Private Beta — Charter Applications Open

Dedicated infrastructure.
Hardware you don’t share.

Compute, storage, and AI inference on hardware you don’t share. No hypervisors, no egress fees.

Early access pricing available to charter applicants.

ONE STACK COMPUTE 64c · 512GB STORAGE molten · vault AI INFERENCE forge moe · gpu smelt://stack-01
Three ways to run it

Bare metal, managed, or fully serviced.

01

Bare metal

Flatcar Linux, full root, IPMI. Your server, you manage it.

02

Managed Kubernetes

Single-tenant RKE2 with a dedicated control plane.

03

Managed services

Postgres, object storage, runners, and private networking.

What sets it apart

No hypervisor. No egress meter. No middleman.

$0
egress fees
99.9%
uptime SLA
25 Gbps
internal fabric
100%
dedicated hardware
Typical cloud
Smelt
Tenancy
Shared hypervisor
Whole dedicated server
Storage egress
Metered per GB
Free
Bulk storage
~$15–23 / TB-mo
~$0.49 / TB-mo
Included
Storage + traffic extra
NVMe + transfer included
Charter members — now recruiting

Deployment 001 — South Florida

A permanent rate, locked in at every future site. Early applicants only.

99 of 99 charter nodes remaining

Not ready to commit? Join the waitlist → — free, no payment required.

FAQ

Questions, answered.

What processors do you use?

AMD EPYC for general Compute; Intel Xeon AMX for Compute Ultra and Forge MoE. We guarantee a performance floor per tier, not a specific SKU.

Where are your servers located?

South Florida (MIA-1) at launch, with more sites on the roadmap.

What network do you offer?

25 Gbps internal between nodes, verifiable with iperf3. Internet upstream is shared; dedicated uplinks available.

Do you offer AI / GPU nodes?

Yes — Forge MoE: Xeon AMX + GPU running mixture-of-experts models on a single node.

When is Smelt not the right fit?

Bursty workloads that autoscale to zero, many-region needs today, or compliance we have not yet completed. For steady-state, dedicated hardware wins.