NEW YORK, March 26, 2026, 10:05 EDT
Micron Technology fell roughly 4.2% to $366.23 early Thursday in New York, trading below its previous close of $382.09 from Wednesday.
This shift packs a punch. Micron holds a crucial spot in the AI memory supply chain, right up there with Samsung Electronics and SK Hynix—the trio controlling most of the world’s memory output. All three have been benefiting from the tight supply caused by AI data-center buildouts. High-bandwidth memory (HBM), stacked right alongside those AI processors, is the hot commodity.
Every notch lower in memory appetite counts. This week, Google Research claimed TurboQuant slashes “key-value cache” memory—used by large language models to remember prior tokens—by a factor of six or more, all without dinging accuracy. On H100 hardware, some computations reportedly jump as much as eightfold. Google Research
Micron’s timing isn’t ideal here, landing just a week after it posted a blowout quarter. The company projected third-quarter revenue at $33.5 billion—well ahead of what Wall Street was expecting. Still, shares dropped after Micron revealed it plans to lift 2026 capital spending by $5 billion, bringing the total to over $25 billion. The bigger budget “makes sense, given the shape of the demand,” Ben Bajarin at Creative Strategies told Reuters. Reuters
Micron’s tone remains upbeat. CEO Sanjay Mehrotra, writing on the company’s investor page, pointed to record highs in revenue, gross margin, EPS, and free cash flow. He described memory as “a strategic asset” for AI-focused customers. Micron Technology
Supply remains tight, at least for now. In comments prepared last week, Micron described NAND demand as “significantly in excess” of available supply. The company also projected that supply-demand dynamics for DRAM—the core memory in phones and servers—and for NAND, which handles data storage, are likely to remain constrained past 2026.
Competitors are betting big that the crunch isn’t going away. SK Hynix, for one, revealed plans this week for a confidential U.S. listing, eyeing up to $14 billion. Just the day before, it announced a nearly $8 billion order with ASML for extreme ultraviolet equipment—the gear behind the sharpest chip designs—to ramp up its advanced memory production.
Micron is moving forward, too. According to Reuters on March 16, the company’s prepping a second plant at the Tongluo location in Taiwan. Meanwhile, Micron also announced the launch of high-volume HBM4 production, targeting Nvidia’s Vera Rubin platform.
The bear scenario doesn’t ask much. Should tools like TurboQuant catch on more quickly and chipmakers push capacity too far, the pricing tailwind that’s bolstered memory shares could lose steam earlier than many expect. Lately, analysts have zeroed in on that combination—less demand per device, looming surplus.
Plenty of Wall Street voices aren’t on board with the bearish view. Morgan Stanley’s Joseph Moore called the drop “a healthy pricing in of durability concerns,” but pushed back, saying memory still sets the pace for AI demand, not the opposite. tipranks.com
