Micron Plunges 13% in a Single Day — A Crash Triggered by One Report?
Claire Weston
SemiAnalysis reported that Nvidia cut next-gen server memory capacity in half to 28TB; the market read it as fading AI demand and Micron lost over 20% in two days — but the memory being cut and the memory driving AI growth are two entirely different products.
What exactly got cut?
Nvidia reduced the modular memory (SOCAMM DRAM) in its next-gen Vera Rubin server rack from 55TB to 28TB — nearly halved.
This means → the cut hit commodity server memory, not high-bandwidth memory (HBM) — the component that actually powers AI compute.
In plain terms = think of a car's trunk versus its engine. Nvidia shrank the trunk, not the engine.
Why did the market react so violently?
Micron closed down 13% at $864.01, its steepest single-day drop since April 2025; combined with a 7.7% slide the day before, the two-day loss topped 20%.
Micron's stock had rallied more than 735% over the prior 12 months — at that valuation, any bearish signal triggers aggressive profit-taking.
This reflects a market that did not distinguish *which* memory was cut — it equated "less memory capacity" with "AI demand is cooling."
Why didn't the HBM certification save the stock?
On the same day, Nvidia CEO Jensen Huang announced that Micron — alongside SK Hynix and Samsung — had passed HBM4 certification, qualifying as a supplier of the latest-generation high-bandwidth memory.
That should have been a major positive. Instead, it was drowned out by the panic selling the report triggered.
This reflects a pattern: when fear dominates trading, bad news travels far faster than good news.
What does the "crowding-out effect" tell us?
Barron's reported that the SOCAMM cut may actually stem from HBM crowding out other memory types — each unit of HBM consumes roughly three times the wafer capacity of standard memory.
This means → the more HBM production scales up, the less capacity is left for SOCAMM and other conventional DRAM. Cutting commodity memory is, paradoxically, a sign that HBM demand is accelerating.
Trendforce analysts noted that as HBM technology continues advancing through 2027, the squeeze on traditional DRAM will intensify — strengthening suppliers' pricing power in HBM negotiations.
Why did the firm's clarification fall flat?
SemiAnalysis founder Dylan Patel posted on X that the report contained "nothing bearish whatsoever" and that critics simply hadn't read the full analysis.
The clarification failed to reverse the sell-off and instead sparked a new controversy: user Vinit Baliyan asked why the full report wasn't shared with the broader subscriber base.
In plain terms = the core accusation is this: institutional clients saw the full report and traded on it first; retail investors saw only the headline and got caught in the downdraft — the information asymmetry itself became the story.
Content is for reference only, not financial advice.