Search accelerators, providers, regions or GPU models GPU model GPU model AMD Instinct MI300X 192GB AMD Instinct MI325X 256GB AMD Instinct MI350X 288GB AMD Instinct MI355X 288GB Intel Gaudi 2 96GB Intel Gaudi 3 128GB NVIDIA A10 24GB NVIDIA A100 40GB — provider does not specify PCIe/SXM NVIDIA A100 40GB PCIe NVIDIA A100 40GB SXM NVIDIA A100 80GB — provider does not specify PCIe/SXM NVIDIA A100 80GB PCIe NVIDIA A100 80GB SXM NVIDIA A100 MIG 20GB slice NVIDIA A100 NVLink 80GB — Hyperstack flavor NVIDIA A100 Provider model only - memory and form factor unspecified NVIDIA A2 16GB NVIDIA A30 24GB NVIDIA A40 48GB NVIDIA B200 SXM 180GB NVIDIA B200 SXM 192GB — GMI Cloud identity NVIDIA B200 SXM 192GB — Hyperstack flavor NVIDIA B300 SXM NVIDIA GB200 Grace Blackwell Superchip GPU NVIDIA GB300 NVL72 / 288GB per GPU NVIDIA GeForce GTX 1080 Ti 11GB - board vendor unspecified NVIDIA GeForce RTX 3060 Reference specification NVIDIA GeForce RTX 3070 Reference specification NVIDIA GeForce RTX 3070 Ti Reference specification NVIDIA GeForce RTX 3080 10GB reference specification NVIDIA GeForce RTX 3080 Ti Reference specification NVIDIA GeForce RTX 3090 Founders / reference NVIDIA GeForce RTX 3090 Ti Reference specification NVIDIA GeForce RTX 4060 Ti SaladCloud public pricing class NVIDIA GeForce RTX 4070 SaladCloud public pricing class NVIDIA GeForce RTX 4070 Ti Reference specification NVIDIA GeForce RTX 4070 Ti SUPER Reference specification NVIDIA GeForce RTX 4080 Founders / reference NVIDIA GeForce RTX 4090 Founders / reference NVIDIA GeForce RTX 5060 SaladCloud public pricing class NVIDIA GeForce RTX 5060 Ti SaladCloud public pricing class NVIDIA GeForce RTX 5070 Reference specification NVIDIA GeForce RTX 5070 Ti Reference specification NVIDIA GeForce RTX 5080 Founders / reference NVIDIA GeForce RTX 5090 Founders / reference NVIDIA GeForce RTX 5090 Laptop SaladCloud public pricing class NVIDIA GH200 96GB HBM3 NVIDIA H100 80GB — provider does not specify PCIe/SXM NVIDIA H100 NVL 94GB NVIDIA H100 NVLink 80GB — Hyperstack flavor NVIDIA H100 PCIe 80GB NVIDIA H100 Provider model only - memory and form factor unspecified NVIDIA H100 SXM NVIDIA H200 141GB — provider does not specify SXM/NVL NVIDIA H200 NVL NVIDIA H200 SXM NVIDIA L4 24GB NVIDIA L40 48GB NVIDIA L40S 48GB NVIDIA Quadro M4000 8GB NVIDIA Quadro P4000 8GB NVIDIA Quadro P5000 16GB NVIDIA Quadro P6000 24GB NVIDIA Quadro RTX 4000 8GB NVIDIA Quadro RTX 5000 16GB NVIDIA Quadro RTX 6000 24GB NVIDIA RTX 4000 Ada 20GB NVIDIA RTX 6000 Ada 48GB NVIDIA RTX A4000 16GB NVIDIA RTX A5000 24GB NVIDIA RTX A6000 48GB NVIDIA RTX PRO 4500 Blackwell Server Edition 32GB NVIDIA RTX PRO 6000 Blackwell 96GB - provider does not specify edition NVIDIA RTX PRO 6000 Blackwell Max-Q Workstation Edition NVIDIA RTX PRO 6000 Blackwell Server Edition NVIDIA T4 16GB NVIDIA V100 16GB PCIe NVIDIA V100 32GB PCIe NVIDIA V100 SXM2 16GB — TensorDock provider label NVIDIA V100 SXM3 32GB — TensorDock provider label NVIDIA V100S 32GB PCIe Unknown VR200 GPU memory GPU memory 8 GB 10 GB 11 GB 12 GB 16 GB 20 GB 