| Provider | Offering | GPU | Form | GPUs | Interconnect | Region | Term | USD / GPU-hour | Scraped (UTC) | Confidence |
|---|---|---|---|---|---|---|---|---|---|---|
| RunPod | rental_unspecified | H200 | NVL | not stated | NVLink | not stated | on_demand | $0.50 RunPod Community Cloud gpuTypes.communityPrice. Host marketplace price on RunPod. | 2026-10-21T23:09:28Z | medium |
| Nebius | vm | H100 | unknown | not stated | unknown | not stated | spot | $0.79 Preemptible column is labelled from, so confidence is low. | 2026-10-21T23:09:17Z | low |
| Nebius | vm | H200 | unknown | not stated | unknown | not stated | spot | $0.79 Preemptible column is labelled from, so confidence is low. | 2026-10-21T23:09:17Z | low |
| Nebius | vm | B200 | unknown | not stated | unknown | not stated | spot | $0.99 Preemptible column is labelled from, so confidence is low. | 2026-10-21T23:09:17Z | low |
| Azure | vm | H100 | PCIe | 1 | unknown | eastus | spot | $1.2899 eastus retail price. NC40adsH100v5 Spot. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | PCIe | 2 | unknown | eastus | spot | $1.2899 eastus retail price. NC80adisH100v5 Spot. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| latitude via Shadeform | vm | H100 | PCIe | 1 | none | US, Dallas, TX | on_demand | $1.66 via Shadeform. Underlying cloud: latitude. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| latitude via Shadeform | vm | H100 | PCIe | 1 | none | US, Ashburn, VA | on_demand | $1.66 via Shadeform. Underlying cloud: latitude. Region: US, Ashburn, VA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Latitude.sh | bare_metal | H100 | unknown | 1 | unknown | not stated | on_demand | $1.68 Latitude.sh says Starting At. Hourly node price divided by the stated GPU count. A monthly figure on the same card was not used. | 2026-10-21T23:10:31Z | low |
| Vast.ai | rental_unspecified | H100 | SXM | not stated | unknown | not stated | on_demand | $1.73 Read from vast.ai/pricing. The public JSON feed was not used because its license restricts index redistribution. | 2026-10-21T23:09:31Z | medium |
| Hyperstack | vm | H100 | unknown | not stated | unknown | not stated | unknown | $1.75 Starting from Hyperstack reservation price. The page does not state the commitment length, so term is unknown. | 2026-10-21T23:09:40Z | low |
| latitude via Shadeform | vm | H100 | NVL | 8 | NVLink | US, Dallas, TX | on_demand | $1.7925 via Shadeform. Underlying cloud: latitude. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Hyperstack | vm | H100 | unknown | not stated | NVLink | not stated | unknown | $1.82 Starting from Hyperstack reservation price. The page does not state the commitment length, so term is unknown. | 2026-10-21T23:09:40Z | low |
| horizon via Shadeform | vm | H100 | SXM | 8 | none | IS, Blonduos 2 | on_demand | $1.95 via Shadeform. Underlying cloud: horizon. Region: IS, Blonduos 2. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| RunPod | rental_unspecified | H100 | PCIe | not stated | unknown | not stated | on_demand | $1.99 RunPod Community Cloud gpuTypes.communityPrice. Host marketplace price on RunPod. | 2026-10-21T23:09:28Z | medium |
| Voltage Park | rental_unspecified | H100 | unknown | not stated | InfiniBand | not stated | on_demand | $1.99 Page says Starting at for NVIDIA HGX H100. GPU count is not on the price line. SXM is mentioned in a separate hardware sentence, so form factor is unknown. The hourly figure is used as printed and is not divided. | 2026-10-21T23:09:51Z | low |
| latitude via Shadeform | vm | H100 | PCIe | 1 | none | US, Dallas, TX | on_demand | $1.99 via Shadeform. Underlying cloud: latitude. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| voltagepark via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Dallas, TX | on_demand | $1.99 via Shadeform. Underlying cloud: voltagepark. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| voltagepark via Shadeform | vm | H100 | SXM | 4 | NVLink | US, Dallas, TX | on_demand | $1.99 via Shadeform. Underlying cloud: voltagepark. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| voltagepark via Shadeform | vm | H100 | SXM | 2 | NVLink | US, Dallas, TX | on_demand | $1.99 via Shadeform. Underlying cloud: voltagepark. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| voltagepark via Shadeform | vm | H100 | SXM | 1 | none | US, Dallas, TX | on_demand | $1.99 via Shadeform. Underlying cloud: voltagepark. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H100 | NVL | 8 | NVLink | not stated | spot | $2.00 via Shadeform. spot_hourly_price in US cents. Underlying cloud: nebius. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H100 | SXM | 1 | none | not stated | spot | $2.00 via Shadeform. spot_hourly_price in US cents. Underlying cloud: nebius. | 2026-10-21T23:09:43Z | medium |
| Azure | vm | H100 | unknown | 8 | unknown | eastus | spot | $2.0441 eastus retail price. ND96isH100v5 Spot. GPU count 8 is numberOfGPUs for STANDARD_ND96IS_H100_V5 on the Azure Machine Learning managed online endpoints VM SKU list. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| Datacrunch | rental_unspecified | H100 | SXM | 1 | unknown | not stated | spot | $2.213 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x H100 SXM5 80GB, 30 CPUs, 120 GB RAM, 80 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Datacrunch | rental_unspecified | H100 | SXM | 1 | unknown | not stated | spot | $2.213 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x H100 SXM5 80GB, 32 CPUs, 170 GB RAM, 80 GB VRAM. | 2026-10-21T23:10:05Z | high |
