Google Dismantles Retired Servers for DDR4 Memory Modules, Redeploys Them in New AI Servers
Summary
According to IT Home on September 2, citing Wccftech, DRAM supply tightness is forcing AI service providers to take extraordinary measures. Google's senior director of supply chain infrastructure revealed that the AI industry has shifted from being 'compute-constrained' to 'memory-constrained', with high-performance memory accounting for about 75% of the bill of materials cost in a typical AI server. To break the bottleneck, Google is dismantling retired servers to recover DDR4 memory modules and using them in new AI servers via dedicated adapter cards, while also optimizing software to reduce memory overhead.
Bubble analysis
This news shows that memory shortages are forcing AI companies to resort to unconventional measures like recycling old memory, reflecting high costs and tight supply in AI hardware investment, which could exacerbate bubble risks as rising costs may squeeze profits and drive up investment demand.
Related signals
linked by shared tags
- Sources: Samsung Allocates Over Half of Its 4nm Capacity to HBM4 Chips#memory
- Rumors of Listed Company's 'Computing Cards Stuck at Customs' Lead to Penalties#supply-chain
- CXMT Announces LPDDR6 Mass Production, Debuting in Xiaomi 18 Fold#memory
- Wistron Turns To Global Markets For $1.5 Billion Raise#ai-server
- Weekly AI Model Rankings: Multiple Chinese Models Enter Frontend Development List, Alibaba's wan3.0 Jumps into Top Three of Video List#google