{"catalog":{"title":"Cloud GPU rental prices: H100, H200, B200 per GPU-hour","path":"/gpu-prices","updatedAt":"2026-10-05T05:12:28.963Z","fields":[{"id":"provider","label":"Provider","type":"select","options":["AWS","Google Cloud","Microsoft Azure","Oracle OCI","CoreWeave","Lambda","Nebius","Crusoe","RunPod","DigitalOcean","Vultr","Hyperstack"]},{"id":"gpu","label":"GPU","type":"select","options":["H100","H200","B200","B300","GB200","MI300X","MI325X","MI355X","L40S","A100"]},{"id":"formFactor","label":"Form factor","type":"text"},{"id":"pricePerGpuHour","label":"Price per GPU-hour","type":"number","format":{"style":"currency","currency":"USD","maximumFractionDigits":4,"minimumFractionDigits":2}},{"id":"billing","label":"Billing","type":"select","options":["On-demand","Spot","Reserved/commit"]},{"id":"region","label":"Region","type":"text"},{"id":"instanceType","label":"Instance type","type":"text"},{"id":"priceBasis","label":"Price basis","type":"text"},{"id":"source","label":"Source","type":"url"},{"id":"checked","label":"Checked","type":"date"}]},"entries":[{"slug":"runpod-l40s-on-demand","name":"RunPod — L40S on-demand","path":"/gpu-prices/runpod-l40s-on-demand","category":null,"updatedAt":"2026-10-05T04:52:59.872Z","fields":{"provider":"RunPod","gpu":"L40S","formFactor":"48GB","pricePerGpuHour":1.09,"billing":"On-demand","region":null,"instanceType":"Secure Cloud pod, L40S","priceBasis":"Listed per hour for one GPU on RunPod's Secure Cloud pod pricing","source":"https://www.runpod.io/pricing","checked":"2026-10-05"}},{"slug":"crusoe-l40s-on-demand","name":"Crusoe — L40S on-demand","path":"/gpu-prices/crusoe-l40s-on-demand","category":null,"updatedAt":"2026-10-05T04:50:45.688Z","fields":{"provider":"Crusoe","gpu":"L40S","formFactor":"48GB","pricePerGpuHour":1.5,"billing":"On-demand","region":null,"instanceType":"NVIDIA L40S 48GB","priceBasis":"Listed per GPU-hour on Crusoe's on-demand pricing","source":"https://www.crusoe.ai/cloud/pricing","checked":"2026-10-05"}},{"slug":"digitalocean-l40s-on-demand","name":"DigitalOcean — L40S on-demand","path":"/gpu-prices/digitalocean-l40s-on-demand","category":null,"updatedAt":"2026-10-05T04:51:06.608Z","fields":{"provider":"DigitalOcean","gpu":"L40S","formFactor":"48GB, 1 GPU per Droplet","pricePerGpuHour":1.57,"billing":"On-demand","region":null,"instanceType":"GPU Droplet, NVIDIA L40S","priceBasis":"Listed per GPU-hour, on-demand GPU Droplet","source":"https://www.digitalocean.com/pricing/gpu-droplets","checked":"2026-10-05"}},{"slug":"runpod-a100-sxm-on-demand","name":"RunPod — A100 SXM 80GB on-demand","path":"/gpu-prices/runpod-a100-sxm-on-demand","category":null,"updatedAt":"2026-10-05T04:52:40.149Z","fields":{"provider":"RunPod","gpu":"A100","formFactor":"SXM 80GB","pricePerGpuHour":1.59,"billing":"On-demand","region":null,"instanceType":"Secure Cloud pod, A100 SXM","priceBasis":"Listed per hour for one GPU on RunPod's Secure Cloud pod pricing","source":"https://www.runpod.io/pricing","checked":"2026-10-05"}},{"slug":"vultr-l40s-on-demand","name":"Vultr — L40S on-demand","path":"/gpu-prices/vultr-l40s-on-demand","category":null,"updatedAt":"2026-10-05T04:53:08.123Z","fields":{"provider":"Vultr","gpu":"L40S","formFactor":"48GB, 1 to 8 GPUs","pricePerGpuHour":1.671,"billing":"On-demand","region":null,"instanceType":"NVIDIA L40S (1, 2, 4 or 8 GPUs)","priceBasis":"Listed per GPU-hour, on-demand pricing; re-checked October 5, 2026 against Vultr's official public plans API (api.vultr.com/v2/plans, plan vcg-l40s-16c-180g-48vram at $1.671/hour for 1 GPU) because the pricing page returned a browser check","source":"https://www.vultr.com/pricing/","checked":"2026-10-05"}},{"slug":"coreweave-l40s-on-demand","name":"CoreWeave — L40S on-demand","path":"/gpu-prices/coreweave-l40s-on-demand","category":null,"updatedAt":"2026-10-05T04:50:33.374Z","fields":{"provider":"CoreWeave","gpu":"L40S","formFactor":"8×L40S 