Memory bandwidth is the hard physical ceiling on AI inference speed. For every token a large-language model generates, the accelerator must repeatedly read billions of model weights from memory at high frequency, and the only memory architecture fast enough to keep pace with modern GPU compute is high-bandwidth memory (HBM), a stacked DRAM design built to maximize throughput at the die level. The companies that design those accelerators and manufacture that memory have taken the sharpest losses in the current chip selloff. Nvidia, SK Hynix, and Samsung Electronics have led the sector since Friday's market close, with combined chip-stock losses now exceeding $1 trillion.
Where these names sit in the stack
Nvidia designs the GPU accelerators at the top of the AI compute hierarchy. SK Hynix and Samsung Electronics are the primary manufacturers of HBM and together supply the memory that ships on those accelerators. The supplier relationship is structural rather than incidental. Nvidia qualifies HBM vendors product by product, which means its accelerator roadmap and the HBM capacity plans at SK Hynix and Samsung are co-developed across successive product generations. That coupling means a sentiment shift on AI capital expenditure reprices all three names simultaneously rather than sequentially, and positions each company as a direct proxy for AI infrastructure spending at the silicon level.
What the numbers cover
Combined chip-sector losses since Friday's close exceed $1 trillion. The source identifies Nvidia, SK Hynix, and Samsung as the leaders of the move but provides no per-company breakdown, no individual share prices, and no percentage declines for any name. The $1 trillion figure is an aggregate across the chip sector, not the combined loss of those three companies alone.
What the repricing signals
The standard "AI boom" narrative concentrates on demand. A sector drawdown of this magnitude is a question about conversion: how much AI demand actually translates into sustained capital expenditure at the chip layer, where procurement lead times are longest and unit economics are most exposed to volume assumptions. Nvidia, SK Hynix, and Samsung are the names closest to that conversion point, which is why all three have led the sector lower since Friday's close.