What This Is
Micron is one of the world's three major memory chip makers, primarily producing DRAM and HBM (high-bandwidth memory, a critical component for AI training). CEO Sanjay Mehrotra said publicly this week that memory supply will be significantly tighter in 2027 and 2028 than in 2026. The backdrop: AI companies are frantically building data centers and consuming all available memory capacity.
Industry View
Wall Street read this as bullish — Micron's stock jumped on the news, and analysts believe the memory cycle has entered a "seller's market," with pricing power back in the chipmakers' hands.
We see dissent too. Semiconductor research firm SemiAnalysis points out that AI inference (i.e., routine inference calls) is the real long-term memory black hole; training is only a short-term peak, and current capacity expansion may not keep pace. Gartner also cautions that the 2025-2026 investment surge could turn into overcapacity after 2027. We think there's another underappreciated risk: memory price hikes will squeeze consumer electronics — phones and PCs carry meaningful memory cost shares, and AI PCs and AI phones will get more expensive.
Impact on Regular People
For enterprise IT: hardware costs for AI deployment over the next two to three years will likely rise, putting pressure on cloud service pricing.
For individual professionals: marginal costs of using AI tools won't change in the short term (hyperscalers will absorb it), but budget approvals will get stricter.
For the consumer market: memory in phones and PCs will likely get pricier; AI PCs and AI phones will become more expensive.