MU Bleeds 7% as Broadcom's AI Guidance Miss Drags the Memory Complex

The Broadcom Domino

This is contagion, not a Micron problem. Broadcom beat on headline Q2 numbers but guided Q3 AI chip revenue to $16 billion against a $17.2 billion estimate and pointedly declined to raise its 2026 AI semiconductor forecast. For a tape that has been pricing AI capex as a one-way escalator, a flat outlook reads as a top. AVGO fell 14% to $411 and dragged the complex with it — Micron down ~7%, SanDisk off 3%, Western Digital off 2%.

Micron has no direct read-through to Broadcom's custom-silicon order book; the link is pure sentiment. When the most-watched AI chip name signals demand might be plateauing, every memory and accelerator name gets repriced first and asked questions later. That's exactly the kind of correlated, narrative-driven move that hits hardest on the names that ran the furthest.

Nothing Actually Broke at Micron

The fundamental case is unchanged from a week ago. Last quarter Micron put up revenue up 57% year-over-year to $13.64 billion, with cloud memory revenue nearly doubling to $5.28 billion at 66% gross margins and HBM order books stretching into 2027. None of that reversed today. What changed is positioning: the stock is up roughly 865% over the past year, recently cleared a $1 trillion market cap, and a chart that vertical has no natural support to lean on when the bid steps away.

The one company-specific data point making rounds — CEO Sanjay Mehrotra selling about $38 million of stock near the highs — is noise, not signal. Those sales ran through a pre-arranged Rule 10b5-1 plan adopted back in January, so the timing is automated, not a discretionary call on the top. Reading it as an insider bailing out is exactly backwards.

What the June 24 Print Decides

The real catalyst is still ahead. Micron reports fiscal Q3 on June 24, and that print — not a Broadcom sympathy move — settles whether this is a flush or the start of something. The bull and bear cases are already drawn: Raymond James just lifted its target to $1,100 from $530 on AI-driven memory demand while reiterating Outperform, but the same shop's Karl Ackerman flagged that DRAM and NAND average selling prices likely peak in mid-2026 and roll over sequentially early next year.

That tension is the whole trade. Pricing power is the entire bull thesis on memory right now, and if the cycle peaks a year sooner than the Street modeled, a parabolic chart has a lot of air beneath it. Earnings either confirm demand and pricing are still accelerating, or they hand the bears the first piece of hard evidence that the cycle is topping. Until then, this is a sentiment tape.