Sign in →
CloudLab Works emblem: Waku the orca ringed by CloudLab and WorksCLOUDLAB WORKScrossed whale bones, one end a wrenchCloud City, headwaters of the agentic cloud revolution!

Nutanix: Invisible Wires: Agentic Cloud · Nutanix Knowledge Base

NC2 on Google Cloud

Nutanix Cloud Clusters (NC2), Google Cloud · verified against the sources below on

Instances

Six instance types, in two families that differ in where AOS keeps its data.

Local SSD (Table 7 in the guide). AOS runs on the node's own NVMe.

InstanceProcessorCores / vCPUsMemoryLocal storageNetwork
z3-highmem-192-highlssd-metalIntel Sapphire Rapids 8481C, 2.2 GHz, 3.0 GHz turbo96 / 1921,536 GB DDR512 x 6 TB NVMe, 72,000 GBup to 200 Gbps, Tier 1
c4-standard-288-lssd-metalIntel Granite Rapids 6985P-C, 3.0 GHz, 3.9 GHz turbo144 / 2881,080 GB DDR56 x 3 TB, 18,000 GBup to 200 Gbps, Tier 1
c4-highmem-288-lssd-metalIntel Granite Rapids 6985P-C, 3.0 GHz, 3.9 GHz turbo144 / 2882,232 GB DDR56 x 3 TB, 18,000 GBup to 200 Gbps, Tier 1

Hyperdisk Balanced (Table 8). No local storage. Capacity is remote and decoupled from the node. Added to the guide on July 27, 2026.

InstanceProcessorCores / vCPUsMemoryRemote storage per hostIOPS / throughput per host
c3-standard-192-metalIntel 4th Gen Xeon (Sapphire Rapids), 2.8 GHz96 / 192768 GB15 to 120 TB30,000 / 1,400 MB/s
c3-highcpu-192-metalsame96 / 192512 GB15 to 120 TB30,000 / 1,400 MB/s
c3-highmem-192-metalsame96 / 1921,536 GB15 to 120 TB30,000 / 1,400 MB/s

Each C3 node carries four data disks and two boot disks (one AHV, one CVM). Three constraints come with Hyperdisk: never change a volume's capacity or performance from the Google Cloud console; disks cannot be added to or removed from an existing node, so more capacity means another node; and a single failed volume is handled by replacing the whole node.

The vCPU count is the cloud provider's billing and quota figure, not a scheduling limit.

Regions

Twenty-three regions (Table 9), and availability is per instance type. Three regions carry all six types: asia-southeast1, europe-west4 and us-central1. Pick the region for the instance, not the other way round.

Regionz3-highmemc4-standardc4-highmemc3-standardc3-highcpuc3-highmem
asia-east1YesNoNoYesNoNo
asia-northeast1YesNoNoNoNoNo
asia-south1YesNoNoYesNoNo
asia-southeast1YesYesYesYesYesYes
australia-southeast2YesNoNoNoNoNo
europe-north2NoNoNoYesNoNo
europe-southwest1NoNoYesNoNoNo
europe-west1YesNoNoYesYesYes
europe-west2YesNoNoNoNoNo
europe-west3YesNoNoYesNoNo
europe-west4YesYesYesYesYesYes
europe-west8YesNoNoNoNoNo
me-central1YesNoNoYesNoYes
me-central2NoNoYesNoNoNo
northamerica-northeast2YesNoNoNoNoNo
southamerica-east1YesNoNoNoNoNo
us-central1YesYesYesYesYesYes
us-east1YesNoNoYesYesYes
us-east4YesNoYesYesYesYes
us-east5YesNoNoYesYesYes
us-south1NoNoYesNoNoNo
us-west1NoNoNoYesYesYes
us-west2YesNoNoNoNoNo

The guide's change log shows how fast this table moves: asia-northeast1 was added July 2, 2026, the C3 family July 27, 2026, and asia-south1 August 14, 2026. Re-read the table rather than trusting this copy.

Cluster limits

Networking

Subnet sizes, from the planning chapter:

SubnetMinimum
Cluster (management) subnet/24, with a /28 for Prism Central carved out of it
Flow Virtual Networking NAT subnet/27
Flow Virtual Networking no-NAT subnet/27

Ranges the cluster reserves for itself, which the VPC must not use: 192.168.5.0/24, 10.100.0.0/16, 10.200.32.0/24, 10.200.0.0/22 and 100.64.1.0/24.

Automation

Cluster lifecycle is the NC2 console API, POST /clusters/gcp on https://cloud.nutanix.com/api/v2, which returns a task to poll. Everything after Prism Central answers is the v4 API; see One API for the estate. The full walk with Pulumi for the landing zone is in the post NC2 on Google Cloud, as code.

Sources

Facts here move. The sources listed above win over this page on any day after the verified date. Personal blog. Alex works at Nutanix; the opinions here are his own and nothing here is Nutanix confidential: every fact is public or his own field experience.