48GB instance","pricePerGpuHour":2.25,"billing":"On-demand","region":"North America","instanceType":"NVIDIA L40S (8 GPUs)","priceBasis":"Instance price $18.00/hour for 8 GPUs, on-demand, North America, divided by 8","source":"https://www.coreweave.com/pricing","checked":"2026-10-05"}},{"slug":"azure-nd-h100-v5-spot","name":"Microsoft Azure — ND H100 v5 spot","path":"/gpu-prices/azure-nd-h100-v5-spot","category":null,"updatedAt":"2026-10-05T05:12:28.963Z","fields":{"provider":"Microsoft Azure","gpu":"H100","formFactor":"8×H100 instance (ND H100 v5)","pricePerGpuHour":2.271192,"billing":"Spot","region":"East US","instanceType":"Standard_ND96isr_H100_v5","priceBasis":"Instance Standard_ND96isr_H100_v5, 8 GPUs, East US, Linux spot $18.169536/hour, divided by 8; source is Microsoft's official Azure Retail Prices API (the Azure Linux VM pricing page loads its prices dynamically and could not be read)","source":"https://prices.azure.com/api/retail/prices?$filter=armRegionName%20eq%20%27eastus%27%20and%20armSkuName%20eq%20%27Standard_ND96isr_H100_v5%27","checked":"2026-10-05"}},{"slug":"crusoe-a100-sxm-on-demand","name":"Crusoe — A100 SXM 80GB on-demand","path":"/gpu-prices/crusoe-a100-sxm-on-demand","category":null,"updatedAt":"2026-10-05T04:50:36.557Z","fields":{"provider":"Crusoe","gpu":"A100","formFactor":"SXM 80GB","pricePerGpuHour":2.3,"billing":"On-demand","region":null,"instanceType":"NVIDIA A100 80GB SXM","priceBasis":"Listed per GPU-hour on Crusoe's on-demand pricing","source":"https://www.crusoe.ai/cloud/pricing","checked":"2026-10-05"}},{"slug":"coreweave-hgx-h100-spot","name":"CoreWeave — HGX H100 spot","path":"/gpu-prices/coreweave-hgx-h100-spot","category":null,"updatedAt":"2026-10-05T04:50:26.876Z","fields":{"provider":"CoreWeave","gpu":"H100","formFactor":"8×H100 80GB instance (HGX)","pricePerGpuHour":2.46375,"billing":"Spot","region":"North America","instanceType":"NVIDIA HGX H100 (8 GPUs)","priceBasis":"Instance spot price $19.71/hour for 8 GPUs, North America, divided by 8","source":"https://www.coreweave.com/pricing","checked":"2026-10-05"}},{"slug":"digitalocean-mi300x-on-demand","name":"DigitalOcean — MI300X on-demand","path":"/gpu-prices/digitalocean-mi300x-on-demand","category":null,"updatedAt":"2026-10-05T04:51:09.840Z","fields":{"provider":"DigitalOcean","gpu":"MI300X","formFactor":"192GB, 1 or 8 GPUs per Droplet","pricePerGpuHour":2.59,"billing":"On-demand","region":null,"instanceType":"GPU Droplet, AMD Instinct MI300X","priceBasis":"Listed per GPU-hour, on-demand GPU Droplet","source":"https://www.digitalocean.com/pricing/gpu-droplets","checked":"2026-10-05"}},{"slug":"coreweave-a100-on-demand","name":"CoreWeave — A100 on-demand","path":"/gpu-prices/coreweave-a100-on-demand","category":null,"updatedAt":"2026-10-05T04:50:14.009Z","fields":{"provider":"CoreWeave","gpu":"A100","formFactor":"8×A100 80GB instance","pricePerGpuHour":2.7,"billing":"On-demand","region":"North America","instanceType":"NVIDIA A100 (8 GPUs)","priceBasis":"Instance price $21.60/hour for 8 GPUs, on-demand, North America, divided by 8","source":"https://www.coreweave.com/pricing","checked":"2026-10-05"}},{"slug":"lambda-a100-sxm-80gb-on-demand","name":"Lambda — A100 SXM 80GB on-demand","path":"/gpu-prices/lambda-a100-sxm-80gb-on-demand","category":null,"updatedAt":"2026-10-05T04:51:50.645Z","fields":{"provider":"Lambda","gpu":"A100","formFactor":"SXM 80GB","pricePerGpuHour":2.79,"billing":"On-demand","region":null,"instanceType":"NVIDIA A100 SXM 80GB on-demand instance (240 vCPUs)","priceBasis":"Listed per GPU-hour in Lambda's on-demand instance table","source":"https://lambda.ai/pricing","checked":"2026-10-05"}},{"slug":"vultr-hgx-a100-on-demand","name":"Vultr — HGX A100 on-demand","path":"/gpu-prices/vultr-hgx-a100-on-demand","category":null,"updatedAt":"2026-10-05T04:53:03.950Z","fields":{"provider":"Vultr","gpu":"A100","formFactor":"HGX, 8×A100 