| latitude via Shadeform | vm | H100 | PCIe | 4 | NVLink | US, Dallas, TX | on_demand | $2.2425 via Shadeform. Underlying cloud: latitude. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| TensorDock | vm | H100 | SXM | not stated | unknown | not stated | on_demand | $2.25 TensorDock says From. GPU count is not stated. Used as a per-GPU from-price because the line is a single GPU card, not a node total. | 2026-10-21T23:09:54Z | low |
| TensorDock | vm | H100 | unknown | not stated | unknown | not stated | on_demand | $2.25 TensorDock says From. GPU count is not stated. Used as a per-GPU from-price because the line is a single GPU card, not a node total. | 2026-10-21T23:09:54Z | low |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Raleigh, NC | on_demand | $2.25 via Shadeform. Underlying cloud: excesssupply. Region: US, Raleigh, NC. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Dallas, TX | on_demand | $2.25 via Shadeform. Underlying cloud: excesssupply. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Azure | vm | H100 | unknown | 8 | NVLink | eastus | spot | $2.2712 eastus retail price. ND96isrH100v5 Spot. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| Datacrunch | serverless | H100 | SXM | 1 | unknown | not stated | spot | $2.43 Datacrunch pricing page, now published on Verda. Serverless containers section. The page says auto-scaling containers with pay-per-use. Row: 1x H100 SXM5 80GB, 21 CPUs, 160 GB RAM, 80 GB VRAM. | 2026-10-21T23:10:05Z | high |
| CoreWeave | bare_metal | H100 | unknown | 8 | unknown | EUROPE | spot | $2.4388 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| boostrun via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Raleigh, NC | on_demand | $2.45 via Shadeform. Underlying cloud: boostrun. Region: US, Raleigh, NC. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| boostrun via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Dallas, TX | on_demand | $2.45 via Shadeform. Underlying cloud: boostrun. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| CoreWeave | bare_metal | H100 | unknown | 8 | unknown | NORTH AMERICA | spot | $2.4638 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| imwt via Shadeform | vm | H100 | PCIe | 8 | none | US, Des Moines, IA | on_demand | $2.48 via Shadeform. Underlying cloud: imwt. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| imwt via Shadeform | vm | H100 | PCIe | 2 | none | US, Des Moines, IA | on_demand | $2.48 via Shadeform. Underlying cloud: imwt. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| imwt via Shadeform | vm | H100 | PCIe | 4 | none | US, Des Moines, IA | on_demand | $2.48 via Shadeform. Underlying cloud: imwt. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| imwt via Shadeform | vm | H100 | PCIe | 1 | none | US, Des Moines, IA | on_demand | $2.48 via Shadeform. Underlying cloud: imwt. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Civo | vm | H100 | SXM | 1 | unknown | not stated | 36mo | $2.49 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H100 | SXM | 8 | unknown | not stated | 36mo | $2.49 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Hyperstack | vm | H100 | unknown | not stated | unknown | not stated | on_demand | $2.50 Hyperstack on-demand table, price per GPU per hour. Node size is not a single number on the row. | 2026-10-21T23:09:40Z | high |
| denvr via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Houston, TX | on_demand | $2.50 via Shadeform. Underlying cloud: denvr. Region: US, Houston, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| hyperstack via Shadeform | vm | H100 | PCIe | 8 | none | CA, Montreal | on_demand | $2.50 via Shadeform. Underlying cloud: hyperstack. Region: CA, Montreal. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| hyperstack via Shadeform | vm | H100 | PCIe | 4 | none | CA, Montreal | on_demand | $2.50 via Shadeform. Underlying cloud: hyperstack. Region: CA, Montreal. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| hyperstack via Shadeform | vm | H100 | PCIe | 1 | none | CA, Montreal | on_demand | $2.50 via Shadeform. Underlying cloud: hyperstack. Region: CA, Montreal. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| hyperstack via Shadeform | vm | H100 | PCIe | 2 | none | CA, Montreal | on_demand | $2.50 via Shadeform. Underlying cloud: hyperstack. Region: CA, Montreal. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Vast.ai | rental_unspecified | H100 | NVL | not stated | unknown | not stated | on_demand | $2.52 Read from vast.ai/pricing. The public JSON feed was not used because its license restricts index redistribution. | 2026-10-21T23:09:31Z | medium |
