Memory Chip Stocks Drop 30% Despite Record AI Demand

Memory chip manufacturers are experiencing sharp stock declines even as artificial intelligence demand reaches historic levels, creating a paradox that reveals shifting priorities in AI infrastructure investment.

Micron Technology has fallen approximately 30% from its late-June peak, according to a report first published by Rich Duprey. SK hynix now trades below its July 10 IPO price, while SanDisk has dropped 35% after rallying roughly 600% earlier in the year. These declines come despite fundamentals that appear exceptionally strong on the surface.

Record revenues mask structural shifts

Micron reported record revenue in its latest earnings release, and management confirmed that high-bandwidth memory remains sold out well into future production quarters. SK hynix has similarly reported robust HBM demand driven by orders for Nvidia's latest AI accelerators. Technology companies are on track to spend more than $700 billion on AI infrastructure in 2026, based on company guidance and earnings releases.

The disconnect between operational performance and stock price suggests investors are pricing in risks that extend beyond current quarter results. One key concern centers on pricing power. A manufacturer expanding HBM capacity could theoretically ship 30% more chips yet generate lower revenue if average selling prices decline by 20%.

Budget reallocation reshapes the opportunity

Hyperscale cloud providers are redirecting portions of their AI budgets toward power infrastructure, liquid cooling systems, and custom chip development. This reallocation reduces memory's share of incremental AI investment even as total spending grows. The shift reflects operational constraints that have become more binding than compute capacity itself—data centers increasingly face power and cooling limitations before they exhaust processing capability.

Custom silicon development by major cloud providers also threatens to commoditize certain memory specifications over time. As hyperscalers design chips optimized for their specific workloads, they gain leverage in memory procurement negotiations and may reduce dependence on cutting-edge HBM for some applications.

Why it matters

The memory stock selloff signals that AI infrastructure investment is maturing beyond a simple race for more chips. As cloud providers confront power, cooling, and integration challenges, capital flows toward enabling infrastructure rather than processors alone. For memory manufacturers, this means competing for a smaller slice of each infrastructure dollar even as the total pie expands—a dynamic that can sustain strong revenues while compressing margins and growth expectations.

The details in this analysis were first reported by Rich Duprey for Yahoo Finance.