rule and may reflect the expectation of future supply increase earlier.

However, all pessimistic expectations are based on the same assumption—that the new capacity will be released in a concentrated manner in 2028, leading to another significant drop in storage prices. Looking back at history, since the 1980s, almost all sharp price declines in the storage industry have originated from rapid supply expansion rather than demand downturn. The demand in the PC, smartphone, and cloud computing era has actually always been growing. This AI cycle may be fundamentally different from the past: AI applications have a higher price elasticity of demand for computing power and storage. When prices drop, the growth rate of demand may exceed the impact of the price decline, weakening the blow of the traditional storage cycle.

The analysis cites a paper titled "The Economics of Digital Intelligence Capital" by Zhang and Zhang in 2026, which states that the demand price elasticity of AI Tokens is about 1.42, meaning that for every 1% price drop, the demand volume can increase by about 1.42%. The paper argues that when the API price drops significantly, developers will not only increase the call volume but also adopt more computational-intensive inference architecture, causing Token consumption to exhibit convex growth. Taking DRAM as an example, if the selling price drops by 30%, under a traditional cycle, revenue and profit often shrink significantly, with Samsung Electronics seeing a 52.8% year-on-year decline in operating profit in 2019. However, in the AI-driven new demand environment, if sales volume increases by around 42% and process upgrades drive a cost reduction of about 15%, industry revenue may remain relatively stable, and the profit decline is expected to narrow to around 15%. The analysis suggests that this difference may determine whether storage manufacturers should still be valued based on a traditional cycle P/E ratio of 5 to 6 times in the future.