GPU-powered Servers
90 cloud providers offer GPU-powered Servers services.
Providers
Alphabetically by provider name.
-
Lium
Our sponsorGPU marketplace billed by the minute with root access
-
3NV
Cloud VMs and bare metal hosted in Tallinn
-
AceCloud
GPU cloud and Kubernetes from Mumbai, Noida and Atlanta
-
AIEEV
GPU inference cloud on RTX-class nodes in South Korea
-
Akamai Cloud
VPS and managed Kubernetes on Akamai's CDN network
-
Alibaba Cloud
Hyperscaler with most regions in mainland China and Asia
-
Amazon Web Services
Hyperscaler with the largest managed service catalogue
-
Beam
Serverless GPUs, sandboxes and on-demand machines
-
Bentaus
Custom AI infrastructure built around the power grid
-
Beyond.pl
DGX B200 nodes and managed Kubernetes from Poznań
-
BreezeHost
VPS, dedicated servers and RTX 5090 instances
-
Canopy Wave
HGX GPU nodes and an open-model API from Iceland
-
Cerebrium
Serverless GPUs for real-time inference that scale to zero
-
Cirrascale
Reserved Nvidia, AMD and Cerebras accelerators in US datacenters
-
Civo
Managed Kubernetes and GPU cloud with free egress
-
CloudFerro
OpenStack cloud in Poland and Germany with Earth observation data
-
Cloudzy
Hourly-billed VPS and GPUs with large transfer allowances
-
Contabo
Budget VPS with high core and RAM counts
-
CoreWeave
Kubernetes-native GPU cloud on bare-metal nodes
-
Crusoe
GPU cloud in datacenters it builds and powers itself
-
CUDO Compute
Single-tenant GPU clusters built and run on contract
-
Cyfuture AI
Nvidia, AMD and Intel GPU cloud hosted in India
-
Database Mart
Windows and Linux VPS with consumer and datacenter GPUs
-
Daytona
Sandbox infrastructure for running AI-generated code
-
Denvr
Canadian AI cloud
-
DigitalOcean
Developer cloud with VPS, managed services and GPU servers
-
EcoHash
RTX PRO 6000 GPU cloud with an open-model API
-
Elastx
Swedish cloud with managed databases and GPUs
-
EmpirioLabs AI
One API for text, image, video and audio models
-
Enverge
GPU cloud on renewable power, billed per minute
-
Exoscale
Cloud under Swiss jurisdiction with zones across Central Europe
-
Fal.ai
Serverless inference for generative image, video and audio models
-
Fluidstack
Managed GPU clusters for AI labs, contracted not self-serve
-
Gcore
CDN and edge cloud with VPS and GPU rentals
-
Geodd
Model inference on serverless endpoints and dedicated GPUs
-
Google Cloud
Hyperscaler built around BigQuery, GKE and TPUs
-
GPUaaS
Wholesale GPU cloud sourced from a partner network
-
GPU.ai
GPU aggregator that routes each launch to the cheapest cloud
-
GPUhub
RTX-class GPU containers in Singapore with free egress
-
Green AI Cloud
Reserved H100 and B200 clusters from Sweden
-
Hetzner
Low-price VPS and dedicated servers with 20 TB of transfer
-
Hexabyte
VPS and object storage with no egress fees
-
HIGHRESO
GPU cloud from its own datacenters in Japan
-
Hikube
Kubernetes-native cloud replicated across three Swiss datacenters
-
Hostiko
VPS and RTX 5090 servers in Ukraine, Poland and Sweden
-
Hostinger
Web hosting, prepaid VPS and hourly GPU rentals
-
Hot Aisle
Self-serve AMD MI300X VMs and bare metal nodes
-
Hugging Face
Model and dataset hub with managed GPU compute
-
HyperAI
GPU containers and notebooks sold on prepaid terms
-
Hyperbolic
GPU cloud sourced from partner datacenters, with managed inference
-
Hyperstack
GPU cloud in Canada, Norway and the US
-
Impossible Cloud
Bare metal GPU servers and S3-compatible object storage
-
Jarvislabs
GPU cloud in India and Finland billed by the minute
-
Koyeb
Serverless microVMs for APIs, inference and databases
-
Lambda Labs
On-demand GPUs and training clusters in owned datacenters
-
Latitude.sh
API-provisioned bare metal and GPUs on five continents
-
Leafcloud
OpenStack cloud in the Netherlands that reuses server heat
-
Leaseweb
Dedicated servers, cloud and CDN on four continents
-
Lyceum
EU-sovereign GPU cloud for training and inference
-
Massed Compute
Nvidia GPU cloud on owned hardware for AI and VFX
-
Microsoft Azure
Hyperscaler tied to Microsoft 365 and OpenAI models
-
Nebius
GPU cloud with managed Kubernetes, Slurm and hosted models
-
Nova Cloud
RTX-class GPU cloud from its own Canadian datacenter
-
Novita
GPU cloud with spot pricing and an open-model API
-
Oblivus
GPU cloud with per-GPU hourly pricing and free bandwidth
-
Oracle Cloud
Hyperscaler with 10 TB of free egress and Oracle Database
-
OVHcloud
VPS, bare metal and managed services with egress included
-
Packet·ai
On-demand Nvidia GPUs with a cheaper shared Dynamic tier
-
Paperspace
GPU cloud with Gradient notebooks, part of DigitalOcean
-
Replicate
Model API with a community library, billed per second
-
Runcrate
GPU cloud with per-second billing and an inference API
-
Runpod
GPU cloud with pods, serverless endpoints and multi-node clusters
-
Salad
Distributed GPU cloud
-
Scaleway
EU-only cloud with managed services, serverless and GPUs
-
Sesterce
GPU cloud with Nvidia, AMD and Intel accelerators
-
Shadeform
One console and API to launch GPUs across other clouds
-
Spheron
GPU marketplace sourced from vetted datacenter partners
-
SwissGPU
Bare metal RTX servers in Switzerland, billed per minute
-
TensorDock
GPU marketplace with virtualized and bare metal servers
-
TensorWave
AMD Instinct GPU cloud on enterprise contracts
-
TheAI Cloud
Bare metal B300 and B200 nodes in the US
-
Theta EdgeCloud
Datacenter GPUs and a community fleet of consumer cards
-
Thunder Compute
GPU instances billed per minute
-
Together AI
Open-model inference API with GPU clusters for training
-
UpCloud
Cloud with MaxIOPS storage and datacenters on four continents
-
Utho
Indian-hosted cloud with managed Kubernetes and databases
-
Vast.ai
GPU marketplace with on-demand, interruptible and reserved rentals
-
Verda
GPU cloud from owned datacenters in Finland and Iceland
-
Vultr
VPS, bare metal and GPUs on six continents
-
Zenlayer
Edge cloud with VPS and GPUs across emerging markets