Voltage Park
Voltage Park is a US-based GPU cloud provider operating 36,000+ owned NVIDIA HGX H100 and Blackwell GPUs across six Tier 3+ data centers, serving AI startups, enterprises, researchers, and AI labs via on-demand rentals, long-term reserved clusters, and the post-merger Lightning AI software stack.
- Company typePrivate
- Founded2023
- HeadquartersBerkeley, United States
- Headcount101–250
- GTM typeB2B
- OfferingSoftware
What Voltage Park does
Voltage Park is a US-based GPU cloud provider that owns and operates a fleet of 36,000+ NVIDIA HGX H100 and Blackwell (B200, GB200, B300, GB300) GPUs across six Tier 3+ data centers in four US states, delivering AI and machine learning infrastructure to AI startups, enterprises, researchers, and AI labs. Its platform spans on-demand self-serve H100 rentals starting at $1.99/hr (with 15-minute provisioning and no ingress/egress/support fees), long-term Dedicated Reserve contracts for 6–12+ month reserved clusters of 32–8,000+ GPUs, bare-metal access with bring-your-own orchestrator (Slurm, Kubernetes, Ray), managed Kubernetes with NVIDIA GPU Operator, and VAST-powered unified storage for trillion-parameter checkpointing. The stack is interconnected by 3,200 Gbps NVIDIA Quantum-2 InfiniBand and supports HIPAA, SOC 2 Type II, and ISO/IEC 27001-compliant enterprise deployments.
The company employs a hybrid go-to-market: product-led growth via a self-serve web dashboard for on-demand buyers, complemented by an enterprise sales motion for reserved Blackwell fleets and channel partnerships (e.g., JBK Group, Matrice.ai) for international Vision AI factory deployments. In March 2025 Voltage Park acquired TensorDock to broaden GPU cloud distribution, and in October 2025 launched the AI Factory platform — a higher-value offering that packages AI blueprints (Image Generator, Video Generator, Podcast Generator, Agentic RAG, Spreadsheet Analyzer, SAM3 Segmentation) and hosted third-party foundation models on top of the GPU fleet. In January 2026, Voltage Park completed a merger with Lightning AI, combining its owned GPU infrastructure with Lightning's PyTorch Lightning software stack under the Lightning AI name at a $2.5B combined valuation, $500M+ ARR (up from $18M in 2024), and a 400,000+ developer and enterprise user base.
Voltage Park firmographics
Firmographics- Name
- Voltage Park
- Legal name
- Voltage Park, Inc.
- Website
- https://voltagepark.com
- Company type
- Private
- Founded year
- 2023
- Operating status
- Acquired
- Headcount range
- 101–250 employees
- Short description
- Voltage Park is a US-based GPU cloud provider operating 36,000+ owned NVIDIA HGX H100 and Blackwell GPUs across six Tier 3+ data centers, serving AI startups, enterprises, researchers, and AI labs via on-demand rentals, long-term reserved clusters, and the post-merger Lightning AI software stack.
- Ownership category
- akta.pro rank
Voltage Park industry classification
Industry- Product category
- AI Cloud Infrastructure (GPU-as-a-Service)
- NAICS
- Computing Infrastructure Providers, Data Processing, Web Hosting, and Related Services (518)
- SIC
- Services-Computer Rental & Leasing (7377)
- akta.pro primary industry
- AI Server Systems & HGX/Accelerator Platforms (HDAAAAAB)
- akta.pro secondary industries
- AI Compute Cloud & GPU-as-a-Service (HDAAAAAK), AI Compiler, Runtime & Kernel Optimization Software (CUDA/ROCm/XLA, graph compilers) (HDAAAAAI)
Keywords
Where Voltage Park is headquartered
LocationHeadquarters
- HQ city
- Berkeley
- HQ country
- United States
- HQ region
- North America
Offices5 records
Markets served
Voltage Park business model
Business model- GTM type
- B2B
- Offering type
- Software
- Cost components
- Infrastructure, Technology or R&D, Personnel, Operations, Marketing or Sales
Revenue model
- On-Demand GPU Compute: Usage-based, self-serve rental of NVIDIA HGX H100 GPUs (Ethernet or 3,200 Gbps InfiniBand) with no minimum terms and per-hour pricing starting at $1.99/hr; customers can spin up nodes in 15 minutes.
- Dedicated Reserve (Long-Term): Recurring long-term contracts (6+ or 12+ months) for reserved, isolated GPU clusters (H100, B200, GB200, B300, GB300) typically 128–8,000+ GPUs, with friendly payment terms and dedicated VPC deployment.
