GPU Cloud Explorer

Thunder Compute offers the NVIDIA RTX A6000 48GB at $0.35 per GPU-hour with per-minute billing. The plan includes 100GB of running storage per GPU and no data egress fee, based on the provider’s official pricing page.

Thunder Compute offers the NVIDIA A100 80GB at $1.09 per GPU-hour with per-minute billing. The plan includes 100GB of running storage per GPU and no data egress fee, based on the provider’s official pricing page.

Runpod lists NVIDIA RTX A6000 48GB GPU Pods on Community Cloud at $0.33 per GPU-hour, compared with $0.59 per GPU-hour on Secure Cloud. Community Cloud pricing reflects a separate capacity and reliability context; actual GPU availability varies.

Thunder Compute offers the NVIDIA L40 48GB at $0.79 per GPU-hour with per-minute billing. The plan includes 100GB of running storage per GPU and no data egress fee, based on the provider’s official pricing page.

Fireworks AI lists B200 180 GB on-demand deployments at $13.00 per GPU-hour from September 1, billed per GPU second with no extra startup-time charge. Region-restricted deployments are priced at a 1.5x premium.

Fireworks AI lists H200 141 GB on-demand deployments at $8.00 per GPU-hour from September 1, billed per GPU second with no extra startup-time charge. Region-restricted deployments are priced at a 1.5x premium.

Fireworks AI lists B300 288 GB on-demand deployments at $15.00 per GPU-hour from September 1, billed per GPU second with no extra startup-time charge. Region-restricted deployments are priced at a 1.5x premium.

Fireworks AI lists H100 80 GB on-demand deployments at $8.00 per GPU-hour from September 1, billed per GPU second with no extra startup-time charge. Region-restricted deployments are priced at a 1.5x premium.

Runpod lists NVIDIA RTX A5000 24GB instances on Community Cloud at $0.16 per GPU-hour and Secure Cloud at $0.27 per GPU-hour. Runpod bills Pods per second with no minimum duration; Community Cloud is lower-cost capacity on third-party hardware, and Runpod states there are no ingress or egress charges.

Fireworks AI lists GB300 288 GB on-demand deployments at $20.00 per GPU-hour from September 1, billed per GPU second with no extra startup-time charge. Region-restricted deployments are priced at a 1.5x premium.

Together AI offers a single-tenant Dedicated Endpoint on 1× NVIDIA H100 80GB at a promotional $3.99 per GPU-hour versus a $5.49 listed rate through September 30, 2026. Dedicated endpoints are billed per minute while running and can be created and managed through Together's API, CLI, or UI.

Together AI GPU Clusters list preemptible NVIDIA HGX H100 capacity at $1.99 per GPU-hour versus $3.99 per GPU-hour on-demand. Preemptible nodes are available on Kubernetes clusters in all regions, can be reclaimed with up to a five-minute drain window, and are billed sub-hourly.

Together AI GPU Clusters list preemptible NVIDIA HGX H200 capacity at $2.99 per GPU-hour versus $5.99 per GPU-hour on-demand. Preemptible nodes are available on Kubernetes clusters in all regions, can be reclaimed with up to a five-minute drain window, and are billed sub-hourly.

Together AI GPU Clusters list preemptible NVIDIA HGX B200 capacity at $4.09 per GPU-hour versus $8.19 per GPU-hour on-demand. Preemptible nodes are available on Kubernetes clusters in all regions, can be reclaimed with up to a five-minute drain window, and are billed sub-hourly.

SaladCloud Community Cloud offers RTX 4070 12 GB at $0.150 per GPU-hour on the High Priority tier, billed per second. The September 12, 2026 pricing change reduced this tier from $0.220/hour; capacity is distributed and workloads should tolerate node reallocation.

SaladCloud Community Cloud offers RTX 3090 24 GB at $0.170 per GPU-hour on the High Priority tier, billed per second. The September 12, 2026 pricing change reduced this tier from $0.250/hour; capacity is distributed and workloads should tolerate node reallocation.

