Micron 41B Quarter|CEO Sold the Peak

The 13% Selloff That Set Up the Blowout

Micron Technology entered its June 24 earnings report carrying its worst single-day drop since June 5. On Tuesday, shares fell 13.18% as semiconductor stocks globally collapsed on a report that SK Hynix was pulling back on AI memory expansion. The read was automatic: if the supplier is blinking, the cycle is topping. It wasn't. Micron reported Q3 revenue of $41.46 billion — against a $36 billion consensus and just $9.3 billion a year ago. GAAP gross margin reached 84.6%, up from 37.7% a year prior, with Q4 guided to approximately 86%. The bottleneck that explains this is not demand — demand is the story everyone knows. The bottleneck is supply, and supply cannot close the gap. CEO Sanjay Mehrotra stated on the earnings call that memory and storage demand significantly exceeds supply and will continue to do so beyond calendar 2027, driven by AI across all segments and structural constraints. Greenfield fab construction takes years and billions. Skilled worker shortages compound the constraint. Mehrotra said Micron does not have a line of sight as to when supply will catch up with demand. That is not a bullish talking point shaped for investors. It is a cost admission — the same structural constraints that lift pricing also cap how fast Micron can grow volume. The Tuesday selloff priced in a different story. The blowout answered it. But answering the bear's four objections is not the same as settling the question of what the stock is worth at a $1.164 trillion market cap. That is the question the reversal opened, not closed.

The CEO Who Sold at the Top Is Also the Man Who Just Called It a Decade

The central conflict in Micron's post-earnings setup is not between bulls and bears abstractly. It is between two concrete, attributable acts by the same person. In May, CEO Sanjay Mehrotra sold 37,435 shares at approximately $979.37 — near the stock's record high, after a 763% year-to-date run. On June 24, the same CEO signed off on Q4 guidance of $50 billion in revenue and $31 in earnings per share, figures that smashed street estimates of $43 billion. He disclosed that Micron has signed 16 long-term Strategic Customer Agreements, locking in three- to five-year supply commitments from hyperscalers, automakers, and AI infrastructure companies. HBM4, built on 1-beta DRAM technology, is in high-volume shipments for the lead customer. Qualification samples have been shipped to multiple end-customers, a direct signal that the HBM relationship is diversifying beyond a single-customer dependency. The insider selling, in isolation, is ambiguous — executives sell for tax and diversification reasons. But the combination is not ambiguous: broad executive selling at peak pricing, followed immediately by the most optimistic multi-year outlook the company has ever issued publicly. Portfolio manager Kyle Taylor at Tridelta Private Wealth made the split explicit: six months ago he loved the stock; today he tolerates it and has been selectively trimming to keep position sizing reasonable. Paul Meeks at Freedom Capital went further, warning that revenue growth is projected to slow to 76% in FY2027 and 7% in FY2028. Neither is short the stock. Both are adjusting exposure against a $1.164 trillion market cap that, as Meeks stated directly, prices in flawless execution. The question this creates for a holder is not whether the Q3 results were real — they were — but whether the CEO's personal capital move reflects information about execution risk at the 2027 inflection that the guidance language does not convey. That information gap is not resolvable from the earnings call alone.

Non-AI Units 10x in One Year — The Frame the Cyclical Bear Needs to Survive

The bear case heading into earnings had a structural fallback: even if HBM demand holds, AI-related strength is masking underlying weakness in consumer memory, mobile, and embedded markets. The Q3 print did not weaken that argument. It inverted it. Micron's Mobile and Client Business Unit generated $11.5 billion in Q3 revenue at an 87% gross margin and 86% operating margin. The Automotive and Embedded Business Unit posted a 79% gross margin and 75% operating margin. Both units saw operating profit more than 10 times the year-prior level. The significance runs deeper than the numbers themselves. A company where only the AI segment prints software-level margins can still be rationalized as a cyclical name with a temporary AI tailwind. A company where mobile, client, automotive, and embedded units simultaneously post margins of that magnitude has a different underlying economics — one where the memory-cycle discount model loses its structural justification. The mechanism is visible in the pool: spending on memory within the semiconductor market exceeded 50% of total semiconductor spend this year, up from 22% a year ago, per data cited by Tridelta's Taylor. That share expansion is not explained by HBM alone. It reflects AI-driven demand spreading across every major end market that touches computation — from smartphones processing on-device AI to robotaxis requiring high-memory autonomous driving stacks. Micron disclosed it shipped first 1-gamma DDR5 samples to a robotaxi customer — a line item buried in the product highlights that marks where the next wave of memory demand is beginning to form before automotive AI workloads scale. The remaining bear argument now requires demonstrating that this cross-segment margin expansion is coincidental and temporary, not structural. That is not impossible. But it is a harder case than the original "AI masking consumer weakness" thesis, which the Q3 numbers directly contradicted.

The $50 Billion Quarter Ahead and the One Variable That Decides the Leaning

Q4 guidance stands at $50 billion in revenue, $31 in earnings per share, and approximately 86% gross margin — against $11.3 billion in Q4 a year ago, a 342% increase year over year. High-volume multi-year Strategic Customer Agreements give near-term revenue visibility that removes most of the quarterly execution risk. The real question sits at the 2027 inflection. HBM4E, built on 1-gamma DRAM technology, has volume production expected in calendar 2027. HBM supply constraints are forecast to ease gradually starting in 2028. Revenue growth is projected to slow to 76% in FY2027 and 7% in FY2028, per analyst estimates in the pool. The bull thesis requires one thing to hold: that memory content per AI accelerator keeps rising even as AI server unit growth eventually moderates. The evidence for this trend is strong — Nvidia's H100 required roughly 80 gigabytes of HBM, the H200 raised that to 141 gigabytes, and Blackwell raised it again to approximately 192 gigabytes. If that trajectory continues through HBM4E and beyond, Micron does not need AI chip volume to stay parabolic. It needs each new GPU generation to require meaningfully more memory. The pool flags the genuine counter-risk: emerging Chinese competitors could narrow supply shortages, and Evercore's Daryanani noted that CEO Mehrotra's 2027 HBM order book commentary on the earnings call would move the stock more than the headline numbers. That commentary was bullish — but the same source noted the analyst who called Nvidia in 2010 excluded Micron from his current top 10 AI stocks. Both views exist in the same pool, grounded in separate analytical frameworks. For a holder: the monitoring variable is whether Q4's earnings call expands or narrows the 2027 HBM4E customer allocation visibility. Expansion confirms the structural re-rating. Flat or narrowing 2027 guidance gives the CEO's personal share sale the most informative read. For a watch-list candidate: Q4 delivery of $50 billion with an expanded 2027 allocation book is the entry signal. Until then, the $1.164 trillion market cap prices in a best-case execution path that Mehrotra himself did not hold at the peak.