instance","pricePerGpuHour":2.8,"billing":"On-demand","region":null,"instanceType":"NVIDIA HGX A100 (8 GPUs)","priceBasis":"Listed per GPU-hour, on-demand pricing; re-checked October 5, 2026 against Vultr's official public plans API (api.vultr.com/v2/plans-metal, plan vbm-112c-2048gb-8-a100-gpu at $22.40/hour for 8 GPUs) because the pricing page returned a browser check","source":"https://www.vultr.com/pricing/","checked":"2026-10-05"}},{"slug":"runpod-h100-pcie-on-demand","name":"RunPod — H100 PCIe on-demand","path":"/gpu-prices/runpod-h100-pcie-on-demand","category":null,"updatedAt":"2026-10-05T04:52:50.018Z","fields":{"provider":"RunPod","gpu":"H100","formFactor":"PCIe 80GB","pricePerGpuHour":2.89,"billing":"On-demand","region":null,"instanceType":"Secure Cloud pod, H100 PCIe","priceBasis":"Listed per hour for one GPU on RunPod's Secure Cloud pod pricing","source":"https://www.runpod.io/pricing","checked":"2026-10-05"}},{"slug":"hyperstack-h100-sxm-on-demand","name":"Hyperstack — H100 SXM on-demand","path":"/gpu-prices/hyperstack-h100-sxm-on-demand","category":null,"updatedAt":"2026-10-05T04:51:40.172Z","fields":{"provider":"Hyperstack","gpu":"H100","formFactor":"SXM 80GB","pricePerGpuHour":3.2,"billing":"On-demand","region":null,"instanceType":"NVIDIA H100 SXM","priceBasis":"Per GPU per hour: \"Pricing Per Hour\" in Hyperstack's on-demand GPU table, whose rows give per-GPU CPU and RAM; Hyperstack's pricebook docs state a flavor's hourly cost is the per-GPU rate multiplied by its GPU count","source":"https://www.hyperstack.cloud/gpu-pricing","checked":"2026-10-05"}},{"slug":"crusoe-mi300x-on-demand","name":"Crusoe — MI300X on-demand","path":"/gpu-prices/crusoe-mi300x-on-demand","category":null,"updatedAt":"2026-10-05T04:50:48.623Z","fields":{"provider":"Crusoe","gpu":"MI300X","formFactor":"192GB","pricePerGpuHour":3.45,"billing":"On-demand","region":null,"instanceType":"AMD MI300X 192GB","priceBasis":"Listed per GPU-hour on Crusoe's on-demand pricing","source":"https://www.crusoe.ai/cloud/pricing","checked":"2026-10-05"}},{"slug":"runpod-h100-sxm-on-demand","name":"RunPod — H100 SXM on-demand","path":"/gpu-prices/runpod-h100-sxm-on-demand","category":null,"updatedAt":"2026-10-05T04:52:53.269Z","fields":{"provider":"RunPod","gpu":"H100","formFactor":"SXM 80GB","pricePerGpuHour":3.49,"billing":"On-demand","region":null,"instanceType":"Secure Cloud pod, H100 SXM","priceBasis":"Listed per hour for one GPU on RunPod's Secure Cloud pod pricing","source":"https://www.runpod.io/pricing","checked":"2026-10-05"}},{"slug":"digitalocean-mi325x-on-demand","name":"DigitalOcean — MI325X on-demand","path":"/gpu-prices/digitalocean-mi325x-on-demand","category":null,"updatedAt":"2026-10-05T04:51:13.249Z","fields":{"provider":"DigitalOcean","gpu":"MI325X","formFactor":"256GB, 1 or 8 GPUs per Droplet","pricePerGpuHour":3.8,"billing":"On-demand","region":null,"instanceType":"GPU Droplet, AMD Instinct MI325X","priceBasis":"Listed per GPU-hour, on-demand GPU Droplet","source":"https://www.digitalocean.com/pricing/gpu-droplets","checked":"2026-10-05"}},{"slug":"crusoe-h100-hgx-on-demand","name":"Crusoe — H100 HGX on-demand","path":"/gpu-prices/crusoe-h100-hgx-on-demand","category":null,"updatedAt":"2026-10-05T04:50:40.289Z","fields":{"provider":"Crusoe","gpu":"H100","formFactor":"HGX 80GB","pricePerGpuHour":3.9,"billing":"On-demand","region":null,"instanceType":"NVIDIA H100 80GB HGX","priceBasis":"Listed per GPU-hour on Crusoe's on-demand pricing","source":"https://www.crusoe.ai/cloud/pricing","checked":"2026-10-05"}},{"slug":"hyperstack-h200-sxm-on-demand","name":"Hyperstack — H200 SXM on-demand","path":"/gpu-prices/hyperstack-h200-sxm-on-demand","category":null,"updatedAt":"2026-10-05T04:51:44.512Z","fields":{"provider":"Hyperstack","gpu":"H200","formFactor":"SXM 141GB","pricePerGpuHour":3.99,"billing":"On-demand","region":null,"instanceType":"NVIDIA H200 