AI GPU Cloud Companies in 2026
AI GPU cloud companies rent graphics processing units (GPUs) over the internet so teams can train, fine-tune and run AI models without buying hardware. The category spans large GPU-specialized clouds, developer-focused on-demand and serverless GPU platforms, general-purpose clouds with GPU instances, and marketplaces that resell or aggregate capacity from many providers.
AI GPU Cloud Companies to know
Companies were chosen for how much GPU rental they actually do (fleet size, number of developers, revenue reported in the press), recent funding or public-market status, and how well they fit this page. That means renting GPU compute on demand, on reservation, serverlessly or through a marketplace. CoreWeave and Lambda also appear on the Neocloud page. They are listed here too because they are among the biggest GPU renters overall. The rest were picked for breadth: serverless GPU platforms (Modal, Runpod), developer clouds that offer GPUs (Vultr, DigitalOcean), and marketplaces or exchanges (Vast.ai, SF Compute, Hyperbolic, Salad). Order reflects overall notability as of October 2026.
CoreWeave
Publicly listed GPU cloud built for AI training and inference at large scale
CoreWeave operates GPU cloud infrastructure for AI and sells it to large customers, including hyperscalers and AI labs. It started in 2017 as a cryptocurrency miner called Atlantic Crypto. In 2025 it acquired Weights & Biases, OpenPipe and Monolith AI, which added software for model development; a proposed $9B acquisition of Core Scientific was rejected by Core Scientific shareholders in October 2025.
- Headquarters
- Livingston, New Jersey, USA
Lambda
GPU cloud offering on-demand instances, 1-Click Clusters and large superclusters
Lambda rents NVIDIA GPU capacity in several forms: on-demand instances, 1-Click Clusters, and very large Superclusters. It also distributes deep learning software (Lambda Stack). Customers range from researchers and startups to frontier labs and hyperscalers.
- Headquarters
- San Jose, California, USA
Modal
Serverless GPU platform for inference, training, batch jobs and sandboxes
Modal runs a serverless cloud where developers define Python functions that scale from zero up to thousands of GPUs, and they pay only for what they use. Its products cover model inference, fine-tuning and training, batch processing, notebooks, and isolated sandboxes for coding agents. It sources GPU capacity from multiple cloud providers.
- Headquarters
- New York, New York, USA
Runpod
Developer GPU cloud with on-demand Pods, serverless endpoints and clusters
Runpod rents GPUs to developers through on-demand Pods, autoscaling serverless endpoints, public model endpoints and multi-node clusters. Workloads include experimentation, fine-tuning, inference and training. The company says it serves more than one million developers.
- Headquarters
- San Francisco, California, USA
Vultr
Independent cloud provider offering GPU instances alongside compute and storage
Vultr is a general-purpose cloud provider. Alongside compute and storage it offers cloud GPU instances and bare-metal servers, and rents both NVIDIA and AMD GPUs for AI workloads. AMD co-led its December 2024 equity round, and it added $329M of credit financing in June 2025.
- Headquarters
- West Palm Beach, Florida, USA
DigitalOcean
Public developer cloud selling GPU Droplets, bare-metal GPUs and serverless inference
DigitalOcean is a developer-focused cloud provider listed on the NYSE. It sells GPU Droplets (GPU virtual machines) and bare-metal GPUs using NVIDIA H100, H200, B300 and AMD Instinct MI300X/MI325X/MI350X chips, as well as a serverless inference platform. Paperspace, a GPU cloud it acquired, is one of its brands.
- Headquarters
- Broomfield, Colorado, USA
Vast.ai
Two-sided GPU rental marketplace connecting independent hosts with AI developers
Vast.ai runs a marketplace where data centers and individual hardware owners list GPUs, and developers rent them for training and inference, at hourly rates the company markets as lower than the major clouds. Besides marketplace rentals it offers clusters, serverless inference and a model library. The company reported more than 17,000 GPUs on the platform as of 2024.
- Headquarters
- Los Angeles, California, USA
SF Compute
GPU compute exchange where buyers reserve cluster time and resell unused capacity
SF Compute (San Francisco Compute Company) sells GPU capacity as VMs, bare metal and managed Slurm clusters. Customers reserve capacity for fixed periods and can resell time they don't use to other buyers through the company's order book. The company began as a way to sublease a GPU cluster an AI lab had bought.
- Headquarters
- San Francisco, California, USA
Hyperbolic
Open-access GPU cloud with on-demand GPUs, reserved clusters and supplier network
Hyperbolic rents NVIDIA H100, H200 and B200 GPUs by the hour without a sales process. It also offers reserved clusters on commitments shorter than a year and private cloud capacity through its supplier network. Its founders describe the aim as putting underused GPU capacity to work for developers.
