Partner Directory
Cloud GPU Providers
Below is a list of cloud GPU providers, along with some example configurations they offer. For each provider, we provide a brief overview and a few example instance types, along with their specifications and hourly pricing. Use these tables to compare offerings across different platforms. (Prices are in USD per hour. Configurations and prices are examples and may vary by region; please refer to the provider’s site for the most current details.)
Premier Partner
Vultr – A USA-based global cloud provider offering GPU instances (A16, A40, L40S, GH200) with both single and multi-GPU configurations across global regions.
Alibaba Cloud – A cloud platform with a strong presence in Asia (headquartered in Singapore). It offers a range of GPU instances on its global infrastructure, suitable for AI, rendering, and HPC workloads, with competitive pricing especially in Asian regions.
Amazon Web Services (AWS) – The largest global cloud provider (USA). AWS offers a broad array of GPU instances integrated into its cloud ecosystem. These range from cost-effective instances for light workloads to multi-GPU clusters for heavy AI training. AWS’s GPU instances (the G- and P- series) benefit from the robust AWS infrastructure and services.
Build AI – A USA-based startup focusing on affordable GPUs for machine learning training. Build AI provides both on-demand and spot-priced GPU servers to help save costs, offering modern NVIDIA A100 and H100 GPUs. This platform is ideal for training deep learning models on a budget, with lower prices for interruptible (spot) instances.
Civo – A UK-based cloud platform known for simplicity and Kubernetes services (K3s). Civo has introduced GPU instances for AI and machine learning workloads. It provides configurations from single GPU VMs to multi-GPU setups, with A100 and L40S GPUs available. Civo’s GPU plans are straightforward, making it easy to deploy and scale GPU workloads in the cloud.
Contabo – A Germany-based provider known for low-cost servers. Contabo offers a couple of high-end GPU configurations on dedicated hardware. These instances pack multiple GPUs (L40S or H100) and provide bare-metal performance at competitive hourly rates (note: Contabo often charges a one-time setup fee for dedicated resources).
CoreWeave – A USA-based GPU cloud built specifically for AI workloads. CoreWeave offers a massive fleet of NVIDIA GPUs with preemptible pricing, low-latency networking, and rapid provisioning. It’s a popular choice among AI startups and enterprise labs running large-scale training or inference jobs.
Crusoe – A USA-based cloud provider focused on sustainable computing (it powers its cloud with stranded energy such as flare gas). Crusoe offers GPU instances for AI and HPC with an emphasis on cost efficiency and green energy. Their instances can be scaled up to 10× the base configuration for larger workloads.
CUDO Compute – A UK-based high-performance GPU cloud provider. CUDO offers a variety of NVIDIA (and some AMD) GPU options, including cutting-edge models like the NVIDIA H100 and even experimental GPUs, aimed at AI researchers and enterprises needing serious compute power. They emphasize performance and have data centers in multiple regions.
DataCrunch – A Finland-based provider that deploys the latest NVIDIA GPUs in European data centers. DataCrunch caters to researchers and startups with high-end hardware (like NVIDIA A100, H100, and even H200 GPUs) at competitive prices. It’s a great choice for those needing powerful GPUs in Europe without using a hyperscaler.
DigitalOcean – A USA-based developer-friendly cloud platform known for simplicity. DigitalOcean has recently introduced GPU “Droplets,” currently offering powerful NVIDIA H100 instances for compute-intensive tasks. This allows developers to integrate high-end GPUs into DigitalOcean’s easy-to-use cloud environment.
Exoscale – A Switzerland-based European cloud provider offering GPU instances suitable for machine learning, rendering, and compute-heavy tasks. Known for GDPR compliance and regional datacenters.
FluidStack – A UK-based cloud platform specialized in AI training and inference. FluidStack aggregates underutilized or distributed GPUs to provide affordable compute. It offers a range of NVIDIA GPU types, from consumer-grade (RTX/A-series) to data center GPUs (A100/H100), allowing customers to train models at lower cost.
