Wall Street Stacks Micron Targets to $2,000 as It Bets $250B on U.S. Fabs

The Target Avalanche

Micron's 24-hour rip to $978.90 — up more than 11% — wasn't one headline, it was the whole sell side moving at once. KeyBanc sits at $1,750 and Cantor Fitzgerald at $2,000, with Daiwa at $1,700 and Citi keeping the stock on an upside-catalyst watch. The loudest note came from BofA's Vivek Arya, who lifted his target to $1,550 on a genuinely contrarian read: cheap, open-weight Chinese AI models don't kill memory demand, they multiply it, because every enterprise that self-hosts open weights still has to load its own HBM.

The supply side backs the case. Micron's HBM is effectively sold out through 2026, with purchase orders extending into 2028 and BofA modeling a $246 billion HBM market by 2030. When capacity is pre-sold two years out, the target hikes read less like a call and more like analysts marking to a demand curve that has already contracted.

The $250 Billion Onshore Bet

Underneath the ratings noise is a structural story the market is only now pricing. Micron is lifting planned U.S. fab and technology spend above $250 billion through 2035, with a stated goal of building roughly 40% of its DRAM onshore across New York, Idaho, and Virginia. That sits on top of a broader U.S. chipmaking push that includes around $3 billion into the domestic semiconductor supply chain, $500 million of financing, and a 10-year wafer supply deal with GlobalWafers to lock in critical materials.

This isn't a story built on a single quarter. Micron already posted record third-quarter results in late June, and the onshoring commitment is what turns a cyclical memory beat into a multi-year capacity narrative — the kind of framing that supports $1,500-plus targets rather than a trade around the next print.

The Counter-Tell

The one voice cutting against the euphoria is the person with the most to gain from it. SK Hynix chairman Chey Tae-won called AI-driven memory prices 'abnormally high,' a rare admission from a supplier that current pricing may not hold. When the people selling the product flag the price, it carries more weight than any bear note.

The tape still argues for momentum over caution — memory demand is modeled to climb 50-60% next year and AI-specific demand as much as 100% — but the setup has air pockets. At roughly 19.6x earnings the stock isn't cheap, it remains about 15% below its June high near $1,150, and there's no fresh company data until the next report on September 22. This is a repricing on analyst conviction and macro flow, not new numbers — which cuts both ways if the memory cycle cools faster than the targets assume.