Nutanix: Invisible Wires: Agentic Cloud · Nutanix Knowledge Base
NC2 on AWS
Instances
Table 12 in the guide lists these EC2 bare-metal types. Rows marked with two asterisks in the guide need AOS 7.5.1.x or later.
| Instance | Physical cores / vCPUs | Memory | Local NVMe |
|---|---|---|---|
i7ie.metal-24xl | 48 / 96 | 768 GB | 60 TB |
i7ie.metal-48xl | 96 / 192 | 1,536 GB | 120 TB |
i7i.metal-24xl | 48 / 96 | 768 GB | 22.5 TB |
i7i.metal-48xl | 96 / 192 | 1,536 GB | 45 TB |
i4i.metal | 64 / 128 | 1,024 GB | 30 TB |
m6id.metal | 64 / 128 | 512 GB | 7.6 TB |
m5d.metal | 48 / 96 | 384 GB | 3.6 TB |
i3.metal (existing clusters only) | 36 / 72 | 512 GB | 15.2 TB |
i3en.metal | 48 / 96 | 768 GB | 60 TB |
z1d.metal | 24 / 48 | 384 GB | 1.8 TB |
g4dn.metal | 48 / 96 | 384 GB, plus 128 GiB GPU memory | 1.8 TB |
m8id.metal-48xl (AOS 7.5.1.x+) | 96 / 192 | 768 GB | 11.4 TB |
m8id.metal-96xl (AOS 7.5.1.x+) | 192 / 384 | 1.5 TB | 22.8 TB |
r8id.metal-48xl (AOS 7.5.1.x+) | 96 / 192 | 1.5 TB | 11.4 TB |
r8id.metal-96xl (AOS 7.5.1.x+) | 192 / 384 | 3 TB | 22.8 TB |
c6id.metal (AOS 7.5.1.x+) | 64 / 128 | 256 GB | 7.6 TB |
r6id.metal (AOS 7.5.1.x+) | 64 / 128 | 1,024 GB | 7.6 TB |
c8id.metal-48xl (AOS 7.5.1.x+) | 96 / 192 | 384 GB | 11.4 TB |
c8id.metal-96xl (AOS 7.5.1.x+) | 192 / 384 | 768 GB | 22.8 TB |
r6idn.metal (AOS 7.5.1.x+) | 64 / 128 | 1,024 GB | 7.6 TB |
x2idn.metal (AOS 7.5.1.x+) | 64 / 128 | 2,048 GB | 3.8 TB |
x2iedn.metal (AOS 7.5.1.x+) | 64 / 128 | 4,096 GB | 3.8 TB |
m6idn.metal (AOS 7.5.1.x+) | 64 / 128 | 512 GB | 7.6 TB |
i3.metal is a legacy type as of the September 8, 2026 change: it cannot create a new cluster, hibernate one, or be added during expansion, including to clusters that already contain it. Existing i3.metal nodes can stay, be removed, or be migrated to a supported type.
The vCPU count is AWS's billing and quota figure, not a scheduling limit.
Mixing instance types
- A cluster may mix at most three instance types, subject to bare-metal availability in the region. The guide's Creating a Heterogeneous Cluster section lists the allowed combinations.
g4dn.metalclusters are homogeneous only.- NC2 needs three or more partitions in the region for a given type; some types are absent from some regions for that reason.
Tenancy
Two EC2 tenancy models. Default tenancy lets a bare-metal instance land on a different physical host after a power cycle and spreads instances across partitions with placement groups; it is also the model that allows Windows Server License Included instances. Dedicated Host pins each instance to a physical host NC2 allocates, and it stays there across power cycles.
Regions
Table 13 in the guide is a matrix of region by instance type, and it runs to several pages because availability differs per type. Read the row for the instance you want; a region count on its own is misleading here.
Cluster limits and networking
- Twenty-eight nodes maximum per cluster.
- Prism Central subnet: /28, and /28 is what the console selects by default.
Automation
Cluster lifecycle is the NC2 console API on https://cloud.nutanix.com/api/v2 (POST /clusters/aws, then poll the task). Everything after Prism Central answers is the v4 API; see One API for the estate. The Terraform walk is in the post NC2 on AWS, as code.
Sources
- Nutanix Cloud Clusters on AWS Deployment and User Guide, hosted PDF dated September 9, 2026 on its cover. Supported Regions and Bare-metal Instances (Tables 12 and 13), the change log entries for September 8, 2026 and July 3, 2026, the tenancy section, the limitations list. Read 2026-09-23. Also on portal.nutanix.com under Cloud Clusters (NC2).
- NC2 API reference: the documented 3-node AWS create example and task polling. Read 2026-09-23.
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.