News / Taiwan / cnyes
Google Scrambles for DRAM as AI Demand Outstrips Supply, Boosting DDR4 Outlook
Electronic Technology · Semiconductors · cnyes · 2026-09-02
2344, 2408, 6770
Google warns of severe memory shortages, revealing it is recycling old DDR4 modules for AI servers, signaling a potential windfall for manufacturers like Nanya and Winbond.
What Happened
Memory Bottleneck Crisis: Google's supply chain infrastructure director, Nikhil Cherian, stated at the SEMICON forum that the industry has shifted from being compute-bound to memory-constrained. Despite aggressive software and hardware optimizations, Google struggles to keep its expensive AI chips fully utilized due to insufficient memory bandwidth.
DDR4 Recycling Strategy: To cope with the severe supply crunch, Google has begun harvesting DDR4 memory from decommissioned servers to integrate into modern AI infrastructure. This move highlights that the insatiable demand for AI memory has spilled over from HBM and DDR5 into legacy DRAM, extending the lifecycle of older components.
Impact on Taiwanese Manufacturers: Industry analysts suggest that Google's reliance on legacy DRAM provides a significant tailwind for Taiwanese memory makers such as Nanya Technology, Winbond, and PSMC. The sustained demand for DDR4 is expected to bolster capacity utilization rates and pricing for these companies in the near term.
Call for Increased Production: Google has issued a direct plea to memory suppliers to accelerate and expand production capacity. As hyperscalers continue to build out AI infrastructure, the persistent supply-demand gap suggests that memory manufacturers will face sustained pressure to ramp up output to meet the company's ambitious AI roadmap.