Micron Technology (MU) briefly crossed a $1 trillion market capitalization for the first time on May 26, after shares spiked as much as 19% intraday and closed up about 17%. The milestone was driven less by a product launch than by a single analyst note and a memory market where supply is effectively gone through year-end.
The catalyst: a 204% price-target hike
UBS analyst Timothy Arcuri raised his Micron price target to $1,625 from $535 — roughly tripling it and making it the most bullish of the 46 brokerages covering the stock. The thesis is structural, not cyclical: Arcuri argues that Micron's shift to multi-year, fixed-price supply agreements with hyperscale AI customers has broken the boom-bust pattern that historically kept memory names trading at low multiples. Despite an eightfold run over the past 12 months, Micron still changes hands at roughly 8x forward earnings versus about 22x for the S&P 500.
"The need for pure memory has increased rapidly over very short periods of time, and clearly, Micron sits at the center of it," said Art Hogan, chief market strategist at B. Riley Wealth.
HBM sold out, pricing power intact
The engine underneath the rally is high-bandwidth memory. Micron's entire 2026 HBM supply is sold out, and CEO Sanjay Mehrotra has said HBM4 capacity is locked under long-term, fixed-price contracts — with the company reportedly filling only 50-65% of key customers' medium-term demand. Volume production of HBM4 for Nvidia's Vera Rubin platform began in Micron's fiscal first quarter.
The shortage is bleeding into commodity memory too: contract DRAM pricing is reportedly up 58-63% and NAND flash 70-75%, squeezing every downstream buyer of server, networking, and edge memory. Micron's cloud-memory segment grew roughly 160% year over year to about $7.7 billion last quarter.
What it means for builders
The read-through for infrastructure teams is blunt: memory, not just GPUs, is now a binding constraint on AI capacity. When the dominant supplier is allocating output by contract and filling barely half of demand, spot availability for anyone without a multi-year agreement gets thin and expensive. That reshapes capacity planning — HBM and DRAM allocation become procurement problems on par with accelerator supply, and the pricing increases flow straight into the cost of standing up new clusters.
Micron joins Samsung, which already crossed $1 trillion, in the memory chapter of the trillion-dollar club; SK Hynix is approaching the same threshold. For enterprises budgeting 2026-2027 AI buildouts, the practical takeaway is to treat memory as a first-class line item with its own lead times and locked-in pricing, rather than a bundled afterthought to GPU spend.
