Lightning AI
AI cloud platform with owned GPU infrastructure
Lightning AI provides a full-stack platform for training and deploying artificial intelligence models. Formed in 2026 by merging with Voltage Park, it operates over 36,000 Nvidia GPUs across seven US data centres, offering guaranteed access to compute resources without long-term contracts.
- Operates more than 36,000 Nvidia GPUs
- Merged with Voltage Park in 2026
- Raised $108 million in venture funding
- PyTorch Lightning has 470M+ downloads
- Hardware bought with no debt or leasing
What it is
Lightning AI is a computing provider that combines software frameworks with physical hardware infrastructure. It allows developers to code, train, and deploy models through a browser-based interface or via secure shell access. The company owns its GPU fleet rather than renting capacity from other providers. This setup aims to give users direct control over their AI development stack while ensuring hardware availability. The platform supports both individual researchers and large enterprise teams needing scalable resources for complex machine learning tasks.
How it came about
The origins of the company lie in PyTorch Lightning, an open-source framework created by William Falcon in 2015 during his time at Columbia University. He released it publicly in 2019 while completing his PhD at New York University and working with Facebook AI Research. In 2023, the team launched Lightning AI Studio, a cloud service for model development. That same year, Voltage Park was established with backing from the Navigation Fund to purchase GPU hardware outright. In 2026, Lightning AI and Voltage Park merged into a single entity, combining the software tools with the physical compute infrastructure.
The compute
The infrastructure relies on a fleet of more than 36,000 Nvidia graphics processing units. These accelerators are housed in Tier 3+ data centres and are available for on-demand use or large-scale training clusters. The system supports parallel experiments and interactive coding environments. Users can run workloads using Kubernetes or Slurm schedulers. The platform emphasizes guaranteed availability and self-serve access, allowing teams to scale from development to testing without long-term contractual commitments. Encryption is applied to data both at rest and during transmission.
Where the GPUs are
Compute resources are located in seven data centres across the United States. These facilities are classified as Tier 3+ or higher, indicating robust redundancy and reliability standards. By owning the hardware directly, the company manages the entire physical layer of its service. This approach avoids reliance on third-party cloud rentals for its primary GPU inventory. The distributed nature of these sites helps provide capacity for various workload sizes, from small debugging sessions to massive parallel training runs.
Customers and deals
The platform serves over 450,000 developers and AI teams globally. Notable users include Infineon, where researchers reported accelerating their work significantly, and NNAISENSE, which used the system for market simulations. Other clients include Slalom, Inworld, and Reticular AI. Academic institutions such as Columbia University also utilise the service for research purposes. The user base spans industries requiring specialized model training, data ingestion, and edge AI deployment, benefiting from reduced setup times and consistent working environments.
Money and ownership
Lightning AI has raised $108 million from venture capital firms to support its growth and infrastructure expansion. The hardware arm, formerly Voltage Park, purchased its equipment without taking on debt or using leasing arrangements. This financial strategy aimed to ensure long-term asset ownership and stability. The company operates on a usage-based pricing model, with rates starting from $4.50 per hour for GPU access. There are no mandatory long-term contracts for standard services, allowing for flexible spending based on actual compute consumption.
Its place in AI
As the creator of PyTorch Lightning, which has seen over 470 million downloads, the company holds a significant position in the open-source AI community. The merger with Voltage Park represents a shift towards vertical integration in the neocloud sector, where software providers acquire physical hardware assets. This model challenges traditional hyperscalers by offering dedicated, owned infrastructure. The platform supports the training of specialized models that often outperform general-purpose alternatives, contributing to advancements in fields like cancer research and autonomous systems.
Questions people ask about Lightning AI
Who founded Lightning AI?
William Falcon began developing the core framework, PyTorch Lightning, in 2015. The current company formed in 2026 through a merger with Voltage Park, which was backed by the Navigation Fund.
How many GPUs does Lightning AI have?
The company operates a fleet of more than 36,000 Nvidia GPUs. These are located across seven Tier 3+ data centres in the United States and are owned outright by the company.
What is the cost of using the service?
Pricing starts from $4.50 per hour for GPU access. The service operates on a self-serve basis with no long-term contracts required, allowing users to pay only for the compute time they consume.
Good reading and sources
There is no English Wikipedia article on Lightning AI yet, so the facts on this page come from the sources listed above: lightning.ai/about, lightning.ai/. Some are the subject's own pages, so its claims about itself are reported as its claims. EveryNeocloud wrote the text in its own words, with AI assistance, and compares it with the sources to make sure no passage is copied.
It may contain mistakes, and figures change over time. Nothing here is investment or buying advice. If something is wrong, tell us. Last refreshed 2026-11-14.