Google Cloud Launches C4N Network-Optimized VMs with 400 Gbps

As organizations scale modern workloads — from high-throughput databases and network appliances to real-time analytics and AI/ML inference — standard virtual machines often become bottlenecked by network and block storage I/O. The challenge is balancing compute efficiency with the high-volume data-transfer demands these applications require, leading to over-provisioning and wasted cost. Google Cloud aims to address this head-on with its new C4N instance family, now generally available.

C4N is built on Google’s custom Titanium offload architecture, which dedicates hardware to networking and storage tasks, freeing up CPU cycles for application work. The instances deliver up to 400 Gbps of network bandwidth and 95 million packets per second (MPPS) — nearly 33% higher network bandwidth per vCPU and 224% faster packet processing than comparable Intel-based offerings at other hyperscalers. When paired with Hyperdisk Extreme, C4N achieves up to 25 GiB/s of storage throughput and 1M IOPS, a 2x improvement over C4. The instances are powered by 5th Gen Intel Xeon Scalable processors and come in nine shapes from 2 to 192 vCPUs, with up to 1.5 TB of DDR5 memory. Smaller shapes (2–16 vCPUs) deliver 25–50 Gbps of network bandwidth, allowing cost-effective I/O acceleration without over-provisioning compute. Customer quotes from Ericsson, Teradata, NetApp, Sycomp, and ClipperDB highlight real-world gains: up to 1.5x more Nginx requests per second, 45% better MySQL QPS, and 3x lower cost per query in TPC-DS benchmarks with ClipperDB.

For serious builders, the key takeaway is that C4N changes the price-performance equation for network- and storage-intensive workloads on Google Cloud. By eliminating the need to over-provision resources just to meet I/O demands, it delivers significant TCO benefits. The instances are available now in us-central1, us-east1, us-east5, us-west1, and europe-west2 with Committed Use Discounts. Whether you’re running virtual appliances, large-scale databases, telco 5G UPF, or CPU-based AI/ML inference, C4N provides predictable, high-throughput I/O at scale. Bare metal and local SSD variants are coming soon, further expanding the options for performance-sensitive environments.