Since bottoming in April 2025, Micron Technology (NASDAQ: MU) has staged one of the semiconductor sector’s most spectacular rallies: a 15-fold gain in just 17 months. But after peaking in June, the stock has given back roughly 22% of its value. With management scheduled to report earnings on Sept. 30, the pullback is now at the center of an intense debate on Wall Street — is this a healthy dip to buy within an AI-driven memory supercycle, or an early warning that the cycle is about to turn?
The engine behind Micron’s surge is high-bandwidth memory (HBM), the 3D-stacked DRAM chip purpose-built for the data throughput demands of large AI models and high-performance computing. HBM has become a critical complement to AI accelerators, and the market is an oligopoly: Micron, Samsung and SK Hynix are the only three manufacturers shipping the product at scale, leaving supply chronically tight.
That imbalance has translated into explosive financial results. For the first three quarters of fiscal 2026, Micron posted $79 billion in revenue, up 203% year over year, while net income soared to $47 billion from just $5.3 billion a year earlier. Analysts expect full-year fiscal 2026 revenue growth to reach 247%, moderating to 88% in fiscal 2027. The results have pushed Micron’s market capitalization to roughly $1.1 trillion, gross margin to 72.6%, and forward price-to-earnings ratio to around 6x — a valuation many bulls argue is too low for the growth on offer.
The supply picture points to a long runway. HBM consumes roughly three times the wafer capacity of conventional DRAM, and output is constrained by EUV lithography tools also needed for advanced logic chips. SK Hynix has projected that the DRAM supply-demand imbalance will peak in 2027, with the market not rebalancing until as early as 2030. Major memory makers have begun signing long-term supply agreements — a first for the industry — reinforcing the view that the AI-led upcycle has structural underpinnings.
Still, memory remains a notoriously commoditized, cyclical business, swinging between shortage-driven price spikes and glut-driven collapses. Today’s record margins and profits sit near what has historically been the top of the cycle. Should AI infrastructure demand show even marginal signs of slowing, the supply-demand balance could flip quickly, dragging memory prices and Micron’s earnings down with it.
History offers a warning. Micron’s shares have repeatedly lost more than 50% of their value in past downturns, including two declines exceeding 80%. The stock’s seemingly cheap valuation can also be misleading: if net income merely fell back to fiscal 2025 levels, its trailing P/E would stretch into triple digits, forcing a painful re-rating.
Seasonality adds another headwind. September is historically the weakest month for U.S. equities, and the effect tends to amplify in midterm-election years. According to Cantor Fitzgerald analysis, the S&P 500 has fallen 5% or more during the September–October period in 15 of the past 24 midterm cycles — a backdrop that typically pressures high-multiple growth names.
Heading into the print, the setup cuts both ways. The HBM shortage is intact and the supercycle is not over, so the quarter is likely to show continued strength and the stock could keep working as a tactical trade. But at these heights, negative news is far more likely to trigger a selloff than in recent quarters, making the name less suited to buy-and-hold investors.
For now, the 22% pullback is neither a signal that the trade is over nor a guaranteed buying opportunity. The Sept. 30 report — and, crucially, management’s guidance — will be the moment that tests whether the AI memory boom has further to run, and whether the cycle is starting to turn.
