🧠 Food for thought

Implications, context, and why it matters.

Why the memory chip shortage looks more lasting than temporary

• Analysts tie the shortage to memory makers moving capacity and spending from conventional DRAM, a common type of memory used in PCs and servers, to high-margin High Bandwidth Memory (HBM), a faster memory used with AI chips 1.
• That shift brings strong returns, with one large memory maker posting a 72% operating margin as demand for AI memory climbs 2.
• As more resources flow into HBM, supply of conventional memory for PC and smartphone makers can tighten, pushing up prices across consumer electronics 3.
• New chip fabrication plants and advanced packaging capacity, which prepares chips for use in products, take years to build, and some large additions are not due until 2027 or 2028 3.

Long-term memory contracts aim to curb boom-bust swings while tightening supply elsewhere

• To meet lasting demand, suppliers including SK Hynix and Samsung Electronics are moving from short-term sales to three- to five-year long-term agreements (LTAs) with hyperscalers, meaning large cloud computing companies such as Microsoft and Google 4.
• This changes how the industry sells memory, with the goal of cutting volatility and giving suppliers a clearer view of future revenue through a built-to-order model 4.
• For large cloud providers, these deals can lock in memory supply and may include prepayments plus less room to adjust prices 4.
• Smaller business buyers may face a harder market because multi-year agreements can lock up future output and keep prices higher for other enterprise and consumer customers 5.