Pricing that fits how you run agents.
Every plan is set up with your team, around what your agents actually need.
Dedicated GPU capacity
NVIDIA clusters reserved for your training and inference.
Priced by capacity and term
Ask about capacity- GB300 NVL72 and HGX B300, with Vera Rubin capacity reserved
- Dedicated to you, never shared with other customers
- High-speed networking and storage included
- US data centers, run by our own engineers
Platform
Build, improve and own your agents on Voltspan Cloud.
Usage-based, with committed plans
Contact us →- Our agents help build and improve yours
- Every run captured, searchable and kept yours
- Training for a model of your own, with weights you own
- Inference for open models and the ones you train
- Isolated sandboxes for agents to act
Your VPC or data center
The full platform, deployed where your data already lives.
Custom
Contact us- Everything in Platform
- Deployed in your cloud account or on-premises
- Single sign-on and access controls
- A dedicated engineering team
- Custom terms and security review
Questions
How is pricing set?
Infrastructure is priced by the capacity you reserve and for how long. Platform is priced by usage, with lower rates on committed plans. We’ll scope it with you on a short call.
Can we start small?
Yes. Most teams start with one agent: capture its runs, find where it fails, and compare models on their own tasks before training anything.
Do you train on our data?
Only to train your model. Your traces are never pooled with other customers, resold, or used to train anyone else’s model.
Who owns the models we train?
You do. The weights are yours, and you can run them on our cloud, in your VPC, on-premises, or take them with you.
Where does our data live?
Where you choose: our cloud in US data centers, your VPC, or your own data center.