SXM","priceBasis":"Per GPU per hour: \"Pricing Per Hour\" in Hyperstack's on-demand GPU table, whose rows give per-GPU CPU and RAM; Hyperstack's pricebook docs state a flavor's hourly cost is the per-GPU rate multiplied by its GPU count","source":"https://www.hyperstack.cloud/gpu-pricing","checked":"2026-10-05"}},{"slug":"lambda-h100-sxm-on-demand","name":"Lambda — H100 SXM on-demand","path":"/gpu-prices/lambda-h100-sxm-on-demand","category":null,"updatedAt":"2026-10-05T04:51:57.097Z","fields":{"provider":"Lambda","gpu":"H100","formFactor":"SXM 80GB","pricePerGpuHour":3.99,"billing":"On-demand","region":null,"instanceType":"NVIDIA H100 SXM on-demand instance (208 vCPUs)","priceBasis":"Listed per GPU-hour in Lambda's on-demand instance table","source":"https://lambda.ai/pricing","checked":"2026-10-05"}},{"slug":"azure-nd-a100-v4-80gb-on-demand","name":"Microsoft Azure — ND A100 v4 (80GB) on-demand","path":"/gpu-prices/azure-nd-a100-v4-80gb-on-demand","category":null,"updatedAt":"2026-10-05T04:52:01.527Z","fields":{"provider":"Microsoft Azure","gpu":"A100","formFactor":"8×A100 80GB instance (NDm A100 v4)","pricePerGpuHour":4.09625,"billing":"On-demand","region":"East US","instanceType":"Standard_ND96amsr_A100_v4","priceBasis":"Instance Standard_ND96amsr_A100_v4, 8 GPUs, East US, Linux pay-as-you-go $32.77/hour, divided by 8; source is Microsoft's official Azure Retail Prices API (the Azure Linux VM pricing page loads its prices dynamically and could not be read)","source":"https://prices.azure.com/api/retail/prices?$filter=armRegionName%20eq%20%27eastus%27%20and%20armSkuName%20eq%20%27Standard_ND96amsr_A100_v4%27","checked":"2026-10-05"}},{"slug":"crusoe-h200-hgx-on-demand","name":"Crusoe — H200 HGX on-demand","path":"/gpu-prices/crusoe-h200-hgx-on-demand","category":null,"updatedAt":"2026-10-05T04:50:42.527Z","fields":{"provider":"Crusoe","gpu":"H200","formFactor":"HGX 141GB","pricePerGpuHour":4.29,"billing":"On-demand","region":null,"instanceType":"NVIDIA H200 141GB HGX","priceBasis":"Listed per GPU-hour on Crusoe's on-demand pricing","source":"https://www.crusoe.ai/cloud/pricing","checked":"2026-10-05"}},{"slug":"digitalocean-hgx-h100-on-demand","name":"DigitalOcean — HGX H100 on-demand","path":"/gpu-prices/digitalocean-hgx-h100-on-demand","category":null,"updatedAt":"2026-10-05T04:51:00.287Z","fields":{"provider":"DigitalOcean","gpu":"H100","formFactor":"HGX 80GB, 1 or 8 GPUs per Droplet","pricePerGpuHour":4.41,"billing":"On-demand","region":null,"instanceType":"GPU Droplet, NVIDIA HGX H100","priceBasis":"Listed per GPU-hour, on-demand GPU Droplet","source":"https://www.digitalocean.com/pricing/gpu-droplets","checked":"2026-10-05"}},{"slug":"digitalocean-hgx-h200-on-demand","name":"DigitalOcean — HGX H200 on-demand","path":"/gpu-prices/digitalocean-hgx-h200-on-demand","category":null,"updatedAt":"2026-10-05T04:51:03.603Z","fields":{"provider":"DigitalOcean","gpu":"H200","formFactor":"HGX 141GB, 1 or 8 GPUs per Droplet","pricePerGpuHour":4.47,"billing":"On-demand","region":null,"instanceType":"GPU Droplet, NVIDIA HGX H200","priceBasis":"Listed per GPU-hour, on-demand GPU Droplet","source":"https://www.digitalocean.com/pricing/gpu-droplets","checked":"2026-10-05"}},{"slug":"nebius-hgx-h100-on-demand","name":"Nebius — HGX H100 on-demand","path":"/gpu-prices/nebius-hgx-h100-on-demand","category":null,"updatedAt":"2026-10-05T04:52:30.335Z","fields":{"provider":"Nebius","gpu":"H100","formFactor":"HGX 80GB, per GPU (16 vCPUs, 200 GB RAM)","pricePerGpuHour":4.5,"billing":"On-demand","region":null,"instanceType":"NVIDIA HGX H100","priceBasis":"On-demand, effective October 1, 2026: listed per GPU-hour in the column \"GPU-hour (Effective October 1, 2026)\", which replaces the earlier on-demand rate of $3.85","source":"https://nebius.com/prices","checked":"2026-10-05"}},{"slug":"runpod-h200-on-demand","name":"RunPod — H200 