{"meta":{"title":"Cloud GPU On-Demand Price Index","description":"On-demand list $\/GPU-hour for 27 vendor offerings across the main data-centre NVIDIA GPUs (H100, H200, A100 80GB, L40S, L4, RTX 4090) at AWS, GCP, Azure, Lambda, RunPod and the Vast.ai marketplace. Hyperscaler per-instance prices are divided to a per-GPU rate; every priced cell is sourced and dated. Snapshot as of 2026-07-16.","generated":"2026-09-09T22:40:02+00:00","version":"1.0","license":"https:\/\/creativecommons.org\/licenses\/by\/4.0\/","license_name":"CC BY 4.0","source":"https:\/\/d-central.tech\/cloud-gpu-pricing\/","record_count":27,"as_of":"2026-07-16","next_review":"2026-08-16","method":"On-demand list prices only (no spot, no committed\/reserved). Hyperscaler (AWS\/GCP\/Azure) prices are per-instance list divided by GPU count. Neocloud (Lambda\/RunPod) and marketplace (Vast.ai) are per-GPU; Vast.ai figures are advertised on-demand floors that fluctuate by host. Sourced via vendor pricing pages and dated aggregators (Vantage, gcloud-compute, getdeploying); independently cross-checked.","disclaimer":"On-demand list prices, snapshot as of 2026-07-16. Cloud GPU prices change and vary by region, instance configuration, commitment tier and availability; hyperscaler figures are per-instance list prices divided by the GPU count; marketplace (Vast.ai) figures are advertised on-demand floors that fluctuate by host. Verify the current price at the linked source before relying on it. Not investment advice."},"spread":[{"gpu":"NVIDIA H100 80GB","hyperscaler_min":6.88,"hyperscaler_max":12.29,"alt_min":1.73,"neocloud_min":2.99,"marketplace_min":1.73,"pct_cheaper_vs_cheapest_hyper":75,"pct_cheaper_vs_dearest_hyper":86},{"gpu":"NVIDIA H200 141GB","hyperscaler_min":13.78,"hyperscaler_max":13.78,"alt_min":3.75,"neocloud_min":4.39,"marketplace_min":3.75,"pct_cheaper_vs_cheapest_hyper":73,"pct_cheaper_vs_dearest_hyper":73},{"gpu":"NVIDIA A100 80GB","hyperscaler_min":3.43,"hyperscaler_max":5.07,"alt_min":0.8,"neocloud_min":1.39,"marketplace_min":0.8,"pct_cheaper_vs_cheapest_hyper":77,"pct_cheaper_vs_dearest_hyper":84},{"gpu":"NVIDIA L40S 48GB","hyperscaler_min":1.86,"hyperscaler_max":1.86,"alt_min":0.47,"neocloud_min":0.99,"marketplace_min":0.47,"pct_cheaper_vs_cheapest_hyper":75,"pct_cheaper_vs_dearest_hyper":75},{"gpu":"NVIDIA L4 24GB","hyperscaler_min":0.71,"hyperscaler_max":0.8,"alt_min":0.2,"neocloud_min":0.39,"marketplace_min":0.2,"pct_cheaper_vs_cheapest_hyper":72,"pct_cheaper_vs_dearest_hyper":75},{"gpu":"NVIDIA RTX 4090 24GB","hyperscaler_min":null,"hyperscaler_max":null,"alt_min":0.34,"neocloud_min":0.69,"marketplace_min":0.34}],"rows":[{"gpu":"NVIDIA H100 80GB","vram_gb":80,"vendor":"AWS","vendor_type":"Hyperscaler","instance":"p5.48xlarge","gpus_per_instance":8,"instance_usd_per_hr":55.04,"usd_per_gpu_hr":6.88,"status":"on_demand","confidence":"high","source_url":"https:\/\/instances.vantage.sh\/aws\/ec2\/p5.48xlarge","note":"8x H100 SXM; $55.04\/hr on-demand us-east-1 \/ 8.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H100 80GB","vram_gb":80,"vendor":"Google Cloud","vendor_type":"Hyperscaler","instance":"a3-highgpu-8g","gpus_per_instance":8,"instance_usd_per_hr":88.49,"usd_per_gpu_hr":11.06,"status":"on_demand","confidence":"high","source_url":"https:\/\/gcloud-compute.com\/a3-highgpu-8g.html","note":"8x H100 SXM; $88.49\/hr on-demand us-central1 \/ 8 (up to ~$15.83 in pricier regions).