The memory shortage is causing a repricing of consumer electronics
Summary
The memory shortage is causing a significant repricing of consumer electronics, particularly impacting affordable smartphones. This trend stems from the fixed wafer capacity of the three major memory manufacturers, which must allocate production across DDR, LPDDR, and HBM. Historically, HBM, used in GPUs for AI data centers, received only 2% of this capacity. However, due to the immense growth in AI, HBM's allocation is projected to surge to 20% by the end of 2026. Crucially, a single gigabyte of HBM requires over three times the wafer capacity of a gigabyte of DDR or LPDDR. Memory companies, having learned from past market oversupply, prioritize under-provisioning capacity. Consequently, the high profit margins and demand for HBM will constrain the production of RAM for consumer devices for several years, a situation already evident in the sub-\$100 smartphone market, vital for regions like Africa and South Asia.
Key takeaway
For product managers developing consumer electronics, especially in cost-sensitive markets, you should anticipate sustained increases in memory component costs over the next few years. This shift, driven by AI's demand for HBM, necessitates re-evaluating bill of materials and pricing strategies. Consider exploring alternative component sourcing or design optimizations to mitigate the impact on device affordability and market competitiveness, particularly for entry-level devices.
Key insights
AI's surging demand for HBM is drastically reallocating memory wafer capacity, driving up consumer electronics costs.
Principles
- Memory manufacturers prioritize under-provisioning capacity.
- HBM production consumes significantly more wafer capacity.
- High HBM demand constrains consumer RAM supply.
In practice
- Expect higher prices for memory-dependent devices.
- Sub-\$100 smartphone market faces significant pressure.
Topics
- Memory Shortage
- HBM
- AI Data Centers
- Consumer Electronics
- Smartphone Market
- Wafer Capacity
Best for: CTO, VP of Engineering/Data, Director of AI/ML, Executive, Investor, Product Manager
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by Simon Willison's Weblog.