Micron raised its U.S. investment to $250 billion through 2035, betting that AI demand for high bandwidth memory (HBM) will outlast the booms and busts that have defined the memory chip market since the 1990s.
Micron just put $250 billion on one claim: that AI infrastructure demand has broken the memory chip market out of the cycle that has humbled every incumbent since the 1990s. The scoreboard is high-bandwidth memory (HBM) share against SK Hynix, and whether Micron's New York fab pours silicon before the cycle turns.
The chipmaker raised its planned U.S. investment to more than $250 billion through 2035, up sharply from the roughly $100 billion to $150 billion framing of recent years. It also poured first concrete at its New York site in Clay, Onondaga County, a complex the company describes as the largest semiconductor manufacturing site in U.S. history. The market read it as material, not aspirational: Micron shares rose nearly 5% on the announcement.
The bet runs against forty years of memory-industry playbook. In every cycle since the 1990s, memory CEOs cut capex at the top and regretted it at the bottom. Micron is doing the opposite. It is pouring concrete into New York at what history would call the late innings and expanding DRAM and HBM capacity at a moment when prior cycles would have signaled retrenchment.
The mechanism is specific. HBM is the stacked memory that feeds NVIDIA, AMD, and custom AI accelerators; without enough HBM, an AI data center does not come online. Companies are continuing to spend aggressively on AI infrastructure, and HBM demand has been structurally decoupled from the PC and smartphone cycles that used to drive the memory market. Micron is targeting higher-volume HBM production alongside DRAM scaling, with a stated long-term goal of producing about 40% of its total DRAM domestically.
That 40% target is aspiration, not current state. It is the kind of corporate milestone that lives or dies on HBM yields and U.S. policy continuity. The $250 billion is a through-2035 envelope, not a single-year spend; how much of it actually gets drawn depends on CHIPS Act disbursement timing, hyperscaler order books, and whether memory pricing stays favorable through 2027 and 2028.
The competitive target is explicit. SK Hynix is the HBM market leader, with Samsung the established second pole, and Micron is the challenger trying to close a real share gap from a U.S. base. The New York complex will host up to four fabs focused on high-volume DRAM production, while Idaho and Virginia sites handle R&D and modernization of existing operations. The wire has anchored the announcement to CEO Sanjay Mehrotra across CNBC, Yahoo Finance, and Benzinga, with the GlobeNewswire press release as the canonical artifact.
The bear case is the one Micron cannot disprove in a press release. Memory cycles have humbled every incumbent in the industry's history; a capex envelope this large raises execution and pricing risk, and HBM share gains are not guaranteed. If AI demand softens before New York ramps, or if HBM yields slip, the through-2035 number will not be the headline investors remember.
The watch items are concrete: HBM share against SK Hynix quarter by quarter, the New York complex's first wafer-out date, and whether domestic DRAM output tracks toward the 40% target as CHIPS Act milestones unlock. The cycle turns every few years. Concrete cures in days.