- Headquarters
- San Francisco, California, USA
Salad
Distributed GPU cloud built on idle consumer PCs for containerized AI workloads
Salad Technologies runs a distributed cloud built from idle gaming PCs and workstations whose owners contribute GPU time. Businesses deploy containerized inference and batch workloads on the network through the Salad Container Engine. Salad also offers dedicated and private cloud tiers, an AI gateway and a transcription API.
- Headquarters
- Salt Lake City, Utah, USA
AI GPU Cloud Companies at a glance
| Company | Headquarters | Focus |
|---|---|---|
| CoreWeave | Livingston, New Jersey, USA | gpu-cloud, neocloud |
| Lambda | San Jose, California, USA | gpu-cloud, neocloud |
| Modal | New York, New York, USA | serverless-gpu, inference |
| Runpod | San Francisco, California, USA | gpu-cloud, serverless-gpu |
| Vultr | West Palm Beach, Florida, USA | gpu-cloud, cloud-provider |
| DigitalOcean | Broomfield, Colorado, USA | gpu-cloud, public-company |
| Vast.ai | Los Angeles, California, USA | gpu-marketplace, on-demand |
| SF Compute | San Francisco, California, USA | gpu-marketplace, compute-exchange |
| Hyperbolic | San Francisco, California, USA | gpu-cloud, gpu-marketplace |
| Salad | Salt Lake City, Utah, USA | distributed-gpu, consumer-gpus |
Frequently asked questions
Which companies are the biggest AI GPU cloud providers?
By scale, CoreWeave (Nasdaq: CRWV, public since March 2025) and Lambda (over $1.5B Series E in November 2025, with a raise of up to $4B reported in October 2026) are the largest GPU-specialized clouds among notable providers. Among developer-focused platforms, Modal reported about $300M in annualized revenue in May 2026, and Runpod raised $100M in June 2026.
What is a serverless GPU provider?
A serverless GPU provider runs your code or model on GPUs only while requests are coming in. It scales down to zero when idle and bills by usage, so you don't rent a whole machine. Modal (serverless functions, inference, sandboxes), Runpod (Serverless Endpoints) and Vast.ai (Serverless) all offer this model.
How do GPU marketplaces differ from GPU clouds?
A GPU cloud such as CoreWeave or Lambda runs its own fleet and rents it directly. A marketplace aggregates capacity from many suppliers. Vast.ai lists GPUs from independent hosts and data centers. SF Compute lets buyers reserve cluster time and resell what they don't use. Salad pools idle consumer GPUs. Hyperbolic adds capacity from a supplier network on top of its on-demand GPUs.
How do AI GPU cloud companies differ from neoclouds?
Neoclouds are GPU-specialized clouds that own and operate large fleets and data centers. CoreWeave and Lambda are examples, and both are also GPU clouds. The GPU cloud category is broader. It also includes serverless platforms like Modal, developer clouds with GPU instances like DigitalOcean and Vultr, and marketplaces like Vast.ai and SF Compute.
Which GPU clouds offer AMD GPUs?
DigitalOcean offers AMD Instinct MI300X, MI325X and MI350X GPUs alongside NVIDIA H100, H200 and B300. Vultr, whose December 2024 $333M round AMD co-led, also offers AMD GPUs.
More AI company directories
- AI Model Hosting CompaniesHugging Face, Baseten, Modal and more
- Healthcare Intelligence CompaniesIQVIA, Definitive Healthcare, Komodo Health and more
- AI Observability CompaniesArize AI, Braintrust, Langfuse and more
- AI Agent Infrastructure CompaniesLangChain, CrewAI, Browserbase and more
- AI Search CompaniesPerplexity, Glean, Exa and more
- Neocloud CompaniesCoreWeave, Nebius, Crusoe and more
About this directory: OpenCurious curates this list for researchers, founders, operators and buyers. Company descriptions and headquarters are based on publicly reported information as of and change often; figures a company reports about itself are its own claims. This is an editorial selection, not a paid ranking: no company paid to be listed, and companies appear in no particular order. Company names and logos are trademarks of their respective owners and are used only to identify each company. See our editorial policy. To suggest a company or a correction, email hello@opencurious.com.
The OpenCurious Newsletter
Stay curious. Get the best of the open web.
New directories, researcher tools, and open-source finds — straight to your inbox. No spam, ever.
- New directories
- Researcher tools
- Open-source finds
Occasional emails from OpenCurious. Unsubscribe anytime. Privacy Policy