COMPUTE

Instances with their own kernel.

Every Instance is a hardware-virtualized virtual machine — its own kernel, its own memory, its own disk. Not a container sharing a host you can’t see.

PACKAGES

One Package shape, five sizes.

Five tiers, one shape. Memory is always twice the vCPU count, and every tier ships the same flat NVMe disk. No families, no matrix to decode — pick the size, everything else is identical.

1 vCPU · 2 GB · 50 GB NVMe
Squire
$5/mo
2 TB transfer per Instance, pooled · 1 IPv4
2 vCPU · 4 GB · 50 GB NVMe
Baron
$12/mo
2 TB transfer per Instance, pooled · 1 IPv4
4 vCPU · 8 GB · 50 GB NVMe
Viscount
$24/mo
2 TB transfer per Instance, pooled · 1 IPv4
8 vCPU · 16 GB · 50 GB NVMe
Earl
$48/mo
2 TB transfer per Instance, pooled · 1 IPv4
16 vCPU · 32 GB · 50 GB NVMe
Duke
$96/mo
2 TB transfer per Instance, pooled · 1 IPv4
IMAGES

Boot the OS you already know.

Boot from a clean, current operating system image.

Ubuntu
Debian
Alpine
Fedora
Rocky
AlmaLinux
Arch
FreeBSD
NETWORKING

Networking, attached to every Instance

Instances are private by default. Add a Reserved IP when something should be reachable, and place Instances on VPCs and subnets you define.

Explore networking →
ON-PREM

Building your own platform on hardware you control?

Explore Triton On-Prem →
STORAGE

Storage, attached to every Instance

Pair Instances with Object Storage for backups, artifacts, and static assets — reachable privately from your VPC.

Object Storage →

Block storage: not available today — see the storage roadmap.

ONE API

One API, three surfaces

The CLI, the API, and the Terraform provider drive the same control plane. Anything the console does, they do.

triton instance create \
  --image ubuntu-24.04 \
  --package baron   # 2 vCPU / 4 GB

From $5/mo — Squire, 1 vCPU / 2 GB / 50 GB NVMe

Start with a server that’s actually yours.