| CoreWeave | bare_metal | H200 | unknown | 8 | unknown | EUROPE | spot | $2.58 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| Civo | vm | H100 | SXM | 1 | unknown | not stated | 24mo | $2.59 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H100 | SXM | 8 | unknown | not stated | 24mo | $2.59 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| RunPod | rental_unspecified | H100 | NVL | not stated | NVLink | not stated | on_demand | $2.59 RunPod Community Cloud gpuTypes.communityPrice. Host marketplace price on RunPod. | 2026-10-21T23:09:28Z | medium |
| Hyperstack | vm | H100 | unknown | not stated | NVLink | not stated | on_demand | $2.60 Hyperstack on-demand table, price per GPU per hour. Node size is not a single number on the row. | 2026-10-21T23:09:40Z | high |
| hyperstack via Shadeform | vm | H100 | NVL | 8 | NVLink | CA, Montreal | on_demand | $2.60 via Shadeform. Underlying cloud: hyperstack. Region: CA, Montreal. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| CoreWeave | bare_metal | H200 | unknown | 8 | unknown | NORTH AMERICA | spot | $2.6162 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| Civo | vm | H100 | SXM | 1 | unknown | not stated | 12mo | $2.69 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H100 | SXM | 8 | unknown | not stated | 12mo | $2.69 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| RunPod | rental_unspecified | H100 | SXM | not stated | unknown | not stated | on_demand | $2.69 RunPod Community Cloud gpuTypes.communityPrice. Host marketplace price on RunPod. | 2026-10-21T23:09:28Z | medium |
| Hyperstack | vm | H100 | SXM | not stated | unknown | not stated | unknown | $2.72 Starting from Hyperstack reservation price. The page does not state the commitment length, so term is unknown. | 2026-10-21T23:09:40Z | low |
| massedcompute via Shadeform | vm | H100 | PCIe | 8 | none | US, Des Moines, IA | on_demand | $2.7275 via Shadeform. Underlying cloud: massedcompute. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | PCIe | 8 | none | US, Kansas City, KS | on_demand | $2.7275 via Shadeform. Underlying cloud: massedcompute. Region: US, Kansas City, KS. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | PCIe | 4 | none | US, Des Moines, IA | on_demand | $2.7275 via Shadeform. Underlying cloud: massedcompute. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | PCIe | 4 | none | US, Kansas City, KS | on_demand | $2.7275 via Shadeform. Underlying cloud: massedcompute. Region: US, Kansas City, KS. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Massed Compute | rental_unspecified | H100 | unknown | 1 | unknown | not stated | on_demand | $2.73 Massed Compute node price divided by the GPU count on the row. | 2026-10-21T23:10:25Z | high |
| massedcompute via Shadeform | vm | H100 | PCIe | 2 | none | US, Des Moines, IA | on_demand | $2.73 via Shadeform. Underlying cloud: massedcompute. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | PCIe | 2 | none | US, Kansas City, KS | on_demand | $2.73 via Shadeform. Underlying cloud: massedcompute. Region: US, Kansas City, KS. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | PCIe | 1 | none | US, Des Moines, IA | on_demand | $2.73 via Shadeform. Underlying cloud: massedcompute. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | PCIe | 1 | none | US, Kansas City, KS | on_demand | $2.73 via Shadeform. Underlying cloud: massedcompute. Region: US, Kansas City, KS. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Civo | vm | H100 | SXM | 1 | unknown | not stated | 6mo | $2.79 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H100 | SXM | 8 | unknown | not stated | 6mo | $2.79 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Hyperstack | vm | H200 | SXM | not stated | unknown | not stated | unknown | $2.79 Starting from Hyperstack reservation price. The page does not state the commitment length, so term is unknown. | 2026-10-21T23:09:40Z | low |
| nebius via Shadeform | vm | H200 | SXM | 1 | none | not stated | spot | $2.80 via Shadeform. spot_hourly_price in US cents. Underlying cloud: nebius. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H200 | NVL | 8 | NVLink | not stated | spot | $2.80 via Shadeform. spot_hourly_price in US cents. Underlying cloud: nebius. | 2026-10-21T23:09:43Z | medium |
| Datacrunch | rental_unspecified | H200 | SXM | 1 | unknown | not stated | spot | $2.886 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x H200 SXM5 141GB, 44 CPUs, 170 GB RAM, 141 GB VRAM. | 2026-10-21T23:10:05Z | high |
| RunPod | rental_unspecified | H100 | PCIe | not stated | unknown | not stated | on_demand | $2.89 RunPod Secure Cloud gpuTypes.securePrice. | 2026-10-21T23:09:28Z | high |
| DeepInfra | managed_inference | H100 | unknown | not stated | unknown | not stated | on_demand | $2.90 DeepInfra dedicated GPU price. Form factor is not stated. | 2026-10-21T23:10:28Z | high |
| Massed Compute | rental_unspecified | H100 | NVL | 2 | NVLink | not stated | on_demand | $2.92 Massed Compute node price divided by the GPU count on the row. | 2026-10-21T23:10:25Z | high |
| Vast.ai | rental_unspecified | H100 | PCIe | not stated | unknown | not stated | on_demand | $2.95 Read from vast.ai/pricing. The public JSON feed was not used because its license restricts index redistribution. | 2026-10-21T23:09:31Z | medium |