on-demand","path":"/gpu-prices/runpod-h200-on-demand","category":null,"updatedAt":"2026-10-05T04:52:56.418Z","fields":{"provider":"RunPod","gpu":"H200","formFactor":"141GB","pricePerGpuHour":4.59,"billing":"On-demand","region":null,"instanceType":"Secure Cloud pod, H200","priceBasis":"Listed per hour for one GPU on RunPod's Secure Cloud pod pricing","source":"https://www.runpod.io/pricing","checked":"2026-10-05"}},{"slug":"google-cloud-a2-ultra-a100-on-demand","name":"Google Cloud — A2 Ultra (A100 80GB) on-demand","path":"/gpu-prices/google-cloud-a2-ultra-a100-on-demand","category":null,"updatedAt":"2026-10-05T04:51:16.916Z","fields":{"provider":"Google Cloud","gpu":"A100","formFactor":"8×A100 80GB instance (A2 Ultra)","pricePerGpuHour":5.0688,"billing":"On-demand","region":"Iowa (us-central1)","instanceType":"a2-ultragpu-8g","priceBasis":"Machine type a2-ultragpu-8g, 8 GPUs, us-central1, on-demand $40.550383123/hour (GPUs, vCPU, memory and Local SSD included), divided by 8","source":"https://cloud.google.com/products/compute/pricing/accelerator-optimized","checked":"2026-10-05"}},{"slug":"aws-p5-h100-capacity-block","name":"AWS — P5 (H100) Capacity Block","path":"/gpu-prices/aws-p5-h100-capacity-block","category":null,"updatedAt":"2026-10-02T12:07:17.899Z","fields":{"provider":"AWS","gpu":"H100","formFactor":"8×H100 instance (p5.48xlarge)","pricePerGpuHour":5.191,"billing":"Reserved/commit","region":"US East (N. Virginia)","instanceType":"p5.48xlarge","priceBasis":"Capacity Block effective hourly rate per accelerator as listed by AWS ($41.528 per instance-hour, 8 GPUs); fee paid up front","source":"https://aws.amazon.com/ec2/capacityblocks/pricing/","checked":"2026-10-02"}},{"slug":"nebius-hgx-h200-on-demand","name":"Nebius — HGX H200 on-demand","path":"/gpu-prices/nebius-hgx-h200-on-demand","category":null,"updatedAt":"2026-10-05T04:52:33.213Z","fields":{"provider":"Nebius","gpu":"H200","formFactor":"HGX 141GB, per GPU (16 vCPUs, 200 GB RAM)","pricePerGpuHour":5.4,"billing":"On-demand","region":null,"instanceType":"NVIDIA HGX H200","priceBasis":"On-demand, effective October 1, 2026: listed per GPU-hour in the column \"GPU-hour (Effective October 1, 2026)\", which replaces the earlier on-demand rate of $4.50","source":"https://nebius.com/prices","checked":"2026-10-05"}},{"slug":"hyperstack-b200-on-demand","name":"Hyperstack — B200 on-demand","path":"/gpu-prices/hyperstack-b200-on-demand","category":null,"updatedAt":"2026-10-05T04:51:36.394Z","fields":{"provider":"Hyperstack","gpu":"B200","formFactor":null,"pricePerGpuHour":6,"billing":"On-demand","region":null,"instanceType":"NVIDIA B200","priceBasis":"Per GPU per hour: \"Pricing Per Hour\" in Hyperstack's on-demand GPU table, whose rows give per-GPU CPU and RAM; Hyperstack's pricebook docs state a flavor's hourly cost is the per-GPU rate multiplied by its GPU count","source":"https://www.hyperstack.cloud/gpu-pricing","checked":"2026-10-05"}},{"slug":"azure-nd-mi300x-v5-on-demand","name":"Microsoft Azure — ND MI300X v5 on-demand","path":"/gpu-prices/azure-nd-mi300x-v5-on-demand","category":null,"updatedAt":"2026-10-05T04:52:18.570Z","fields":{"provider":"Microsoft Azure","gpu":"MI300X","formFactor":"8×MI300X instance (ND MI300X v5)","pricePerGpuHour":6,"billing":"On-demand","region":"East US 2","instanceType":"Standard_ND96isr_MI300X_v5","priceBasis":"Instance Standard_ND96isr_MI300X_v5, 8 GPUs, East US 2, Linux pay-as-you-go $48.00/hour, divided by 8; source is Microsoft's official Azure Retail Prices API (the Azure Linux VM pricing page loads its prices dynamically and could not be read)","source":"https://prices.azure.com/api/retail/prices?