Memory executives expect RAM shortage to continue through 2028

Micron CEO Sanjay Mehrotra expects demand for Micron's memory products to exceed available supply for at least the next couple of years, while Samsung executives also expect the memory shortage to persist. Capacity is being prioritized for AI-oriented high-bandwidth memory and server DRAM, supporting suppliers' pricing and demand outlooks but constraining memory availability for consumer devices.

Analysis

The investable implication is not simply higher MU pricing: prolonged allocation of leading-edge wafer capacity toward HBM raises the opportunity cost of commodity DRAM/NAND production, creating a broader memory upcycle with unusually disciplined supply. MU has the most direct operating leverage, but SK Hynix (000660 KS) likely retains the strongest near-term HBM mix advantage; Samsung (005930 KS) is the swing supplier whose qualification progress at major AI customers could cap pricing upside. The key variable is HBM yield and customer qualification, not aggregate AI-server demand alone.

Over the next 1-3 months, memory contract-price checks and hyperscaler capex guidance are the relevant catalysts. If server DRAM tightness spreads into DDR5 used in enterprise systems, MU’s gross-margin trajectory can exceed consensus even if unit-bit growth moderates; conversely, a rapid Samsung HBM supply ramp would shift the market from scarcity premium to share competition before end-demand weakens. Consumer-device OEMs with low inventory and fixed launch calendars—particularly PC and handset assemblers—face a margin squeeze or bill-of-material pass-through, favoring component suppliers over hardware brands.

Consensus may be underestimating the duration of supply discipline because adding cleanroom capacity does not immediately create qualified HBM output; packaging, yield learning, and customer validation are binding constraints. The contrarian risk is that memory equities already discount a multi-year shortage: a single large customer redesign, AI capex pause, or evidence of HBM inventory building could compress MU’s cycle premium sharply. Falsify the bullish thesis on sequential contract-price deceleration combined with MU guiding gross margin below consensus, rather than on isolated spot-price volatility.

Key Decisions for Investors

• Maintain/add long MU on pullbacks ahead of the next earnings and contract-price cycle; target a 6-12 month holding period, with upside driven by gross-margin revisions rather than volume. Reduce if management signals HBM allocation growth is being offset by falling conventional DRAM pricing or if gross-margin guidance misses consensus materially.
• Express relative value through long MU / short a consumer-hardware basket such as HPQ and DELL for 1-3 months, sized modestly: memory inflation is a direct revenue/margin tailwind for MU but an input-cost headwind for OEMs with competitive pricing. Exit if PC/server OEMs demonstrate broad successful price pass-through or DRAM contract prices flatten.
• Watch Samsung Electronics (005930 KS) HBM qualification and yield disclosures as the principal supply-side hedge to MU. Do not add aggressively to MU if Samsung gains material leading-AI-accelerator qualification faster than expected; that would likely narrow HBM pricing and valuation premiums within 1-2 quarters.
• Monitor enterprise-server ODM and hyperscaler capex commentary for evidence that AI-system demand is converting into memory orders rather than inventory accumulation. If lead times decline while reported HBM inventory rises, treat it as an alert to take profits rather than a reason to buy the dip.