| Civo | vm | H100 | SXM | 1 | unknown | not stated | on_demand | $2.99 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H100 | SXM | 8 | unknown | not stated | on_demand | $2.99 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H200 | SXM | 1 | unknown | not stated | 36mo | $2.99 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H200 | SXM | 8 | unknown | not stated | 36mo | $2.99 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| OVHcloud | rental_unspecified | H100 | PCIe | not stated | unknown | not stated | on_demand | $2.99 OVHcloud says From. GPU count is not in this sentence, so the hourly figure was not divided. | 2026-10-21T23:10:28Z | low |
| horizon via Shadeform | vm | H200 | SXM | 8 | none | IS, Blonduos 3 | on_demand | $3.00 via Shadeform. Underlying cloud: horizon. Region: IS, Blonduos 3. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Civo | vm | H200 | SXM | 1 | unknown | not stated | 24mo | $3.09 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H200 | SXM | 8 | unknown | not stated | 24mo | $3.09 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Massed Compute | rental_unspecified | H100 | NVL | 1 | NVLink | not stated | on_demand | $3.11 Massed Compute node price divided by the GPU count on the row. | 2026-10-21T23:10:25Z | high |
| Massed Compute | rental_unspecified | H100 | SXM | 1 | unknown | not stated | on_demand | $3.14 Massed Compute node price divided by the GPU count on the row. | 2026-10-21T23:10:25Z | high |
| massedcompute via Shadeform | vm | H100 | NVL | 8 | NVLink | US, Des Moines, IA | on_demand | $3.14 via Shadeform. Underlying cloud: massedcompute. Region: US, Des Moines, IA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | NVL | 8 | NVLink | US, Kansas City, MO | on_demand | $3.14 via Shadeform. Underlying cloud: massedcompute. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | NVL | 8 | NVLink | US, Baltimore, MD | on_demand | $3.14 via Shadeform. Underlying cloud: massedcompute. Region: US, Baltimore, MD. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | NVL | 8 | NVLink | US, Denver, CO | on_demand | $3.14 via Shadeform. Underlying cloud: massedcompute. Region: US, Denver, CO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| massedcompute via Shadeform | vm | H100 | NVL | 8 | NVLink | US, Kansas City, KS | on_demand | $3.14 via Shadeform. Underlying cloud: massedcompute. Region: US, Kansas City, KS. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Datacrunch | serverless | H200 | SXM | 1 | unknown | not stated | spot | $3.17 Datacrunch pricing page, now published on Verda. Serverless containers section. The page says auto-scaling containers with pay-per-use. Row: 1x H200 SXM5 141GB, 21 CPUs, 160 GB RAM, 141 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Civo | vm | H200 | SXM | 1 | unknown | not stated | 12mo | $3.19 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H200 | SXM | 8 | unknown | not stated | 12mo | $3.19 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| RunPod | rental_unspecified | H100 | NVL | not stated | NVLink | not stated | on_demand | $3.19 RunPod Secure Cloud gpuTypes.securePrice. | 2026-10-21T23:09:28Z | high |
| Hyperstack | vm | H100 | SXM | not stated | unknown | not stated | on_demand | $3.20 Hyperstack on-demand table, price per GPU per hour. Node size is not a single number on the row. | 2026-10-21T23:09:40Z | high |
| Vast.ai | rental_unspecified | H200 | NVL | not stated | unknown | not stated | on_demand | $3.20 Read from vast.ai/pricing. The public JSON feed was not used because its license restricts index redistribution. | 2026-10-21T23:09:31Z | medium |
| hyperstack via Shadeform | vm | H100 | SXM | 8 | none | US, Houston, TX | on_demand | $3.20 via Shadeform. Underlying cloud: hyperstack. Region: US, Houston, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| hyperstack via Shadeform | vm | H100 | SXM | 8 | none | CA, Montreal | on_demand | $3.20 via Shadeform. Underlying cloud: hyperstack. Region: CA, Montreal. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Scaleway | rental_unspecified | H100 | PCIe | 2 | unknown | fr-par-2 | on_demand | $3.2202 Scaleway catalog price. isApproximation is true, so this is not treated as a firm list price. converted from EUR at ECB USD per 1 EUR = 1.1234 | 2026-10-21T23:10:17Z | low |
| Scaleway | rental_unspecified | H100 | PCIe | 1 | unknown | fr-par-2 | on_demand | $3.2202 Scaleway catalog price. isApproximation is true, so this is not treated as a firm list price. converted from EUR at ECB USD per 1 EUR = 1.1234 | 2026-10-21T23:10:17Z | low |
| Scaleway | rental_unspecified | H100 | PCIe | 2 | unknown | pl-waw-2 | on_demand | $3.2202 Scaleway catalog price. isApproximation is true, so this is not treated as a firm list price. converted from EUR at ECB USD per 1 EUR = 1.1234 | 2026-10-21T23:10:17Z | low |
| Scaleway | rental_unspecified | H100 | PCIe | 1 | unknown | pl-waw-2 | on_demand | $3.2202 Scaleway catalog price. isApproximation is true, so this is not treated as a firm list price. converted from EUR at ECB USD per 1 EUR = 1.1234 | 2026-10-21T23:10:17Z | low |