$filter=armSkuName%20eq%20%27Standard_ND96isr_MI300X_v5%27%20and%20armRegionName%20eq%20%27eastus2%27","checked":"2026-10-05"}},{"slug":"coreweave-hgx-h100-on-demand","name":"CoreWeave — HGX H100 on-demand","path":"/gpu-prices/coreweave-hgx-h100-on-demand","category":null,"updatedAt":"2026-10-05T04:50:23.856Z","fields":{"provider":"CoreWeave","gpu":"H100","formFactor":"8×H100 80GB instance (HGX)","pricePerGpuHour":6.155,"billing":"On-demand","region":"North America","instanceType":"NVIDIA HGX H100 (8 GPUs)","priceBasis":"Instance price $49.24/hour for 8 GPUs, on-demand, North America, divided by 8","source":"https://www.coreweave.com/pricing","checked":"2026-10-05"}},{"slug":"coreweave-hgx-h200-on-demand","name":"CoreWeave — HGX H200 on-demand","path":"/gpu-prices/coreweave-hgx-h200-on-demand","category":null,"updatedAt":"2026-10-05T04:50:30.166Z","fields":{"provider":"CoreWeave","gpu":"H200","formFactor":"8×H200 141GB instance (HGX)","pricePerGpuHour":6.305,"billing":"On-demand","region":"North America","instanceType":"NVIDIA HGX H200 (8 GPUs)","priceBasis":"Instance price $50.44/hour for 8 GPUs, on-demand, North America, divided by 8","source":"https://www.coreweave.com/pricing","checked":"2026-10-05"}},{"slug":"google-cloud-a3-high-h100-spot","name":"Google Cloud — A3 High (H100) spot","path":"/gpu-prices/google-cloud-a3-high-h100-spot","category":null,"updatedAt":"2026-10-05T04:51:25.383Z","fields":{"provider":"Google Cloud","gpu":"H100","formFactor":"8×H100 instance (A3 High)","pricePerGpuHour":6.6203,"billing":"Spot","region":"Iowa (us-central1)","instanceType":"a3-highgpu-8g","priceBasis":"Machine type a3-highgpu-8g, 8 GPUs, us-central1, \"Current Spot pricing (USD)\" column $52.962296548/hour, divided by 8 (not the DWS Flex-start $38.32, DWS Calendar Mode $41.60, 1-year CUD $61.383674231 or 3-year CUD $38.864383195 columns); the same spot rate is listed on Google's Spot VM pricing page (cloud.google.com/spot-vms/pricing), where a3-highgpu-1g is $6.620287068/hour","source":"https://cloud.google.com/products/compute/pricing/accelerator-optimized","checked":"2026-10-05"}},{"slug":"lambda-b200-sxm6-on-demand","name":"Lambda — B200 SXM6 on-demand","path":"/gpu-prices/lambda-b200-sxm6-on-demand","category":null,"updatedAt":"2026-10-05T04:51:54.202Z","fields":{"provider":"Lambda","gpu":"B200","formFactor":"SXM6 180GB","pricePerGpuHour":6.69,"billing":"On-demand","region":null,"instanceType":"NVIDIA B200 SXM6 on-demand instance (208 vCPUs)","priceBasis":"Listed per GPU-hour in Lambda's on-demand instance table","source":"https://lambda.ai/pricing","checked":"2026-10-05"}},{"slug":"runpod-b200-on-demand","name":"RunPod — B200 on-demand","path":"/gpu-prices/runpod-b200-on-demand","category":null,"updatedAt":"2026-10-05T04:52:43.471Z","fields":{"provider":"RunPod","gpu":"B200","formFactor":"180GB","pricePerGpuHour":6.79,"billing":"On-demand","region":null,"instanceType":"Secure Cloud pod, B200","priceBasis":"Listed per hour for one GPU on RunPod's Secure Cloud pod pricing","source":"https://www.runpod.io/pricing","checked":"2026-10-05"}},{"slug":"aws-p5en-h200-capacity-block","name":"AWS — P5en (H200) Capacity Block","path":"/gpu-prices/aws-p5en-h200-capacity-block","category":null,"updatedAt":"2026-10-02T12:09:08.236Z","fields":{"provider":"AWS","gpu":"H200","formFactor":"8×H200 instance (p5en.48xlarge)","pricePerGpuHour":6.865,"billing":"Reserved/commit","region":"US East (N. Virginia)","instanceType":"p5en.48xlarge","priceBasis":"Capacity Block effective hourly rate per accelerator as listed by AWS ($54.920 per instance-hour, 8 GPUs); fee paid up front","source":"https://aws.amazon.com/ec2/capacityblocks/pricing/","checked":"2026-10-02"}},{"slug":"hyperstack-b300-on-demand","name":"Hyperstack — B300 on-demand","path":"/gpu-prices/hyperstack-b300-on-demand","category":null,"updatedAt":"2026-10-05T04:53:15.522Z","fields":{"provider":"Hyperstack","gpu":"B300","formFactor":"288GB","pricePerGpuHour":7.4,"billing":"On-demand","region":null,"instanceType":"NVIDIA B300","priceBasis":"Per GPU per hour: \"Pricing Per Hour\" in Hyperstack's on-demand GPU table, whose rows give per-GPU CPU and RAM; Hyperstack's pricebook docs