Nvidia’s warning that memory costs have become increasingly elevated adds another dimension to the global artificial-intelligence investment cycle, where demand for advanced computing hardware is putting pressure on multiple parts of the semiconductor supply chain. If memory availability remains constrained through fiscal 2028, as Nvidia has indicated, investors may need to reassess how long the current pricing environment can support earnings across the memory industry.

Nvidia Signals a Longer Period of Memory Supply Constraints

Memory has become an increasingly important component of AI infrastructure as data centers deploy large numbers of GPUs and accelerators. High-bandwidth memory, or HBM, is particularly critical because it allows AI processors to access large amounts of data at high speeds, while conventional DRAM and other memory products remain essential throughout data-center systems.

Nvidia’s assessment that memory pricing has reached “extreme” levels suggests that demand is continuing to outpace available supply. Semiconductor manufacturers have been expanding capacity, but advanced memory production requires significant capital investment, specialized manufacturing processes and lengthy qualification periods. As a result, supply cannot necessarily respond quickly to sudden increases in AI-related demand.

A constraint that persists into FY2028 would be significant because it would extend the period during which memory manufacturers could potentially maintain stronger pricing power. However, actual pricing and profitability will continue to depend on production increases, customer inventories and the pace of AI infrastructure investment.

Why MU and SK Hynix Could Benefit From Elevated Memory Pricing

Micron and SK Hynix are among the companies most directly exposed to the expanding demand for advanced memory. Both have significant positions in HBM and other memory technologies used in AI servers, making their financial performance increasingly connected to the capital-spending plans of major cloud providers and AI developers.

When memory prices rise while production costs remain relatively controlled, manufacturers can experience substantial improvements in revenue and operating leverage. That dynamic can create sharp swings in semiconductor earnings because memory is historically a cyclical industry in which supply-demand imbalances have a significant effect on profitability.

The argument surrounding valuations therefore centers on whether current earnings multiples adequately account for a potentially extended period of elevated demand. If supply remains constrained for longer than previously expected, earnings estimates could remain supported or potentially move higher. Conversely, rapid capacity expansion or weaker AI investment could cause pricing conditions to normalize more quickly.

AI Demand Changes the Memory Cycle

The current memory cycle differs from traditional semiconductor upturns because AI infrastructure is generating demand for increasingly specialized products. HBM, in particular, has become strategically important to the performance of advanced AI accelerators, creating an additional bottleneck alongside GPU and networking capacity.

For investors, the broader issue is whether AI-related memory demand can remain strong enough to offset the cyclical risks historically associated with the industry. Nvidia’s warning provides evidence that memory supply remains an important constraint, but it does not guarantee that elevated pricing or current profitability levels will persist.

Going forward, investors will monitor HBM production capacity, memory contract pricing, capital expenditure plans, customer inventories and Nvidia’s AI infrastructure demand. The central question for MU and SK Hynix will be whether constrained supply lasts long enough to support earnings at levels that justify a higher valuation, or whether the market begins anticipating a normalization of memory prices before FY2028.