SaladCloud Community Cloud offers RTX 5080 16 GB at $0.260 per GPU-hour on the High Priority tier, billed per second. The September 12, 2026 pricing change reduced this tier from $0.420/hour; capacity runs on SaladCloud's distributed Community Cloud and workloads should tolerate node reallocation.

Hostinger lists a single NVIDIA RTX PRO 6000 (Server) Blackwell GPU instance with 96 GB VRAM from $0.65/hour. GPU compute is billed from a shared credit balance with per-minute usage, no long-term commitment, and region-dependent live availability; listed prices exclude VAT.

Runpod Secure Cloud lists NVIDIA B300 Pods with 288 GB VRAM at $8.99 per GPU-hour, billed per second. Each listed Pod provides 251 GB system RAM and 32 vCPUs; regional availability may vary.

Hostinger lists a single NVIDIA L40S GPU instance with 48 GB VRAM from $0.96/hour. GPU compute is billed from a shared credit balance with per-minute usage, no long-term commitment, and region-dependent live availability; listed prices exclude VAT.

Lambda lists a self-serve 1× NVIDIA A100 SXM GPU instance with 80 GB VRAM at $2.79 per GPU-hour. Capacity is first-come and applicable sales tax/VAT/GST may apply.

Modal lists NVIDIA RTX PRO 6000 GPU task compute at $0.000842 per second, equivalent to $3.0312 per GPU-hour at the published base rate. Region selection or execution settings can change effective cost, and the pricing page does not establish real-time GPU availability.

Hostinger lists a single NVIDIA B200 Blackwell GPU instance with 192 GB VRAM from $4.88/hour. GPU compute is billed from a shared credit balance with per-minute usage, no long-term commitment, and region-dependent live availability; listed prices exclude VAT.

DigitalOcean Paperspace lists the NVIDIA A100-80G machine with 80 GB VRAM at $3.18 per hour on demand. Machine compute charges stop when powered off, while attached storage and public IPs may continue billing; capacity is unconfirmed.

DigitalOcean Paperspace lists the NVIDIA H100 80GB machine at $5.95 per hour on demand. Machine compute billing stops when powered off, while storage and other add-ons may continue billing; access is subject to approval and capacity.

Lambda lists a self-serve 1× NVIDIA A100 SXM GPU instance with 40 GB VRAM at $1.99 per GPU-hour. Capacity is first-come and applicable sales tax/VAT/GST may apply.

Hostinger lists a single NVIDIA RTX 4090 GPU instance with 24 GB VRAM from $0.42/hour. GPU compute is billed from a shared credit balance with per-minute usage, no long-term commitment, and region-dependent live availability; listed prices exclude VAT.

Modal lists NVIDIA B300 GPU task compute at $0.001972 per second, equivalent to $7.0992 per GPU-hour at the published base rate. Region selection or execution settings can change effective cost, and the pricing page does not establish real-time GPU availability.

Lambda lists a self-serve 1× NVIDIA B200 SXM6 GPU instance at $6.69 per GPU-hour with 180 GB VRAM. Capacity is first-come and applicable sales tax/VAT/GST may apply.

Nebius NVIDIA H200 on-demand costs USD 5.40 per GPU-hour for one GPU with 141 GB VRAM; the listed configuration includes 16 vCPUs and 200 GB RAM. The platform is documented in eu-north1; eu-north2 (conditional); eu-west1; us-central1, with storage charged separately.

Nebius NVIDIA RTX PRO 6000 on-demand is USD 1.80 per GPU-hour for a 96 GB VRAM GPU, with 24 vCPUs and 218 GB RAM listed. Platform regions include us-central1, uk-south2 and eu-south1; storage is billed separately.

Nebius NVIDIA HGX H100 on-demand costs USD 4.50 per GPU-hour from October 1, 2026, in eu-north1. Each GPU has 80 GB memory and the listed configuration has 16 vCPUs and 200 GB RAM. Storage costs extra and dynamic spot pricing is separate.

Nebius NVIDIA B200 on-demand costs USD 8.50 per GPU-hour for one GPU with 180 GB VRAM; the listed configuration includes 20 vCPUs and 224 GB RAM. The platform is documented in us-central1; me-west1, with storage charged separately.

