Accessible and affordable cloud AI infrastructure
Last reviewed Mar 14, 2026
Voltage Park provides on‑demand and reserved bare‑metal clusters built on NVIDIA HGX H100, with Blackwell systems (B200/B300, GB200/GB300) offered on long‑term contracts. Merged with Lightning AI to create an integrated cloud platform for AI development.
Hourly on-demand pricing. Click column headers to sort.
Prices last updated: September 12, 2026
| H100 SXM | $1.99/hr 1× | 1× |
Configurations, price rank, and alternatives for one GPU at a time.
NVIDIA HGX H100 GPUs in Dell PowerEdge XE9680 servers with 1TB RAM and v52 CPUs
Six U.S.-based data centers with redundancy, availability, and reliability across power, cooling, network, and security
NVIDIA Quantum-2 InfiniBand network with up to 3,200Gbps of aggregate bandwidth for high-speed communication
Direct hardware access for maximum performance without virtualization overhead
Integrated platform of prebuilt blueprints — image, video and podcast generation, agentic RAG, spreadsheet analysis, SAM3 image/video segmentation, markdown conversion — plus a model playground
Bare-metal access, managed Kubernetes, and virtual machines to match different workload needs
Separate storage and observability products offered alongside compute
On-demand clusters can be provisioned on Ethernet or 3,200 Gbps InfiniBand to balance speed against budget
Top-tier firewalls and comprehensive security protocols including encryption, access controls, and regular audits
Self-serve GPU instances available in minutes with no long-term contracts
Long‑term GPU reservations with favorable payment terms; H100 and Blackwell options.
| Option | Details |
|---|---|
| On-Demand (Ethernet) | Self-serve, pay-as-you-go bare-metal H100 nodes on Ethernet with no minimum term |
| On-Demand (InfiniBand) | Self-serve, pay-as-you-go bare-metal H100 nodes on 3,200 Gbps InfiniBand with no minimum term |
| Long-Term Reserve | Dedicated reserved capacity on 6+ month contracts with negotiated payment terms; the only route to Blackwell systems |
Six Tier 3+ designed data centers across four U.S. states
24/7 expert support included at no extra cost; self-serve help center at support.voltagepark.com (also reachable at docs.voltagepark.com); infrastructure partnership with Penguin Solutions for setup and optimization
Sign up for Voltage Park services through their website
Select between on-demand access or long-term reservation
Specify the number of GPUs and networking requirements
Launch your cluster and access via SSH or preferred tools
Voltage Park offers various GPU types including H100 SXM. Check the pricing table above for current availability and pricing.
Create an account, Choose deployment option, Configure your cluster, Deploy and access
Voltage Park's main advantages include: Great GPU availability for on-demand use, No separate ingress, egress, or support charges, Fast deployment (spin up nodes in 15 minutes), Bare metal access for maximum performance, Exceptional customer service and support, NVIDIA Cloud Partner status, Scales from a single GPU up to 8,176-GPU InfiniBand clusters.
Voltage Park's main limitations include: U.S.-only regions (limited global presence), Low GPU variety (primarily H100s), Relatively new provider (founded 2023).
Find the best prices for the same GPUs and models from other providers