| Paperspace/DigitalOcean | vm | H100 | unknown | 1 | unknown | not stated | 12mo | $3.26 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| Paperspace/DigitalOcean | vm | H100 | unknown | 8 | unknown | not stated | 12mo | $3.26 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| Civo | vm | H200 | SXM | 1 | unknown | not stated | 6mo | $3.29 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H200 | SXM | 8 | unknown | not stated | 6mo | $3.29 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Lambda | vm | H100 | PCIe | 1 | unknown | not stated | on_demand | $3.29 Lambda tab 1x. | 2026-10-21T23:09:07Z | high |
| scaleway via Shadeform | vm | H100 | PCIe | 2 | none | FR, Paris | on_demand | $3.30 via Shadeform. Underlying cloud: scaleway. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| scaleway via Shadeform | vm | H100 | PCIe | 2 | none | PL, Warsaw | on_demand | $3.30 via Shadeform. Underlying cloud: scaleway. Region: PL, Warsaw. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| scaleway via Shadeform | vm | H100 | PCIe | 1 | none | PL, Warsaw | on_demand | $3.30 via Shadeform. Underlying cloud: scaleway. Region: PL, Warsaw. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| scaleway via Shadeform | vm | H100 | PCIe | 1 | none | FR, Paris | on_demand | $3.30 via Shadeform. Underlying cloud: scaleway. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| verda via Shadeform | vm | H100 | SXM | 1 | none | FI, Central | on_demand | $3.31 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Paperspace/DigitalOcean | vm | H200 | unknown | 1 | unknown | not stated | 12mo | $3.40 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| Paperspace/DigitalOcean | vm | H200 | unknown | 8 | unknown | not stated | 12mo | $3.40 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| Civo | vm | H200 | SXM | 1 | unknown | not stated | on_demand | $3.49 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | H200 | SXM | 8 | unknown | not stated | on_demand | $3.49 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| DeepInfra | managed_inference | H200 | unknown | not stated | unknown | not stated | on_demand | $3.49 DeepInfra dedicated GPU price. Form factor is not stated. | 2026-10-21T23:10:28Z | high |
| Jarvislabs | vm | H100 | SXM | 1 | unknown | not stated | on_demand | $3.49 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| Jarvislabs | vm | H100 | SXM | 2 | unknown | not stated | on_demand | $3.49 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| Jarvislabs | vm | H100 | SXM | 4 | unknown | not stated | on_demand | $3.49 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| Jarvislabs | vm | H100 | SXM | 8 | unknown | not stated | on_demand | $3.49 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| Scaleway | rental_unspecified | H100 | SXM | 8 | NVLink | fr-par-2 | on_demand | $3.5571 Scaleway catalog price. isApproximation is true, so this is not treated as a firm list price. converted from EUR at ECB USD per 1 EUR = 1.1234 | 2026-10-21T23:10:17Z | low |
| Scaleway | rental_unspecified | H100 | SXM | 4 | NVLink | fr-par-2 | on_demand | $3.5867 Scaleway catalog price. isApproximation is true, so this is not treated as a firm list price. converted from EUR at ECB USD per 1 EUR = 1.1234 | 2026-10-21T23:10:17Z | low |
| RunPod | rental_unspecified | H200 | SXM | not stated | unknown | not stated | on_demand | $3.59 RunPod Community Cloud gpuTypes.communityPrice. Host marketplace price on RunPod. | 2026-10-21T23:09:28Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 8 | NVLink | not stated | spot | $3.60 via Shadeform. spot_hourly_price in US cents. Underlying cloud: excesssupply. | 2026-10-21T23:09:43Z | medium |
| Massed Compute | rental_unspecified | H200 | NVL | 1 | NVLink | not stated | on_demand | $3.62 Massed Compute node price divided by the GPU count on the row. | 2026-10-21T23:10:25Z | high |
| Scaleway | rental_unspecified | H100 | SXM | 2 | NVLink | fr-par-2 | on_demand | $3.7183 Scaleway catalog price. isApproximation is true, so this is not treated as a firm list price. converted from EUR at ECB USD per 1 EUR = 1.1234 | 2026-10-21T23:10:17Z | low |
| boostrun via Shadeform | vm | B200 | NVL | 8 | NVLink | US, Dallas, TX | on_demand | $3.74 via Shadeform. Underlying cloud: boostrun. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| boostrun via Shadeform | vm | B200 | NVL | 8 | NVLink | US, Charlotte, NC | on_demand | $3.74 via Shadeform. Underlying cloud: boostrun. Region: US, Charlotte, NC. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Civo | vm | B200 | unknown | 8 | unknown | not stated | 36mo | $3.79 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| RunPod | rental_unspecified | H200 | NVL | not stated | NVLink | not stated | on_demand | $3.79 RunPod Secure Cloud gpuTypes.securePrice. | 2026-10-21T23:09:28Z | high |
| Massed Compute | rental_unspecified | H200 | NVL | 4 | NVLink | not stated | on_demand | $3.82 Massed Compute node price divided by the GPU count on the row. | 2026-10-21T23:10:25Z | high |
| Azure | vm | H100 | PCIe | 1 | unknown | eastus | 36mo | $3.839 eastus retail price. NC40adsH100v5. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | PCIe | 2 | unknown | eastus | 36mo | $3.839 eastus retail price. NC80adisH100v5. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| nebius via Shadeform | vm | H100 | NVL | 8 | NVLink | FI, Helsinki | on_demand | $3.8662 via Shadeform. Underlying cloud: nebius. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Crusoe | vm | H100 | unknown | not stated | unknown | not stated | on_demand | $3.90 Crusoe instance price, already per GPU-hour. HGX is not recorded as SXM. | 2026-10-21T23:09:18Z | high |