- AI Factory Platform: Higher-value AI Factory deployments for enterprises (incl. templates like Vision AI, Agentic RAG, Spreadsheet Analyzer) combining GPU infrastructure with bundled AI software tooling.
- Managed Kubernetes / Virtual Machines / Storage add-ons: Adjacent managed services layered on the GPU compute base, including Managed Kubernetes, GPU-powered Virtual Machines, bare-metal access, and VAST-backed storage for AI workloads.
Pricing tiers
| Model | Billing | Price |
|---|---|---|
| Usage-based | Pay-as-you-go | On-Demand H100 — Ethernet (self-serve) |
| Usage-based | Pay-as-you-go | On-Demand H100 — 3,200 Gbps InfiniBand |
| Subscription | Multi-year contract | Dedicated Reserve — Long-Term H100 (and Blackwell B200/GB200/B300/GB300) |
| Other | Multi-year contract | AI Cloud Credit for Research program |
Go-to-market motion4 records
Distribution channels5 records
Marketing channels9 records
Voltage Park product offering
Product offeringCore offering
Voltage Park operates an AI cloud platform delivering owned NVIDIA HGX H100 and Blackwell (B200, GB200, B300, GB300) GPU infrastructure on a self-serve on-demand basis starting at $1.99/hr and via long-term Dedicated Reserve contracts. The platform bundles compute, 3,200 Gbps InfiniBand fabric, VAST-powered unified storage, managed Kubernetes, bare-metal access, virtual machines, and the AI Factory software layer into a vertically integrated AI infrastructure service for startups, enterprises, researchers, and AI labs.
Differentiator
Problem solved
Functional benefit
Brands
- Voltage Park AI Cloud: AI Cloud product suite offering on-demand and dedicated reserve H100/Blackwell GPU compute, storage, bare-metal access, managed Kubernetes, virtual machines, and observability.
- AI Factory
- Tensordock
Products and services
- Voltage Park AI Cloud Foundational AI compute platform bundling on-demand and reserved NVIDIA HGX H100 and Blackwell (B200, GB200, B300, GB300) GPU clusters with 3.2 TB/s NVIDIA Quantum-2 InfiniBand networking, VAST-powered storage, bare-metal access, managed Kubernetes, virtual machines, and observability across six Tier 3+ U.S. data centers.
- AI Factory Vertically integrated AI software platform (now branded GraphN post-Lightning AI merger) offering ready-to-use Factory Blueprints (Image Generator, Video Generator, Podcast Generator, Agentic RAG, Spreadsheet Analyzer, SAM3 Image and Video Segmentation, Markdown Creator) and hosted foundation models (gpt-oss-120b, Qwen2.5-VL-7B-Instruct, Qwen3-Next-80B-A3B-Instruct) integrated with Voltage Park's 36,000+ owned GPU infrastructure.
- On-Demand GPU (NVIDIA HGX H100) Self-serve, bare-metal NVIDIA HGX H100 clusters available with 15-minute provisioning, in InfiniBand (3,200 Gbps, 8-1,016 HGX H100 GPUs) and Ethernet (1-1,016 HGX H100 GPUs) variants. Pricing starts at $1.99/hr with no ingress, egress, or support fees.
- Dedicated Reserve Long-term (6+ or 12+ month) reserved GPU capacity for sustained AI workloads, configurable from 32 to 8,000+ H100 GPUs per cluster and available on NVIDIA HGX H100, B200, GB200, B300, and GB300 platforms with friendly payment terms and physically isolated, bare-metal performance.
- NVIDIA Blackwell GPUs (B200, GB200, B300, GB300) NVIDIA Blackwell-generation GPU clusters available for long-term reserve: HGX B200 (second-generation Transformer Engine, FP8 for GPT MoE training), HGX GB200 (real-time trillion-parameter LLM inference), HGX B300 (ConnectX-8 networking, NVIDIA Mission Control), and HGX GB300 (Blackwell Ultra + NVL72 for massive-scale AI reasoning inference).
- Storage High-throughput, AI-native storage layer powered by VAST AI OS, delivering consistent throughput across training, inference, and evaluation workloads, with a unified namespace for file, object, and block access, per-tenant isolation, and built-in HIPAA, FINRA, and SEC compliance support.
- Bare-Metal Access Direct hardware access to Voltage Park's NVIDIA HGX H100 fleet with no virtualization overhead, predictable performance, non-blocking InfiniBand fabric, VPC-level isolation, and the ability to bring your own orchestrator (Slurm, Kubernetes, Ray, etc.) for full control of CPU, GPU, and I/O resources.
