Skyrocketing Memory Prices Forcing Tech Giants to Become Hardware Scavengers? Google Dismantling Old Servers Amid Critical Shortage
· Source: eu.36kr.com
In short
Rising DRAM prices and a structural shift of memory supply toward HBM have moved the AI industry's bottleneck from computing power to memory, with high-performance memory now accounting for roughly 75% of the material cost of a single AI server. Google's supply chain infrastructure director Nikhil Cherian revealed the company has built an internal recycling supply chain, dismantling decommissioned servers to recover DDR4 modules and using custom adapters to connect them to new-generation AI servers, while also optimizing software, KV Cache compression and its new TPU8i chips to cut memory consumption. The article argues this reflects extreme cost control, a drive for supply chain resilience and a dangerous systemic signal that even a giant like Google can no longer secure affordable general-purpose memory.
Read Original Article — Read the full article on the publisher’s site.
In the issue of week 38, 2026: AI Moves Closer to Supply Chain Decisions, but Data and Returns Still Lag
Tags: supply chain, memory, hardware, data centers, cost control, recycling