state a flavor's hourly cost is the per-GPU rate multiplied by its GPU count","source":"https://www.hyperstack.cloud/gpu-pricing","checked":"2026-10-05"}},{"slug":"runpod-b300-on-demand","name":"RunPod — B300 on-demand","path":"/gpu-prices/runpod-b300-on-demand","category":null,"updatedAt":"2026-10-05T04:52:46.909Z","fields":{"provider":"RunPod","gpu":"B300","formFactor":"288GB HBM3e","pricePerGpuHour":7.89,"billing":"On-demand","region":null,"instanceType":"Secure Cloud pod, B300","priceBasis":"Listed per hour for one GPU on RunPod's Secure Cloud pod pricing","source":"https://www.runpod.io/pricing","checked":"2026-10-05"}},{"slug":"nebius-hgx-b200-on-demand","name":"Nebius — HGX B200 on-demand","path":"/gpu-prices/nebius-hgx-b200-on-demand","category":null,"updatedAt":"2026-10-05T04:52:22.357Z","fields":{"provider":"Nebius","gpu":"B200","formFactor":"HGX, per GPU (20 vCPUs, 224 GB RAM)","pricePerGpuHour":8.5,"billing":"On-demand","region":null,"instanceType":"NVIDIA HGX B200","priceBasis":"On-demand, effective October 1, 2026: listed per GPU-hour in the column \"GPU-hour (Effective October 1, 2026)\", which replaces the earlier on-demand rate of $7.15","source":"https://nebius.com/prices","checked":"2026-10-05"}},{"slug":"coreweave-hgx-b200-on-demand","name":"CoreWeave — HGX B200 on-demand","path":"/gpu-prices/coreweave-hgx-b200-on-demand","category":null,"updatedAt":"2026-10-05T04:50:20.087Z","fields":{"provider":"CoreWeave","gpu":"B200","formFactor":"8×B200 180GB instance (HGX)","pricePerGpuHour":8.6,"billing":"On-demand","region":"North America","instanceType":"NVIDIA HGX B200 (8 GPUs)","priceBasis":"Instance price $68.80/hour for 8 GPUs, on-demand, North America, divided by 8","source":"https://www.coreweave.com/pricing","checked":"2026-10-05"}},{"slug":"nebius-hgx-b300-on-demand","name":"Nebius — HGX B300 on-demand","path":"/gpu-prices/nebius-hgx-b300-on-demand","category":null,"updatedAt":"2026-10-05T04:52:26.638Z","fields":{"provider":"Nebius","gpu":"B300","formFactor":"HGX, per GPU (24 vCPUs, 346 GB RAM)","pricePerGpuHour":9.5,"billing":"On-demand","region":null,"instanceType":"NVIDIA HGX B300","priceBasis":"On-demand, effective October 1, 2026: listed per GPU-hour in the column \"GPU-hour (Effective October 1, 2026)\", which replaces the earlier on-demand rate of $7.85","source":"https://nebius.com/prices","checked":"2026-10-05"}},{"slug":"coreweave-gb200-nvl72-on-demand","name":"CoreWeave — GB200 NVL72 on-demand","path":"/gpu-prices/coreweave-gb200-nvl72-on-demand","category":null,"updatedAt":"2026-10-05T04:50:17.066Z","fields":{"provider":"CoreWeave","gpu":"GB200","formFactor":"4×GB200 186GB instance (NVL72)","pricePerGpuHour":10.5,"billing":"On-demand","region":"North America","instanceType":"NVIDIA GB200 NVL72 (4 GPUs)","priceBasis":"Instance price $42.00/hour for 4 GPUs, on-demand, North America, divided by 4","source":"https://www.coreweave.com/pricing","checked":"2026-10-05"}},{"slug":"azure-nd-h200-v5-on-demand","name":"Microsoft Azure — ND H200 v5 on-demand","path":"/gpu-prices/azure-nd-h200-v5-on-demand","category":null,"updatedAt":"2026-10-05T04:52:15.415Z","fields":{"provider":"Microsoft Azure","gpu":"H200","formFactor":"8×H200 instance (ND H200 v5)","pricePerGpuHour":10.6,"billing":"On-demand","region":"East US 2","instanceType":"Standard_ND96isr_H200_v5","priceBasis":"Instance Standard_ND96isr_H200_v5, 8 GPUs, East US 2, Linux pay-as-you-go $84.80/hour, divided by 8; source is Microsoft's official Azure Retail Prices API (the Azure Linux VM pricing page loads its prices dynamically and could not be read)","source":"https://prices.azure.com/api/retail/prices?