Google Cloud makes Z4M storage-optimised machines generally available
Google Cloud has made its Z4M machine series generally available as virtual-machine and bare-metal instances. Designed for AI training, inference, checkpointing and distributed KV caches, the systems provide up to 168,000 GiB of local SSD and 400 Gbps Cloud RDMA networking. The architecture targets workloads where storage latency leaves GPUs or TPUs waiting.