Overview
- Announced at Supercomputing ’25, Arm will integrate NVLink IP so licensees can build CPU SoCs that connect to Nvidia accelerators over a coherent fabric instead of PCIe.
- Nvidia says the move reduces design complexity, cuts development costs, and speeds time to market for specialized server builds.
- The expansion brings NVLink Fusion support to two major microarchitecture developers and four CPU developers overall, according to Tom’s Hardware.
- Hyperscalers including Microsoft, Amazon, and Google are developing or deploying Arm-based CPUs, positioning them to pair those chips with Nvidia GPUs.
- Nvidia already sells Grace Blackwell, which couples multiple GPUs with an Nvidia Arm CPU, and it agreed to invest $5 billion in Intel to enable NVLink with Xeon as part of a broader multi-vendor strategy.