| Modal | serverless | H100 | SXM | not stated | unknown | not stated | on_demand | $3.9492 Modal GPU task price. Converted from per second to per hour. Node size is not stated. | 2026-10-21T23:09:29Z | high |
| nebius via Shadeform | vm | H100 | SXM | 1 | none | FI, Helsinki | on_demand | $3.98 via Shadeform. Underlying cloud: nebius. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Civo | vm | B200 | unknown | 8 | unknown | not stated | 24mo | $3.99 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Hyperstack | vm | H200 | SXM | not stated | unknown | not stated | on_demand | $3.99 Hyperstack on-demand table, price per GPU per hour. Node size is not a single number on the row. | 2026-10-21T23:09:40Z | high |
| Lambda | vm | H100 | SXM | 8 | unknown | not stated | on_demand | $3.99 Lambda tab 8x. | 2026-10-21T23:09:07Z | high |
| RunPod | rental_unspecified | H100 | SXM | not stated | unknown | not stated | on_demand | $3.99 RunPod Secure Cloud gpuTypes.securePrice. | 2026-10-21T23:09:28Z | high |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | IN, Mumbai | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: IN, Mumbai. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Chicago, IL | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Chicago, IL. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Phoenix, AZ | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Phoenix, AZ. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, San Jose, Ca | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, San Jose, Ca. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Dulles, VA | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dulles, VA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Dallas, TX | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Atlanta, GA | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Atlanta, GA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Austin, TX | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Austin, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 8 | NVLink | US, Salt Lake City, UT | on_demand | $3.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Salt Lake City, UT. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 8 | NVLink | JP, Kagawa | on_demand | $4.00 via Shadeform. Underlying cloud: excesssupply. Region: JP, Kagawa. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Lambda | vm | H100 | SXM | 4 | unknown | not stated | on_demand | $4.09 Lambda tab 4x. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | H100 | SXM | 4 | none | US, Dallas, TX | on_demand | $4.09 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 4 | none | US, Austin, TX | on_demand | $4.09 via Shadeform. Underlying cloud: lambdalabs. Region: US, Austin, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 4 | none | US, Atlanta, GA | on_demand | $4.09 via Shadeform. Underlying cloud: lambdalabs. Region: US, Atlanta, GA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Datacrunch | rental_unspecified | B200 | SXM | 1 | unknown | not stated | spot | $4.095 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x B200 SXM6 180GB, 30 CPUs, 170 GB RAM, 180 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Vast.ai | rental_unspecified | H200 | unknown | not stated | unknown | not stated | on_demand | $4.16 Read from vast.ai/pricing. The public JSON feed was not used because its license restricts index redistribution. | 2026-10-21T23:09:31Z | medium |
| Datacrunch | rental_unspecified | B200 | SXM | 1 | unknown | not stated | spot | $4.177 Datacrunch pricing page, now published on Verda. Confidential computing section. The page says data and the model stay encrypted in memory, in transit, and at rest. It does not say virtual machine or bare metal. Row: 1x B200 CC SXM6 180GB, 30 CPUs, 170 GB RAM, 180 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Lambda | vm | H100 | SXM | 2 | unknown | not stated | on_demand | $4.19 Lambda tab 2x. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | H100 | SXM | 2 | none | US, Atlanta, GA | on_demand | $4.19 via Shadeform. Underlying cloud: lambdalabs. Region: US, Atlanta, GA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 2 | none | US, Austin, TX | on_demand | $4.19 via Shadeform. Underlying cloud: lambdalabs. Region: US, Austin, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 2 | none | US, Dallas, TX | on_demand | $4.19 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| verda via Shadeform | vm | H200 | SXM | 4 | NVLink | FI, Central | on_demand | $4.205 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| verda via Shadeform | vm | H200 | SXM | 8 | NVLink | FI, Central | on_demand | $4.205 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| CoreWeave | bare_metal | B200 | unknown | 8 | unknown | NORTH AMERICA | spot | $4.2637 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| Crusoe | vm | H200 | unknown | not stated | unknown | not stated | on_demand | $4.29 Crusoe instance price, already per GPU-hour. HGX is not recorded as SXM. | 2026-10-21T23:09:18Z | high |