- Managed Kubernetes Fully managed upstream Kubernetes clusters with pre-installed NVIDIA GPU Operator and MPI Operator, built-in Prometheus and Grafana GPU observability, VAST CSI integration, automated patching, encrypted ETCD snapshots, and compatibility with Terraform, Argo, Helm, Kubeflow, and Ray.
Quantifiable outcome
- ~30% lower cost vs traditional hyperscalers (no ingress/egress/support fees)
- +4 more outcomes
Companies that use Voltage Park
Customer profileNamed customers25 records
Segments5 records
Ideal customer profiles5 records
Voltage Park technology and API
TechnologyTechnology focussed Yes
API detail
- Has API
- Yes
- API docs
- API detail
Core technology
AI maturity
App detail
Integration11 records
AI capability13 records
Feature6 records
Voltage Park partnerships and signals
Strategic signalPartnerships
Ten partnerships are on record, tiered strategic, flagship, core and minor.
- JBK GroupstrategicPartnered with Voltage Park and Matrice.ai to deploy Vision AI factories across Qatar for enhanced automation, extending Voltage Park's reach into the Middle East market.
- Matrice.aistrategicStrategic funding led by Voltage Park to industrialize real-time Vision AI; co-deployment partner with JBK Group for Vision AI factories across Qatar.
- TensorDockstrategicVoltage Park acquired TensorDock in March 2025 to expand GPU cloud services for AI and machine learning; TensorDock now operates under the Voltage Park umbrella.
- NVIDIAflagshipVoltage Park sources its full GPU fleet (HGX H100, H200, B200, GB200, B300, GB300) from NVIDIA and was named a Reference Platform NVIDIA Cloud Partner in 2025. NVIDIA Quantum-2 InfiniBand (3,200 Gbps) also underpins the network fabric.
- Dell TechnologiescoreDell PowerEdge XE9680 servers house Voltage Park's NVIDIA HGX H100 GPUs (8-GPU SXM5, 2x Intel Xeon Platinum 8470, 1TB ECC DDR5 RAM). Dell provides 24/7 hardware support and is featured as a top-tier global partner.
- Penguin SolutionscoreFeatured as a global infrastructure partner helping build one of the world's largest AI clouds alongside Dell, VAST, and NVIDIA.
- VAST DatacoreVAST AI OS provides Voltage Park's unified storage layer (file/object/block) for AI workloads including trillion-parameter model checkpointing; supports HIPAA, FINRA, SEC compliance. Co-hosted joint webinars ("How VAST Data & Supermicro Power Voltage Park's AI Cloud") and joint press release.
- Lightning AIflagshipCompleted a full merger in January 2026 to form an integrated AI-native cloud combining Lightning's software tools (PyTorch Lightning, AI Studio) with Voltage Park's 36,000+ GPU fleet. Combined entity operates under the Lightning AI name at $500M+ ARR and $2.5B valuation, serving 400,000+ developers.
- Amazon Web Services (AWS)coreLightning AI (post-merger) interoperates with AWS — customers can continue using Lightning alongside AWS and burst workloads into Lightning's GPU infrastructure when needed.
- SupermicrominorCo-presented webinars ("How VAST Data & Supermicro Power Voltage Park's AI Cloud for Builders, Startups & Enterprises") and participated in the Supermicro Open Storage Summit 2025.
Scale indicators13 records
Recent moves6 records
Expansion highlights6 records
Voltage Park competitors and assessment
Company assessmentDirect peers
- CoreWeave: Largest pure-play neocloud offering NVIDIA GPU compute (H100, H200, Blackwell) on owned infrastructure with InfiniBand fabric. Directly comparable on product, target customer (AI labs and enterprises), and vertically integrated AI cloud positioning; CoreWeave is the named #1 competitor with Voltage Park at #3.
- Nebius Group: Vertically integrated AI infrastructure company operating owned NVIDIA GPU data centers globally. Directly comparable on owned-hardware neocloud model and frontier AI lab customer base; the named #2 neocloud competitor behind which Voltage Park ranks third.
- Crusoe Energy: Purpose-built GPU cloud provider operating AI data centers with a renewable/stranded-energy focus. Closely comparable on owned GPU infrastructure serving AI labs and startups, with similar sustainability positioning.
- Lambda Labs: GPU cloud and deep learning workstation provider offering NVIDIA H100/B200 clusters to AI startups and enterprises. Comparable on GPU-as-a-service product, target market (AI researchers and labs), and pricing transparency.
Emerging players
- Together AI: AI cloud platform combining GPU infrastructure with open-source model inference and fine-tuning services. Comparable on the AI-native cloud + software stack, especially post-Lightning AI merger; overlaps with Voltage Park's AI Factory blueprints.