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H100 80GB","vram_gb":80,"vendor":"Microsoft Azure","vendor_type":"Hyperscaler","instance":"ND96isr H100 v5","gpus_per_instance":8,"instance_usd_per_hr":98.32,"usd_per_gpu_hr":12.29,"status":"on_demand","confidence":"high","source_url":"https:\/\/instances.vantage.sh\/azure\/vm\/nd96isrh100-v5","note":"8x H100 SXM; $98.32\/hr pay-as-you-go \/ 8.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H100 80GB","vram_gb":80,"vendor":"Lambda Labs","vendor_type":"Neocloud","instance":"8x H100 SXM","gpus_per_instance":1,"instance_usd_per_hr":3.99,"usd_per_gpu_hr":3.99,"status":"on_demand","confidence":"high","source_url":"https:\/\/lambda.ai\/pricing","note":"Per-GPU on the 8-GPU on-demand node ($4.29 for a single 1x H100).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H100 80GB","vram_gb":80,"vendor":"RunPod","vendor_type":"Neocloud","instance":"H100 SXM (Secure Cloud)","gpus_per_instance":1,"instance_usd_per_hr":2.99,"usd_per_gpu_hr":2.99,"status":"on_demand","confidence":"high","source_url":"https:\/\/www.runpod.io\/gpu-models\/h100-sxm","note":"Secure Cloud on-demand ($2.69 Community Cloud).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H100 80GB","vram_gb":80,"vendor":"Vast.ai","vendor_type":"Marketplace","instance":"H100 SXM","gpus_per_instance":1,"instance_usd_per_hr":1.73,"usd_per_gpu_hr":1.73,"status":"marketplace_floor","confidence":"medium","source_url":"https:\/\/vast.ai\/pricing\/gpu\/H100-SXM","note":"Marketplace floor; verified DC hosts ~$1.50-1.87, community lower. Varies by host.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H200 141GB","vram_gb":141,"vendor":"AWS","vendor_type":"Hyperscaler","instance":"p5e.48xlarge","gpus_per_instance":8,"instance_usd_per_hr":null,"usd_per_gpu_hr":null,"status":"no_public_on_demand","confidence":"high","source_url":"https:\/\/aws.amazon.com\/ec2\/instance-types\/p5\/","note":"AWS sells H200 (p5e) via EC2 Capacity Blocks (reserved\/scheduled) only; no published on-demand list price.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H200 141GB","vram_gb":141,"vendor":"Google Cloud","vendor_type":"Hyperscaler","instance":"a3-ultragpu-8g","gpus_per_instance":8,"instance_usd_per_hr":null,"usd_per_gpu_hr":null,"status":"no_clean_on_demand","confidence":"medium","source_url":"https:\/\/instances.vantage.sh\/gcp\/a3-ultragpu-8g","note":"The published a3-ultragpu on-demand SKU (~$12.27\/node) appears to exclude the H200 accelerator cost (billed separately), which would price it implausibly below GCP's own H100 \u2014 excluded pending an all-in on-demand figure.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H200 141GB","vram_gb":141,"vendor":"Microsoft Azure","vendor_type":"Hyperscaler","instance":"ND96isr H200 v5","gpus_per_instance":8,"instance_usd_per_hr":110.24,"usd_per_gpu_hr":13.78,"status":"on_demand","confidence":"high","source_url":"https:\/\/instances.vantage.sh\/azure\/vm\/nd96isrh200-v5","note":"8x H200; $110.24\/hr \/ 8 (cheapest regions ~$10.60\/GPU).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H200 141GB","vram_gb":141,"vendor":"Lambda Labs","vendor_type":"Neocloud","instance":"","gpus_per_instance":1,"instance_usd_per_hr":null,"usd_per_gpu_hr":null,"status":"not_on_demand","confidence":"high","source_url":"https:\/\/lambda.ai\/pricing","note":"H200 not on Lambda's self-serve on-demand table (1-Click Clusters \/ reserved only).