ChipsSeptember 2, 2026via SiliconAngle
Lenovo and Broadcom target server costs with memory tiering
Why it matters
As AI workloads drive up DRAM costs, memory tiering (shifting cold data to cheaper NVMe flash) is becoming a practical lever for data-center economics. This validation by major vendors signals the architecture is production-ready and could reshape how practitioners budget for inference clusters and agentic workloads.
Key signals
- Memory tiering shifts infrequently accessed pages from DRAM to NVMe flash in VMware Cloud Foundation 9.1
- Broadcom and Lenovo validated capability on Lenovo ThinkAgile VX
- Technology targets rising DRAM costs as a cost-reduction lever
- Implication: relevant to AI workload cost optimization and on-prem deployment economics
- Memory tiering shifts infrequently accessed pages from DRAM to NVMe flash
- Technology validated in VMware Cloud Foundation 9.1
- Broadcom and Lenovo partnership on Lenovo ThinkAgile VX
- Rising DRAM prices driving adoption of tiering strategies
- Targets data-center server cost reduction
The hook
Memory tiering cuts server costs as DRAM prices spike — Broadcom and Lenovo prove the play works at scale.
Memory tiering is emerging as a way to reduce server costs as dynamic random-access memory prices rise. In VMware Cloud Foundation 9.1, the technology shifts less frequently accessed pages to lower-cost nonvolatile memory express flash while keeping active data in DRAM. Broadcom Inc. and Lenovo Grou…