24 GB 32 GB 40 GB 48 GB 80 GB 94 GB 96 GB 128 GB 141 GB 180 GB 186 GB 192 GB 256 GB 288 GB Region Region Amsterdam, Netherlands Atlanta, United States Busan, South Korea California, United States Canberra, Australia Cardiff, United Kingdom Chennai, India Deployment dependent, Multiple / Unspecified Dubai, United Arab Emirates Europe, Multiple / Unspecified Falkenstein, Germany Frankfurt, Germany Gävle, Sweden Geneva, Switzerland Global / region-aware, Multiple / Unspecified Grand Rapids, Michigan, United States Hanoi, Vietnam Helsinki, Finland Houston, Texas, United States Iceland, Iceland Illinois, United States Indonesia (location unspecified), Indonesia Iowa, United States Ireland, Ireland Ishikari, Japan Ishikawa, Japan Israel, Israel Jakarta, Indonesia Japan (location unspecified), Japan Johannesburg, South Africa Kansas City, United States Kuala Lumpur, Malaysia London, United Kingdom Madrid, Spain Malaysia (location unspecified), Malaysia Memphis, Tennessee, United States Milan, Italy Netherlands, Netherlands New South Wales, Australia New York Metro, United States Noida, India North America, Multiple / Unspecified Northern Virginia, United States Osaka, Japan Paris / availability-zone dependent, France Paris, France Phoenix, United States Pune, India Region dependent, Multiple / Unspecified São Paulo State, Brazil Seoul, South Korea Silicon Valley, United States Singapore, Singapore South Korea (location unspecified), South Korea Taiwan (location unspecified), Taiwan Texas, United States Thailand (location unspecified), Thailand Tokyo, Japan Toronto, Canada United States, United States Unspecified, Multiple / Unspecified Vienna, Austria Vietnam (location unspecified), Vietnam Virginia, United States Warsaw, Poland Washington State, United States Wyoming, United States Zagreb, Croatia Zurich, Switzerland Provider Provider Ace Cloud AWS Biznet Gio Civo CMC Cloud CoreWeave Crusoe Cloud CUDO Compute Denvr Data DigitalOcean E2E Networks Elice Cloud Exoscale Float16 Cloud FPT AI Factory Gcore GMI Cloud GMO GPU Cloud Google Cloud GreenNode Hetzner Highreso GPUSOROBAN Hot Aisle Hyperstack IP ServerOne JarvisLabs Kakao Cloud Lambda Latitude.sh Leaseweb Massed Compute Microsoft Azure NAVER Cloud Platform Nebius Novita AI OVHcloud Paperspace Runpod Sakura Cloud SaladCloud Scaleway SIAM.AI Cloud Singtel RE:AI Sunteco Cloud Taiwan AI Cloud TensorDock Thunder Compute Vast.ai Verda Voltage Park Yotta Shakti Cloud Network Network 10 Gbps 10 Gbps public / 25 Gbps private 2.5 Gbps 20 Gbps 400Gbps InfiniBand per GPU + Azure VNet Azure VM network (documented maximum) EFA EFAv2 Ethernet;external/public path not specified Google AI network fabric Google VPC GPUDirect-TCPXO InfiniBand NVLink Public network: 5 Gbps baseline, up to 15 Gbps burst; private network: 2.5 Gbps Term Term 1 year+ 12-month reserved 2 weeks–1 year Capacity Block Committed Dedicated monthly Flex-start Inference platform On-Demand On-demand cluster Pre-order Preemptible Private Quote Reserved Reserved / contractual Serverless Serverless spot Spot