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H200 141GB","vram_gb":141,"vendor":"RunPod","vendor_type":"Neocloud","instance":"H200 SXM (Secure Cloud)","gpus_per_instance":1,"instance_usd_per_hr":4.39,"usd_per_gpu_hr":4.39,"status":"on_demand","confidence":"high","source_url":"https:\/\/www.runpod.io\/gpu-models\/h200","note":"Secure Cloud on-demand ($3.59 Community Cloud).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA H200 141GB","vram_gb":141,"vendor":"Vast.ai","vendor_type":"Marketplace","instance":"H200","gpus_per_instance":1,"instance_usd_per_hr":3.75,"usd_per_gpu_hr":3.75,"status":"marketplace_floor","confidence":"medium","source_url":"https:\/\/vast.ai\/pricing\/gpu\/H200","note":"Marketplace floor; varies by host (H200 NVL PCIe variant lists lower).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA A100 80GB","vram_gb":80,"vendor":"AWS","vendor_type":"Hyperscaler","instance":"p4de.24xlarge","gpus_per_instance":8,"instance_usd_per_hr":27.447,"usd_per_gpu_hr":3.43,"status":"on_demand","confidence":"high","source_url":"https:\/\/instances.vantage.sh\/aws\/ec2\/p4de.24xlarge","note":"8x A100 80GB SXM; $27.447\/hr \/ 8 (p4de = 80GB; p4d = 40GB).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA A100 80GB","vram_gb":80,"vendor":"Google Cloud","vendor_type":"Hyperscaler","instance":"a2-ultragpu-1g","gpus_per_instance":1,"instance_usd_per_hr":5.07,"usd_per_gpu_hr":5.07,"status":"on_demand","confidence":"high","source_url":"https:\/\/gcloud-compute.com\/a2-ultragpu-1g.html","note":"A2 Ultra = A100 80GB; $5.07\/hr us-central1 (the $4.03 figure elsewhere is the 40GB A2 High-GPU).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA A100 80GB","vram_gb":80,"vendor":"Microsoft Azure","vendor_type":"Hyperscaler","instance":"NC24ads A100 v4","gpus_per_instance":1,"instance_usd_per_hr":3.67,"usd_per_gpu_hr":3.67,"status":"on_demand","confidence":"high","source_url":"https:\/\/instances.vantage.sh\/azure\/vm\/nc24ads-v4","note":"1x A100 80GB PCIe; $3.673\/hr ($0.735 elsewhere is spot).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA A100 80GB","vram_gb":80,"vendor":"Lambda Labs","vendor_type":"Neocloud","instance":"8x A100 SXM","gpus_per_instance":1,"instance_usd_per_hr":2.79,"usd_per_gpu_hr":2.79,"status":"on_demand","confidence":"high","source_url":"https:\/\/lambda.ai\/pricing","note":"A100 80GB SXM per-GPU on the 8-GPU node ($1.99 is the 40GB PCIe variant).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA A100 80GB","vram_gb":80,"vendor":"RunPod","vendor_type":"Neocloud","instance":"A100 PCIe 80GB (Secure Cloud)","gpus_per_instance":1,"instance_usd_per_hr":1.39,"usd_per_gpu_hr":1.39,"status":"on_demand","confidence":"high","source_url":"https:\/\/www.runpod.io\/pricing","note":"Secure Cloud A100 PCIe ($1.49 for A100 SXM).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA A100 80GB","vram_gb":80,"vendor":"Vast.ai","vendor_type":"Marketplace","instance":"A100 SXM4 80GB","gpus_per_instance":1,"instance_usd_per_hr":0.8,"usd_per_gpu_hr":0.8,"status":"marketplace_floor","confidence":"medium","source_url":"https:\/\/vast.ai\/pricing","note":"Representative mid-market on-demand (~$0.52-1.10 by host; verified hosts ~$0.67).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA L40S 48GB","vram_gb":48,"vendor":"AWS","vendor_type":"Hyperscaler","instance":"g6e.xlarge","gpus_per_instance":1,"instance_usd_per_hr":1.861,"usd_per_gpu_hr":1.86,"status":"on_demand","confidence":"high","source_url":"https:\/\/instances.vantage.sh\/aws\/ec2\/g6e.xlarge","note":"1x L40S; $1.861\/hr on-demand us-east-1 (8-GPU g6e.48xlarge is $3.77\/GPU with heavier bundled CPU\/RAM).