Fly.io – A USA-based application hosting platform. Primarily known for deploying web apps to global regions, Fly.io also offers GPU-backed machines in select regions. These GPU instances allow running AI inference or smaller training jobs at the edge, benefiting from Fly.io’s distributed infrastructure.
Google Cloud – A USA-based global cloud platform by Google. Google Cloud’s GPU offerings include the latest NVIDIA chips (L4, A100, etc.), integrated with Google’s AI and data services. It offers flexible instance types; for example, the G2 series with L4 GPUs for inference, and the A2 series with A100 GPUs for heavy training. Users can scale from 1 GPU up to 16 GPUs in a single VM.
Green AI Cloud – A Sweden-based provider focused on sustainability. It offers high-performance GPU servers powered by renewable energy. Green AI Cloud specializes in large multi-GPU configurations (for example, servers with up to 8× NVIDIA H100 or H200 GPUs) aimed at organizations that need extreme performance with a smaller carbon footprint.
Hetzner – A Germany-based provider known for highly cost-effective hosting. Hetzner’s GPUs are offered as add-ons to its dedicated servers rather than on-demand VMs. This provides bare-metal performance at low hourly rates (note: a one-time setup fee applies). It’s a great option for steady, long-running GPU workloads where cost savings are paramount.
Hyperstack – A UK-based cloud specializing in affordable GPUs. Hyperstack offers a wide range of NVIDIA GPU configurations, from single low-cost GPUs (starting at just $0.15/h) to multi-GPU setups. It’s known for having some of the lowest prices in the market, making it popular for budget-conscious AI practitioners and hobbyists.
Koyeb – A France-based developer-centric cloud platform offering container-based GPU compute services. Designed for deployment automation and scalable GPU-backed inference.
Lambda Labs – A USA-based GPU cloud provider and hardware vendor specializing in deep learning. Lambda Labs offers on-demand and reserved GPU servers with a wide range of configurations, from single-GPU instances (including GeForce/RTX cards) to multi-GPU powerhouse servers (8× A100 or H100). It’s a go-to for many ML engineers due to its focus on AI (and integrations like pre-built ML environments).
Linode – A USA-based cloud provider (part of Akamai) known for simple and affordable VPS. Linode introduced GPU instances (“Accelerators”) that utilize NVIDIA RTX 6000 GPUs.
Massed Compute – A USA-based GPU infrastructure provider offering on-demand access to GPUs. Offers a variety of NVIDIA GPU options with flexible scaling.
Microsoft Azure – A USA-based global cloud platform by Microsoft. Offers GPU-enabled VMs (N-series) using NVIDIA T4, A100, H100, etc. for AI and HPC workloads.
Nebius – A Netherlands-based AI cloud platform focused on high-end GPU compute including H100 and H200 for European workloads.
Oblivus – A UK-based cloud GPU provider offering simple, low-cost access to GPUs like A4000, A5000, A100, and H100 for AI workloads.
Oracle Cloud – A USA-based enterprise cloud provider offering AMD MI300X GPU servers for massive AI and HPC compute workloads.
OVHcloud – A France-based global cloud provider with competitively priced GPU instances such as V100, L4, and H100 across European data centers.
Replicate – A USA-based service for running ML models via API. Offers GPU-backed infrastructure for pay-as-you-go inference and fine-tuning.
RunPod – A USA-based provider that lets you spin up GPU containers in seconds across 30+ regions. It offers flexible, on-demand GPU cloud instances with auto-scaling and free data transfers. RunPod’s platform is popular for its ease of use and low-cost spot pricing for AI workloads.
Scaleway – A France-based cloud provider offering a range of NVIDIA GPU options (L4, L40S, H100, etc.) for AI, video rendering, and HPC, with pricing optimized for European customers.
Sesterce – A France-based cloud GPU marketplace originally known for mining hardware. Offers a wide range of low-cost and modern GPU compute options including A4000, A100, and MI300X.
TensorWave – A USA-based cloud provider offering AMD Instinct GPU compute (MI300X) targeted at foundation model training. Pricing is quote-based.
Vultr – A USA-based global cloud provider offering GPU instances (A16, A40, L40S, GH200) with both single and multi-GPU configurations across global regions.