$filter=armSkuName%20eq%20%27Standard_ND96isr_H200_v5%27%20and%20armRegionName%20eq%20%27eastus2%27","checked":"2026-10-05"}},{"slug":"google-cloud-a3-ultra-h200-on-demand","name":"Google Cloud — A3 Ultra (H200) on-demand","path":"/gpu-prices/google-cloud-a3-ultra-h200-on-demand","category":null,"updatedAt":"2026-10-05T04:51:32.869Z","fields":{"provider":"Google Cloud","gpu":"H200","formFactor":"8×H200 instance (A3 Ultra)","pricePerGpuHour":10.6009,"billing":"On-demand","region":"Iowa (us-central1)","instanceType":"a3-ultragpu-8g","priceBasis":"Machine type a3-ultragpu-8g, 8 GPUs, us-central1, on-demand $84.806908493/hour (GPUs, vCPU, memory and Local SSD included), divided by 8","source":"https://cloud.google.com/products/compute/pricing/accelerator-optimized","checked":"2026-10-05"}},{"slug":"google-cloud-a3-high-h100-on-demand","name":"Google Cloud — A3 High (H100) on-demand","path":"/gpu-prices/google-cloud-a3-high-h100-on-demand","category":null,"updatedAt":"2026-10-05T04:51:20.547Z","fields":{"provider":"Google Cloud","gpu":"H100","formFactor":"8×H100 instance (A3 High)","pricePerGpuHour":11.0613,"billing":"On-demand","region":"Iowa (us-central1)","instanceType":"a3-highgpu-8g","priceBasis":"Machine type a3-highgpu-8g, 8 GPUs, us-central1, on-demand $88.490000119/hour (GPUs, vCPU, memory and Local SSD included), divided by 8","source":"https://cloud.google.com/products/compute/pricing/accelerator-optimized","checked":"2026-10-05"}},{"slug":"google-cloud-a3-mega-h100-on-demand","name":"Google Cloud — A3 Mega (H100) on-demand","path":"/gpu-prices/google-cloud-a3-mega-h100-on-demand","category":null,"updatedAt":"2026-10-05T04:51:29.147Z","fields":{"provider":"Google Cloud","gpu":"H100","formFactor":"8×H100 instance (A3 Mega)","pricePerGpuHour":11.6751,"billing":"On-demand","region":"Iowa (us-central1)","instanceType":"a3-megagpu-8g","priceBasis":"Machine type a3-megagpu-8g, 8 GPUs, us-central1, on-demand $93.400712807/hour (GPUs, vCPU, memory and Local SSD included), divided by 8","source":"https://cloud.google.com/products/compute/pricing/accelerator-optimized","checked":"2026-10-05"}},{"slug":"azure-nd-h100-v5-on-demand","name":"Microsoft Azure — ND H100 v5 on-demand","path":"/gpu-prices/azure-nd-h100-v5-on-demand","category":null,"updatedAt":"2026-10-05T04:52:05.738Z","fields":{"provider":"Microsoft Azure","gpu":"H100","formFactor":"8×H100 instance (ND H100 v5)","pricePerGpuHour":12.29,"billing":"On-demand","region":"East US","instanceType":"Standard_ND96isr_H100_v5","priceBasis":"Instance Standard_ND96isr_H100_v5, 8 GPUs, East US, Linux pay-as-you-go $98.32/hour, divided by 8; source is Microsoft's official Azure Retail Prices API (the Azure Linux VM pricing page loads its prices dynamically and could not be read)","source":"https://prices.azure.com/api/retail/prices?$filter=armRegionName%20eq%20%27eastus%27%20and%20armSkuName%20eq%20%27Standard_ND96isr_H100_v5%27","checked":"2026-10-05"}},{"slug":"aws-p6-b200-capacity-block","name":"AWS — P6-B200 Capacity Block","path":"/gpu-prices/aws-p6-b200-capacity-block","category":null,"updatedAt":"2026-10-02T12:09:13.590Z","fields":{"provider":"AWS","gpu":"B200","formFactor":"8×B200 instance (p6-b200.48xlarge)","pricePerGpuHour":12.355,"billing":"Reserved/commit","region":"US East (N. Virginia)","instanceType":"p6-b200.48xlarge","priceBasis":"Capacity Block effective hourly rate per accelerator as listed by AWS ($98.84 per instance-hour, 8 GPUs); fee paid up front","source":"https://aws.amazon.com/ec2/capacityblocks/pricing/","checked":"2026-10-02"}},{"slug":"aws-p6-b300-capacity-block","name":"AWS — P6-B300 Capacity Block","path":"/gpu-prices/aws-p6-b300-capacity-block","category":null,"updatedAt":"2026-10-02T12:09:18.616Z","fields":{"provider":"AWS","gpu":"B300","formFactor":"8×B300 instance (p6-b300.48xlarge)","pricePerGpuHour":14.04,"billing":"Reserved/commit","region":"US East (N. Virginia)","instanceType":"p6-b300.48xlarge","priceBasis":"Capacity Block effective hourly rate per accelerator as listed by AWS ($112.32 per instance-hour, 8 GPUs); fee paid up front","source":"https://aws.amazon.com/ec2/capacityblocks/pricing/","checked":"2026-10-02"}}]}