Contabo GPU VPS provides one physically dedicated NVIDIA RTX PRO 6000 Blackwell Server Edition with 96 GB VRAM, 18 AMD EPYC vCPUs, 96 GB RAM and 900 GB NVMe for €999/month excluding VAT. The current official product page advertises €799.20/month for the first 24 months on a 24-month term; Europe and US Central are offered, with 1 static IPv4, a /64 IPv6 subnet and unlimited traffic subject to fair use.

Runpod Secure Cloud lists NVIDIA B200 Pods with 180 GB VRAM at $7.99 per GPU-hour, billed per second. Each listed Pod provides 283 GB system RAM and 28 vCPUs; regional availability may vary.

Runpod lists the NVIDIA H200 141GB Pod on Secure Cloud at an advertised starting rate of USD 5.29 per GPU-hour, billed per second. The published configuration shows 24 vCPUs and 276 GB system RAM; actual region inventory and storage charges vary.

Runpod lists H100 NVL Secure Cloud GPU Pods at USD 3.19 per GPU-hour with 94 GB VRAM per GPU, 94 GB system RAM and 16 vCPUs. Pods are billed per second; advertised regional capacity may vary.

DigitalOcean's on-demand NVIDIA HGX H200 GPU Droplet costs $4.47 per GPU-hour with 141 GB GPU memory, 24 vCPUs, 240 GiB RAM, a 720 GiB NVMe boot disk, 5 TiB scratch storage and 15,000 GiB included transfer. Billing is per second with a 60-second or $0.01 minimum, and powered-off GPU Droplets continue billing until destroyed.

DigitalOcean on-demand AMD MI300X GPU Droplet is listed at USD 2.59 per GPU-hour for one 192 GB GPU, with 20 vCPUs, 240 GiB system RAM, 720 GiB NVMe boot disk, 5 TiB NVMe scratch disk and 15,000 GiB transfer. GPU Droplets bill per second, but the official pricing and documentation disagree on the minimum billable duration; powered-off Droplets remain billable until destroyed.

DigitalOcean lists its on-demand NVIDIA L40S GPU Droplet at $1.57 per GPU-hour with 48 GB GPU memory, 64 GiB instance memory, 8 vCPUs and a 500 GiB NVMe boot disk. Billing is per second, but DigitalOcean's official pricing and documentation pages disagree on the minimum billable duration (5 minutes versus 60 seconds/$0.01).

DigitalOcean's on-demand NVIDIA HGX H100 GPU Droplet costs $4.41 per GPU-hour with 80 GB VRAM, 20 vCPUs, 240 GiB RAM, a 720 GiB NVMe boot disk, 5 TiB scratch disk and 15,000 GiB included transfer. Billing is per second, but the provider's official pages conflict on whether the minimum charge covers 5 minutes or 60 seconds/$0.01.

DigitalOcean offers the NVIDIA RTX 4000 Ada Generation as an on-demand GPU Droplet at $0.76/hour. The configuration has 20 GB total GPU memory, 32 GiB Droplet memory, 8 vCPUs, and a 500 GiB NVMe boot disk; billing is per second with a 60-second or $0.01 minimum.

DigitalOcean offers an on-demand H100 8x GPU Droplet at US$35.28/hour, with 8 NVIDIA H100 GPUs (640 GB total GPU memory), 160 vCPUs, 1,920 GiB RAM and a 2,046 GiB NVMe boot disk. Billing is per second with a 60-second or US$0.01 minimum, and powered-off GPU Droplets continue to incur charges until destroyed.

DigitalOcean GPU Droplet — MI300X 8x is an on-demand GPU VM priced at $20.72/hour with 1536 GB total GPU memory. Billing is per second with a 60-second or $0.01 minimum; powered-off GPU Droplets continue billing until destroyed.

DigitalOcean lists its on-demand NVIDIA RTX 6000 Ada GPU Droplet at $1.57 per GPU-hour with 48 GB GPU memory, 64 GiB instance memory, 8 vCPUs and a 500 GiB NVMe boot disk. Billing is per second, but DigitalOcean's official pricing and documentation pages disagree on the minimum billable duration (5 minutes versus 60 seconds/$0.01).