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA L40S 48GB","vram_gb":48,"vendor":"RunPod","vendor_type":"Neocloud","instance":"L40S (Secure Cloud)","gpus_per_instance":1,"instance_usd_per_hr":0.99,"usd_per_gpu_hr":0.99,"status":"on_demand","confidence":"high","source_url":"https:\/\/www.runpod.io\/gpu-models\/l40s","note":"Secure Cloud on-demand ($0.79 Community Cloud).","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA L40S 48GB","vram_gb":48,"vendor":"Vast.ai","vendor_type":"Marketplace","instance":"L40S","gpus_per_instance":1,"instance_usd_per_hr":0.47,"usd_per_gpu_hr":0.47,"status":"marketplace_floor","confidence":"medium","source_url":"https:\/\/vast.ai\/pricing\/gpu\/L40S","note":"Marketplace floor; interruptible runs lower. Not offered by GCP\/Azure\/Lambda.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA L4 24GB","vram_gb":24,"vendor":"AWS","vendor_type":"Hyperscaler","instance":"g6.xlarge","gpus_per_instance":1,"instance_usd_per_hr":0.8048,"usd_per_gpu_hr":0.8,"status":"on_demand","confidence":"high","source_url":"https:\/\/instances.vantage.sh\/aws\/ec2\/g6.xlarge","note":"1x L4; $0.8048\/hr on-demand.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA L4 24GB","vram_gb":24,"vendor":"Google Cloud","vendor_type":"Hyperscaler","instance":"g2-standard-4","gpus_per_instance":1,"instance_usd_per_hr":0.7068,"usd_per_gpu_hr":0.71,"status":"on_demand","confidence":"high","source_url":"https:\/\/gcloud-compute.com\/g2-standard-4.html","note":"1x L4; $0.7068\/hr us-central1.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA L4 24GB","vram_gb":24,"vendor":"RunPod","vendor_type":"Neocloud","instance":"L4 (Secure Cloud)","gpus_per_instance":1,"instance_usd_per_hr":0.39,"usd_per_gpu_hr":0.39,"status":"on_demand","confidence":"high","source_url":"https:\/\/www.runpod.io\/gpu-models\/l4","note":"Secure Cloud on-demand, per-second billing.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA L4 24GB","vram_gb":24,"vendor":"Vast.ai","vendor_type":"Marketplace","instance":"L4","gpus_per_instance":1,"instance_usd_per_hr":0.2,"usd_per_gpu_hr":0.2,"status":"marketplace_floor","confidence":"medium","source_url":"https:\/\/vast.ai\/pricing\/gpu\/L4","note":"Marketplace floor (typical ~$0.33). Not offered by Azure\/Lambda.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA RTX 4090 24GB","vram_gb":24,"vendor":"RunPod","vendor_type":"Neocloud","instance":"RTX 4090 (Secure Cloud)","gpus_per_instance":1,"instance_usd_per_hr":0.69,"usd_per_gpu_hr":0.69,"status":"on_demand","confidence":"high","source_url":"https:\/\/www.runpod.io\/gpu-models\/rtx-4090","note":"Secure Cloud on-demand ($0.34 Community Cloud). Consumer GPU: not offered by AWS\/GCP\/Azure\/Lambda.","as_of":"2026-07-16","next_review":"2026-08-16"},{"gpu":"NVIDIA RTX 4090 24GB","vram_gb":24,"vendor":"Vast.ai","vendor_type":"Marketplace","instance":"RTX 4090","gpus_per_instance":1,"instance_usd_per_hr":0.34,"usd_per_gpu_hr":0.34,"status":"marketplace_floor","confidence":"medium","source_url":"https:\/\/vast.ai\/pricing\/gpu\/RTX-4090","note":"Marketplace floor (~$0.30-0.50 on-demand). Consumer GPU: hyperscalers do not offer it.","as_of":"2026-07-16","next_review":"2026-08-16"}]}