Voltage Park
GPU cloud for AI workloads offering on-demand and reserved NVIDIA GPU compute, plus AI Factory blueprints.
At a Glance
About Voltage Park
Voltage Park is an AI cloud provider that owns its GPU hardware and software and offers on-demand and reserved compute for AI and machine learning workloads. The homepage announces that Voltage Park has merged with Lightning AI.
What It Is
Voltage Park provides GPU infrastructure for training, fine-tuning and inference. Users can self-serve on-demand nodes or sign long-term reserved contracts. The site lists NVIDIA Hopper (H100) and Blackwell-generation GPUs, hosted in Tier 3+ data centers in the US.
Beyond raw compute, the site lists cloud services including storage, bare-metal access, managed Kubernetes, virtual machines and observability. It also lists an AI Factory with prebuilt blueprints such as an image generator, video generator, podcast generator, agentic RAG, spreadsheet analyzer, SAM3 segmentation, markdown creator and a model playground.
Infrastructure and Deployment Options
The infrastructure page describes Dell PowerEdge XE9680 servers with NVIDIA HGX H100 GPUs and 3,200 Gbps NVIDIA Quantum-2 InfiniBand networking. The company states 99.982% uptime. Reserved clusters range from 64 to 4,064 HGX H100 GPUs per cluster. The enterprise page also lists HGX B200, B300, GB200 and GB300 systems; some are liquid-cooled.
Enterprise Security and Model Support
The enterprise page describes isolated VPC deployments with data sovereignty and model-agnostic inference for open or closed models, as well as fine-tuning. The company states it holds HIPAA and SOC 2 certifications and more than 200 controls. It also lists 24/7/365 support.
Community Discussions
Be the first to start a conversation about Voltage Park
Share your experience with Voltage Park, ask questions, or help others learn from your insights.
Pricing
On-Demand H100 GPU
Self-serve on-demand NVIDIA HGX H100 GPU rental with no contract; one H100 GPU for 1 hour starting at $1.99.
- Pay only for what you need
- 1-1016 HGX H100 GPUs (Ethernet) / 8-1016 HGX H100 GPUs (3200 Gbps Infiniband)
- Self-serve. Up within 15 mins. No minimum terms
- No hidden ingress, egress or support costs
Long-Term Reserve
Reserved H100 GPU capacity on 6+ month contracts; pricing via Contact Us.
- 6+ month contracts with friendly payment terms
- 32-8000+ H100 GPUs in a single configuration
- 3200 Gbps Infiniband
Capabilities
Key Features
- On-demand GPU nodes
- Dedicated reserved GPU clusters
- NVIDIA H100 and Blackwell GPUs
- InfiniBand networking
- Bare-metal access
- Managed Kubernetes
- Virtual machines
- Storage
- Observability
- Isolated VPC deployments
- Model-agnostic inference and fine-tuning
- AI Factory blueprints
- 24/7 support