Simple licensing for sovereign infrastructure
One unit, one annual subscription, one signed license file that works offline. Your cluster never depends on a connection to us.
The terms, in five lines
-
Priced per GPU node
Every GPU compute node is one license unit. Head, login, and build nodes are not counted. Two node classes, for nodes with 1–4 GPUs and nodes with 5–8 GPUs, with no difference by GPU generation. A license starts at two GPU nodes.
-
A subscription that never stops the cluster
The license is an annual subscription. If you do not renew, the last imported release and your licensed capacity stay in service, and jobs, draining, and recovery continue. Renewal covers new releases, support, and capacity changes.
-
A license file, not a license server
The license is a signed file installed offline. There is no license server, no outbound connection, and no telemetry.
-
24 months per release line
Each minor release line is supported for 24 months from its first general release. Security fixes arrive on that line as signed release bundles, through the same path you use in an air-gapped network.
-
Standard support included
Standard support is part of the subscription. Requests are handled at three priority levels, with response and resolution targets defined in the support agreement.
Demo first, then a paid pilot
Prove BareStax on your own hardware before you commit.
-
Guided demo
A remote session on the product, built around your team's questions.
-
Technical session
Your hardware, fabric, storage, identity, and transfer method.
-
Paid pilot
A separate head node and up to 4 GPU nodes for 90 days, starting when the hardware is ready.
-
Acceptance drills
Node join, two-node validation, a canary update, drain and recovery, an air-gapped import, audit export, and a control plane restore.
-
Production order
The pilot license is time-limited. If you order within 60 days, the full pilot fee counts toward the first year.
How your quote is built
We quote after we have discussed your cluster scope and support needs. The quote is based on:
- the number of GPU nodes and their class, 1–4 or 5–8 GPUs;
- your support needs; and
- the pilot, if you start with one.
Your team can review the technical overview and both checklists before any commitment. Resources