- Fireworks AI: AI inference platform operating its own GPU infrastructure for serving open and custom models. Comparable as a Voltage Park customer and as an inference-focused neocloud with overlapping enterprise AI use cases.
Broad incumbents
- Amazon Web Services (EC2 / Bedrock): Hyperscaler offering P5/P6 GPU instances and Bedrock generative AI services. The broadest incumbent competitor Voltage Park explicitly benchmarks against on price and capacity; AWS also interoperates with Lightning AI post-merger.
- Google Cloud (A3 / Vertex AI): Hyperscaler with A3 H100 instances, custom TPU accelerators, and Vertex AI platform. Major incumbent competitor for both GPU compute and managed AI services targeting the same AI lab and enterprise workloads.
- Microsoft Azure: Hyperscaler providing ND H100/B200 instances and Azure AI Foundry services. Competes head-to-head for enterprise AI workloads that Voltage Park targets, especially given Azure's deep OpenAI partnership.
- Oracle Cloud Infrastructure (OCI): Hyperscaler that has built a sizable GPU cloud business (OCI Compute with H100/B200) targeting AI startups and enterprises. Comparable as a broad incumbent alternative to Voltage Park's owned-infrastructure model.
Market position
Strengths5 records
Weaknesses5 records
Competitive moat5 records
Key risks6 records
Key highlights7 records
Customer concentration
Voltage Park social profiles
Digital presenceVoltage Park compliance and trust
Trust signalCompliance3 records
Voltage Park financial estimates
Financial estimateRevenue estimate
Valuation estimate
Voltage Park leadership team
Management profileNumber of profiles
Profiles8 records
Voltage Park subsidiaries and ownership
Company hierarchySubsidiaries1 record
Voltage Park funding detail
Funding detailFunding overview
Funding rounds1 record
Investors
Funding detail is available on the Subscription and Enterprise plan.Contact sales →
Voltage Park M&A and investment
M&A and investmentM&A1 record
Investments1 record
M&A and investment is available on the Subscription and Enterprise plan.Contact sales →
Frequently asked questions about Voltage Park
What does Voltage Park do?
Voltage Park operates an AI cloud platform delivering owned NVIDIA HGX H100 and Blackwell (B200, GB200, B300, GB300) GPU infrastructure on a self-serve on-demand basis starting at $1.99/hr and via long-term Dedicated Reserve contracts. The platform bundles compute, 3,200 Gbps InfiniBand fabric, VAST-powered unified storage, managed Kubernetes, bare-metal access, virtual machines, and the AI Factory software layer into a vertically integrated AI infrastructure service for startups, enterprises, researchers, and AI labs.
Is Voltage Park a public or private company?
Voltage Park is a private company. It is classified as venture growth investor backed and is currently acquired.
When was Voltage Park founded?
Voltage Park was founded in 2023. It employs 101 to 250 people.
Where is Voltage Park based?
Voltage Park is headquartered in Berkeley, United States, in the North America region.
How does Voltage Park make money?
Four revenue lines are on record. On-Demand GPU Compute is the primary driver. The others are dedicated Reserve (Long-Term), AI Factory Platform and managed Kubernetes / Virtual Machines / Storage add-ons.
Who are Voltage Park's main competitors?
Direct peers on record are CoreWeave, Nebius Group, Crusoe Energy and Lambda Labs. Emerging players are Together AI and Fireworks AI. Broad incumbents are Amazon Web Services (EC2 / Bedrock), Google Cloud (A3 / Vertex AI), Microsoft Azure and Oracle Cloud Infrastructure (OCI).
Does Voltage Park have an API?
Yes. Voltage Park offers a programmatic API for deploying and managing GPU compute resources. Per the Virtual Machines page, deployments are "Accessible through the Voltage Park web console or API, with a guided deployment flow that makes it easy to spin up infrastructure even for first-time users." The On-Demand GPU cloud and its API is described as offering "the fastest click-to-production platform on the planet" with capabilities for self-serve provisioning, scaling GPU clusters, and managing bare-metal HGX H100 deployments. Developer documentation is at docs.voltagepark.com.
What industry is Voltage Park in?
Voltage Park's product category is AI Cloud Infrastructure (GPU-as-a-Service). Its primary akta.pro industry code is HDAAAAAB, AI Server Systems & HGX/Accelerator Platforms, with a secondary code of HDAAAAAK, AI Compute Cloud & GPU-as-a-Service. Its NAICS code is 518 and its SIC code is 7377.