Google Turns to Server Recycling Amid Memory Price Surge

The sudden surge in memory prices has forced tech giants to get creative with their resources. Google, one of the largest internet companies, is now dismantling old servers to recover usable DDR4 memory modules. According to Nikhil Cherian, Senior Director of Supply Chain Infrastructure at Google, the AI industry has shifted from being computing power-constrained to memory-constrained.

High-performance memory currently accounts for about 75% of the material cost of a single AI server. To break through the memory bottleneck, Google is working on both hardware and software solutions. On the hardware side, the company has built an internal recycling supply chain to dismantle decommissioned servers and recover DDR4 memory modules.

Google has also designed hardware adapters to connect previous-generation DDR4 memory to new AI servers. The company is reintroducing decommissioned servers to remove the DDR4 modules inside. At the software level, Google continues to optimize underlying function libraries, model architecture, and KV Cache compression technology to reduce memory consumption per unit of computing power.

The structural imbalance on the supply side is the root cause of the problem. The three core giants - Samsung, SK Hynix, and Micron - have allocated most of their advanced process production capacity and packaging resources to HBM. This has led to a serious shortage in the consumer market, making it difficult for companies like Google to buy cost-effective general-purpose memory.

Google dismantling old servers is not just about scavenging for resources but also about cost control. The company aims to minimize its expenses and maintain its profit margins. By reusing existing hardware assets, Google can reduce its reliance on expensive brand-new memory and extend the life of its infrastructure systems.