| Lambda | vm | H100 | SXM | 1 | unknown | not stated | on_demand | $4.29 Lambda tab 1x. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | H100 | PCIe | 1 | none | IN, Mumbai | on_demand | $4.29 via Shadeform. Underlying cloud: lambdalabs. Region: IN, Mumbai. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | PCIe | 1 | none | US, Chicago, IL | on_demand | $4.29 via Shadeform. Underlying cloud: lambdalabs. Region: US, Chicago, IL. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | PCIe | 1 | none | US, Phoenix, AZ | on_demand | $4.29 via Shadeform. Underlying cloud: lambdalabs. Region: US, Phoenix, AZ. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | PCIe | 1 | none | US, San Jose, Ca | on_demand | $4.29 via Shadeform. Underlying cloud: lambdalabs. Region: US, San Jose, Ca. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | PCIe | 1 | none | US, Dulles, VA | on_demand | $4.29 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dulles, VA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | PCIe | 1 | none | US, Dallas, TX | on_demand | $4.29 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | PCIe | 1 | none | US, Salt Lake City, UT | on_demand | $4.29 via Shadeform. Underlying cloud: lambdalabs. Region: US, Salt Lake City, UT. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| CoreWeave | bare_metal | B200 | unknown | 8 | unknown | EUROPE | spot | $4.3587 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | H100 | SXM | 1 | none | US, Dallas, TX | on_demand | $4.36 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dallas, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 1 | none | US, Atlanta, GA | on_demand | $4.36 via Shadeform. Underlying cloud: lambdalabs. Region: US, Atlanta, GA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | H100 | SXM | 1 | none | US, Austin, TX | on_demand | $4.36 via Shadeform. Underlying cloud: lambdalabs. Region: US, Austin, TX. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Paperspace/DigitalOcean | vm | H100 | unknown | 1 | unknown | not stated | on_demand | $4.41 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| Paperspace/DigitalOcean | vm | H100 | unknown | 8 | unknown | not stated | on_demand | $4.41 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| digitalocean via Shadeform | vm | H100 | SXM | 8 | NVLink | CA, Toronto | on_demand | $4.41 via Shadeform. Underlying cloud: digitalocean. Region: CA, Toronto. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H100 | SXM | 8 | NVLink | US, New York, NY | on_demand | $4.41 via Shadeform. Underlying cloud: digitalocean. Region: US, New York, NY. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H100 | SXM | 8 | NVLink | NL, Amsterdam | on_demand | $4.41 via Shadeform. Underlying cloud: digitalocean. Region: NL, Amsterdam. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H100 | SXM | 1 | none | US, New York, NY | on_demand | $4.41 via Shadeform. Underlying cloud: digitalocean. Region: US, New York, NY. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H100 | SXM | 1 | none | CA, Toronto | on_demand | $4.41 via Shadeform. Underlying cloud: digitalocean. Region: CA, Toronto. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H100 | SXM | 1 | none | NL, Amsterdam | on_demand | $4.41 via Shadeform. Underlying cloud: digitalocean. Region: NL, Amsterdam. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| verda via Shadeform | vm | H200 | SXM | 1 | none | FI, Central | on_demand | $4.41 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Azure | vm | H100 | unknown | 8 | unknown | eastus | unknown | $4.4244 5 Years eastus retail price. ND96isH100v5. GPU count 8 is numberOfGPUs for STANDARD_ND96IS_H100_V5 on the Azure Machine Learning managed online endpoints VM SKU list. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 5 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 43800 hours. 5 Years is not a fixed bucket in the term list, so term is unknown. | 2026-10-21T23:09:07Z | high |
| Datacrunch | rental_unspecified | H100 | SXM | 1 | unknown | not stated | on_demand | $4.426 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x H100 SXM5 80GB, 30 CPUs, 120 GB RAM, 80 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Datacrunch | rental_unspecified | H100 | SXM | 1 | unknown | not stated | on_demand | $4.426 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x H100 SXM5 80GB, 32 CPUs, 170 GB RAM, 80 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Paperspace/DigitalOcean | vm | H200 | unknown | 1 | unknown | not stated | on_demand | $4.47 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| Paperspace/DigitalOcean | vm | H200 | unknown | 8 | unknown | not stated | on_demand | $4.47 DigitalOcean GPU Droplet. Listed as HGX, which is not recorded as SXM. Page says choose 1 or 8 GPUs at this per-GPU price. | 2026-10-21T23:09:38Z | high |
