Needham Tripled Its Micron Target to $1,550. The Whole Street Is Catching Up.

The Re-Rating, Not the Earnings

Micron reported after the June 24 close and the stock has been grinding higher ever since. This particular leg — up 5.75% over six hours to $1,218 — is not a fresh catalyst. It's the sell-side racing to reprice a name that already gapped roughly 15% on the print. Needham took its target from $500 to $1,550, Bank of America reset to $1,500, and Stifel ($1,500), Wedbush ($1,300), Bernstein ($1,300) and Rosenblatt ($1,200) all stacked in between. The day before earnings, plenty of street targets still carried a $500-handle. A day later, those numbers are stale, and the stock is doing the work of closing the gap to where analysts now say it belongs.

Why the Multiple Still Looks Cheap

The entire bull case fits in one number: even at $1,218, MU trades under 12x forward earnings. Needham's $1,550 isn't a heroic multiple — it's 10x its FY2028 non-GAAP EPS estimate of $155. What actually moved is the earnings base underneath the multiple. Q3 printed record revenue of about $41.5B and a gross margin near 84.6%, Q4 is guided to roughly $50B, and HBM3E/HBM4 capacity is booked through calendar 2027 with demand stretching into 2028. When supply is sold out and a chunk of pricing is contracted, the cyclical de-rating that normally caps memory stocks doesn't show up on schedule. That's how a name up several hundred percent YTD can still look cheap on forward numbers.

The Other Side

The skeptic's read is that most of these targets are reactive, not predictive. Raymond James doubled from $530 to $1,100, Needham more than tripled off a $500 base, and several others moved in a single day — that's the street chasing a print, not getting ahead of it, and it often marks local enthusiasm rather than the start of a new leg. Earnings gaps in semis are also prone to mean-reversion in the first 24-48 hours, and Seeking Alpha has already flagged emerging bearish momentum signals even as the fundamental story stays intact. A stock that has run this far leaves little margin for error: one softer NAND datapoint or a single cautious data-center read, and the same forward multiple that screens cheap gets re-examined in a hurry.