What's happening

Semiconductor equities experienced a broad and steep decline on June 23, 2026, as news that SK Hynix plans to slow its expansion of high-bandwidth memory (HBM) production rippled through global markets and triggered a sector-wide sell-off. The PHLX Semiconductor Index fell approximately 8% on the session, reflecting the breadth of the move across chip-related names. Micron Technology (MU), a direct competitor in the HBM market, saw the steepest losses among major U.S. names, with shares falling between 10.96% and 13%. Nvidia (NVDA) declined 3% to 3.45%, AMD dropped 5% to 5.84%, Intel (INTC) fell 4.17% to nearly 5%, and Taiwan Semiconductor Manufacturing Company (TSM) also traded lower. The sell-off in Asian memory makers Samsung Electronics and SK Hynix preceded and appeared to catalyze the weakness in U.S.-listed semiconductor names, according to reporting from CNBC and Reuters. The June 23 decline followed a prior session in which broader technology stocks, including members of the Magnificent 7, had already shown weakness, suggesting the move was part of a multi-day rotation away from AI-exposed equities.

Why it matters for markets

High-bandwidth memory has become a critical component in the AI chip supply chain, serving as the primary memory architecture paired with leading AI accelerators including Nvidia's data center GPUs. SK Hynix's reported decision to moderate HBM production expansion raises questions about the pace at which memory supply can scale to meet projected AI infrastructure demand — and, conversely, whether that demand trajectory itself may be moderating. Micron, which has positioned HBM as a central growth driver and reported revenue of $58.12 billion, saw the largest single-session percentage decline among the major U.S. names on June 23, underscoring how directly its valuation is tied to AI memory demand expectations.

The scale of the PHLX Semiconductor Index's approximately 8% decline in a single session illustrates the degree to which the sector has become concentrated around AI spending narratives. Nvidia, with a market capitalization of $4.85 trillion and a 52-week range of $145.50 to $236.54, carries outsized weight in semiconductor and broader technology indices, meaning its 3% to 3.45% decline alone has significant index-level implications. AMD, trading at a price-to-earnings ratio of 174.4, and TSM, which manufactures advanced chips for both Nvidia and AMD, also registered meaningful losses, indicating the concern is not isolated to memory but extends to the broader AI accelerator ecosystem. Broader questions about whether AI capital expenditure — much of it debt-funded — can sustain current growth rates added a macro dimension to what began as a supply-chain-specific catalyst.

Sectors and assets to watch

The most directly affected names remain those with the highest revenue exposure to AI infrastructure and HBM specifically. Micron Technology (MU), with a 52-week range of $103.38 to $1,213.56 and 53,000 employees dedicated largely to DRAM, NAND, and HBM production, is the U.S. company most immediately exposed to any shift in HBM demand or competitive supply dynamics initiated by SK Hynix's capacity decisions. Taiwan Semiconductor Manufacturing Company (TSM), which manufactures chips for Nvidia, AMD, and other fabless AI chip designers, reported revenue of approximately $4.10 trillion (TWD) and carries a market cap of $2.26 trillion; its 6.69% decline on June 23 reflects its role as a central node in the AI chip supply chain. Intel (INTC), which is simultaneously a chip designer and an expanding contract manufacturer through Intel Foundry Services, fell between 4.17% and nearly 5% on the session despite having comparatively less direct HBM exposure than Micron.

Nvidia (NVDA) and AMD (AMD) represent the demand side of the HBM equation — their AI accelerators, including Nvidia's A100 and H100 data center GPUs and AMD's Instinct accelerators, require HBM to function at peak performance. Any sustained reduction in HBM availability or a repricing of HBM supply contracts would have direct implications for the cost structure and delivery timelines of these products. The sector-wide nature of the June 23 decline, spanning memory, logic, foundry, and accelerator names, suggests investors are reassessing the interconnected dependencies across the entire AI semiconductor value chain.

What to watch next

Market participants and analysts will be monitoring any formal guidance or official announcements from SK Hynix and Samsung Electronics regarding the scope and duration of their HBM capacity adjustments, as well as any response from Micron on its own production plans. Upcoming earnings calls and capital expenditure disclosures from major cloud and hyperscaler companies — which are the primary buyers of AI accelerators and, by extension, the ultimate source of HBM demand — will be closely scrutinized for signals about whether AI infrastructure spending is decelerating. The trajectory of the PHLX Semiconductor Index in the sessions following June 23 will indicate whether the sell-off represents a discrete repricing event or the beginning of a more sustained rotation out of AI-exposed semiconductor names. Any commentary from Nvidia, AMD, or TSM on order books, lead times, or supply agreements related to HBM will also be closely watched.