DigitalOcean offers the NVIDIA H200 8x configuration as an on-demand GPU Droplet at $35.76/hour. The configuration has 1128 GB total GPU memory, 1920 GiB Droplet memory, 192 vCPUs, and a 2046 GiB NVMe boot disk; billing is per second with a 60-second or $0.01 minimum.

AWS EC2 Capacity Blocks prices p6-b300.48xlarge at $129.168 per instance-hour ($16.146 per B300 accelerator-hour) effective October 7, 2026, with 8 NVIDIA B300 GPUs and 2.1 TB aggregate HBM. Capacity Blocks are reserved for a defined duration and charged at the prevailing reservation rate at purchase; OS charges can apply separately.

Modal lists NVIDIA A100 80GB at USD 0.000694/GPU-second, equivalent to USD 2.4984/GPU-hour at its base serverless GPU task rate, with 80 GB VRAM. CPU, memory, network egress, region selection and non-preemptible settings can add charges; real-time GPU availability is not established by the pricing page.

Modal lists NVIDIA H100 SXM5 GPU allocation at USD $0.001097 per GPU-second (equivalent to $3.9492 per GPU-hour) with 80 GB of VRAM. CPU and memory are billed separately; region selection and non-preemptible execution may add surcharges, and actual regional capacity is not independently checked.

Modal lists NVIDIA L40S at USD 0.000542/GPU-second, equivalent to USD 1.9512/GPU-hour at its base serverless GPU task rate, with 48 GB VRAM. CPU, memory, network egress, region selection and non-preemptible settings can add charges; real-time GPU availability is not established by the pricing page.

Modal lists NVIDIA H200 SXM at USD 0.001261/GPU-second, equivalent to USD 4.5396/GPU-hour at its base serverless GPU task rate, with 141 GB VRAM. CPU, memory, network egress, region selection and non-preemptible settings can add charges; real-time GPU availability is not established by the pricing page.

Modal lists NVIDIA B200 GPU task compute at $0.001736 per GPU-second (equivalent to $6.2496 per GPU-hour) for a single 180 GB B200 GPU. This is a base GPU-only rate; CPU, RAM, egress and optional region or non-preemptible multipliers may increase total cost, and real-time stock is not confirmed.

CloudRift lists a 1x NVIDIA A100 SXM4 (80 GB VRAM) on-demand GPU instance at USD 1.05 per GPU-hour. Rates and deployment availability may vary by region and demand; reserved commitments have separate prices.

CloudRift lists a 1x NVIDIA RTX 4090 (24 GB VRAM) on-demand GPU instance at USD 0.39 per GPU-hour. Rates and deployment availability may vary by region and demand; reserved commitments have separate prices.

Modal lists NVIDIA A100 40GB (40 GB VRAM) serverless GPU tasks at USD 0.000583 per GPU-second, equivalent to USD 2.0988 per GPU-hour for the GPU alone. CPU, memory, and egress may be billed separately; region selection and non-preemptible execution can increase rates.

Modal offers NVIDIA L4 24 GB GPU allocations for serverless Functions at a base rate of $0.000222 per GPU-second ($0.7992 per GPU-hour). CPU, memory, network egress beyond plan allowances, region selection and non-preemptible execution may add charges; real-time GPU availability is not established by the pricing page.

Modal lists NVIDIA A10 (24 GB VRAM) serverless GPU tasks at USD 0.000306 per GPU-second, equivalent to USD 1.1016 per GPU-hour for the GPU alone. CPU, memory, and egress may be billed separately; region selection and non-preemptible execution can increase rates.

Modal lists NVIDIA T4 (16 GB VRAM) serverless GPU tasks at USD 0.000164 per GPU-second, equivalent to USD 0.5904 per GPU-hour for the GPU alone. CPU, memory, and egress may be billed separately; region selection and non-preemptible execution can increase rates.

CloudRift lists a 1x NVIDIA V100 SXM2 (16 GB VRAM) on-demand GPU instance at USD 0.25 per GPU-hour. Rates and deployment availability may vary by region and demand; reserved commitments have separate prices.