Bottom line: AI data centers increasingly require high-performance memory (HBM), pressuring the limited wafer production capacity of memory manufacturers. This is displacing affordable memory for smartphones and other consumer products, leading to significant price increases.
The persistent memory shortage is driving up prices for consumer electronics. As artificial intelligence and data centers require increasingly more high-performance memory (HBM), competition for limited production capacity between different memory types is intensifying dramatically.
David Oks provides the clearest explanation yet for why consumer products with memory should become significantly more expensive in the coming years. The core of the problem: memory manufacturers – of which only three major companies remain – have a fixed capacity for wafer processing. This fixed wafer capacity is divided between DDR memory for desktops and servers, LPDDR memory for mobile phones and energy-efficient devices, and HBM memory for GPUs.
Until recently, HBM received only two percent of wafer capacity. The enormous growth in AI data centers has raised this share to an expected 20 percent by the end of 2026. A decisive factor: a single gigabyte of HBM requires more than three times as much wafer capacity as a gigabyte of DDR or LPDDR.
Memory manufacturers have learned from the demise of their competitors and are investing deliberately conservatively in fab capacity – preferring too little rather than too much. The high profit margins and demand for HBM will constrain RAM production for consumer devices for several years to come. The smartphone market under $100, which is of enormous importance for markets such as Africa and South Asia, is already particularly affected.