Google Cloud launches C4N virtual machines with 400 Gbps bandwidth, and the cloud infrastructure race just got more interesting
The new network-optimized VMs deliver up to 8x better egress bandwidth, positioning Google to compete more aggressively for high-performance computing workloads
Google Cloud just flipped the switch on its C4N network-optimized virtual machines, making them generally available on Compute Engine. The machines deliver up to 400 Gbps VM-to-VM bandwidth and can process 95 million packets per second. For anyone running data-intensive workloads in the cloud, those numbers matter a lot.
The C4N instances were first previewed at Google Cloud Next ’26 in April. Now they’re available for production use, and the specs represent a meaningful leap over Google’s standard C4 series: up to 4x better bandwidth per virtual CPU and up to 8x higher egress bandwidth when routing through internet gateways.
What the C4N actually does differently
In technical terms, the C4N achieves its performance through Titanium offload technology, Google’s custom silicon designed to handle network processing without burdening the main CPU. The result is that 400 Gbps bandwidth figure, which represents the kind of throughput previously reserved for bare-metal servers or highly specialized configurations.
On the storage side, C4N supports up to 25 GiB/s throughput and nearly 1 million IOPS through Hyperdisk Extreme integration.
The instances run on fifth-generation Intel Xeon processors, pairing mainstream server chips with Google’s proprietary networking silicon.
Google is targeting specific use cases with the C4N: network and security appliances, high-performance databases, Telco 5G infrastructure, real-time data analytics, and distributed file systems.
The bigger cloud infrastructure picture
The C4N launch fits into what Google calls its “fluid compute” strategy, an approach aimed at enabling seamless scaling across general-purpose and specialized infrastructure.
The 400 Gbps bandwidth figure is particularly noteworthy. It positions Google Cloud at the leading edge of VM networking performance among major cloud providers.
What this means for investors and the market
Google Cloud has been the third-place player behind AWS and Azure for years.
A 4x improvement in bandwidth per vCPU means fewer instances needed to achieve the same throughput, which directly impacts cost efficiency. The 8x improvement in egress bandwidth addresses one of the most persistent complaints about cloud networking: getting data out of the cloud efficiently.
Nearly 1 million IOPS and 25 GiB/s throughput through Hyperdisk Extreme puts the C4N in competition with dedicated storage appliances that cost significantly more.