Volt Forge
GPU-as-a-service. Dedicated leases in your namespace.
Reserved NVIDIA capacity in your customer's metro. A scoped kubeconfig into a dedicated namespace, not a shared queue. Your data never leaves the city.
$2.36/GPU/hrNVIDIA B200, 36-month reserved — 31% below CoreWeave list
NVIDIA B200 and L40S
Dedicated GPU capacity in-metro. B200 for frontier training and inference, L40S for cost-efficient serving.
Scoped kubeconfig
We hand you a scoped kubeconfig into a dedicated namespace. Bring your own workloads. Zero egress by default.
Reserved discounts
Reserve capacity and pay less. 45% off on-demand at 12 months, 60% off at 36 months.
Reserved terms
Forge is sold as reserved capacity. The longer you commit, the lower the rate. There's no egress fee and no inter-pod transfer — data residency is structural, not contractual.
- On-demand: full flexibility at list rate
- 12-month reserved: 45% off on-demand
- 36-month reserved: 60% off on-demand, lands at $2.36/GPU/hr on B200
Run frontier models in your metro.
Zero egress, in-metro serving, at Bedrock-beating prices. Your data never leaves the city.