✦ The $200 Billion American Build-OutNarrow moat

The longest-dated bet is poured in concrete. In June 2025 Micron expanded its American investment plan to roughly $200 billion — about $150 billion of fabs and $50 billion of R&D — spanning two fabs in Idaho, up to four in New York and an expansion in Virginia, with a stated aim of making about 40% of its DRAM in the United States¹. The first checkpoint is dated: ID1 in Boise is to produce its first DRAM wafers in the second half of 2027, with the New York megafabs following around the decade's turn and HBM advanced packaging brought onshore as well². Washington's contribution — a $6.165 billion CHIPS award³ — covers barely three percent of the plan; this is overwhelmingly Micron's own supercycle cash being converted into American capacity.

The strategic logic is that geography has quietly become part of the product. Micron is the only American DRAM maker, its two rivals are Korean, and the buyers funding the AI build-out — hyperscalers under sovereign-supply-chain pressure, a government treating memory as strategic — increasingly price provenance. Forty percent American DRAM would make that distinction structural rather than rhetorical.

The bet's risk wears a date too. Memory troughs arrive roughly every three to four years, and ID1's ramp lands in late 2027 — squarely where the next downturn would fall if history rhymes — with American operating costs meeting trough pricing. Watch whether ID1's first-wafer date holds, and what the demand picture looks like as New York's capacity decisions firm up. A fab that ramps into a boom is a moat; one that ramps into a glut is a very expensive reminder of which industry this still is.