| digitalocean via Shadeform | vm | H200 | SXM | 1 | none | US, New York, NY | on_demand | $4.47 via Shadeform. Underlying cloud: digitalocean. Region: US, New York, NY. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H200 | SXM | 1 | none | US, Atlanta, US | on_demand | $4.47 via Shadeform. Underlying cloud: digitalocean. Region: US, Atlanta, US. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H200 | NVL | 8 | NVLink | US, New York, NY | on_demand | $4.47 via Shadeform. Underlying cloud: digitalocean. Region: US, New York, NY. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| digitalocean via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Atlanta, US | on_demand | $4.47 via Shadeform. Underlying cloud: digitalocean. Region: US, Atlanta, US. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Civo | vm | B200 | unknown | 8 | unknown | not stated | on_demand | $4.49 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Civo | vm | B200 | unknown | 8 | unknown | not stated | 6mo | $4.49 Civo shows $ and does not spell a currency code. Recorded as USD. Hourly node price divided by the stated GPU count. | 2026-10-21T23:10:14Z | high |
| Datacrunch | serverless | B200 | SXM | 1 | unknown | not stated | spot | $4.50 Datacrunch pricing page, now published on Verda. Serverless containers section. The page says auto-scaling containers with pay-per-use. Row: 1x B200 SXM6 180GB, 28 CPUs, 165 GB RAM, 180 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Nebius | vm | H100 | unknown | not stated | unknown | not stated | on_demand | $4.50 On-demand column is labelled GPU-hour. Region is not on the row. HGX is not recorded as SXM. | 2026-10-21T23:09:17Z | high |
| excesssupply via Shadeform | vm | B200 | NVL | 8 | NVLink | US, Kansas City, MO | on_demand | $4.5137 via Shadeform. Underlying cloud: excesssupply. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | FI, Helsinki | on_demand | $4.5163 via Shadeform. Underlying cloud: excesssupply. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | FR, Paris | on_demand | $4.5163 via Shadeform. Underlying cloud: excesssupply. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Kansas City, MO | on_demand | $4.5163 via Shadeform. Underlying cloud: excesssupply. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Kansas City, MO | on_demand | $4.5163 via Shadeform. Underlying cloud: excesssupply. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | FI, Helsinki | on_demand | $4.5163 via Shadeform. Underlying cloud: excesssupply. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | NVL | 8 | NVLink | FR, Paris | on_demand | $4.5163 via Shadeform. Underlying cloud: excesssupply. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H200 | NVL | 8 | NVLink | FR, Paris | on_demand | $4.5163 via Shadeform. Underlying cloud: nebius. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H200 | NVL | 8 | NVLink | FI, Helsinki | on_demand | $4.5163 via Shadeform. Underlying cloud: nebius. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H200 | NVL | 8 | NVLink | US, Kansas City, MO | on_demand | $4.5163 via Shadeform. Underlying cloud: nebius. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Modal | serverless | H200 | SXM | not stated | unknown | not stated | on_demand | $4.5396 Modal GPU task price. Converted from per second to per hour. Node size is not stated. | 2026-10-21T23:09:29Z | high |
| Jarvislabs | vm | H200 | SXM | 1 | unknown | not stated | on_demand | $4.59 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| Jarvislabs | vm | H200 | SXM | 2 | unknown | not stated | on_demand | $4.59 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| Jarvislabs | vm | H200 | SXM | 4 | unknown | not stated | on_demand | $4.59 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| Jarvislabs | vm | H200 | SXM | 8 | unknown | not stated | on_demand | $4.59 Jarvislabs on-demand per GPU-hour. Configurations 1, 2, 4 and 8 are stated at this rate. | 2026-10-21T23:10:20Z | high |
| verda via Shadeform | vm | H200 | SXM | 2 | NVLink | FI, Central | on_demand | $4.615 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 1 | none | FR, Paris | on_demand | $4.63 via Shadeform. Underlying cloud: excesssupply. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 1 | none | US, Kansas City, MO | on_demand | $4.63 via Shadeform. Underlying cloud: excesssupply. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 1 | none | FI, Helsinki | on_demand | $4.63 via Shadeform. Underlying cloud: excesssupply. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 1 | none | FI, Helsinki | on_demand | $4.63 via Shadeform. Underlying cloud: excesssupply. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 1 | none | FR, Paris | on_demand | $4.63 via Shadeform. Underlying cloud: excesssupply. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| excesssupply via Shadeform | vm | H200 | SXM | 1 | none | US, Kansas City, MO | on_demand | $4.63 via Shadeform. Underlying cloud: excesssupply. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H200 | SXM | 1 | none | FR, Paris | on_demand | $4.63 via Shadeform. Underlying cloud: nebius. Region: FR, Paris. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H200 | SXM | 1 | none | FI, Helsinki | on_demand | $4.63 via Shadeform. Underlying cloud: nebius. Region: FI, Helsinki. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| nebius via Shadeform | vm | H200 | SXM | 1 | none | US, Kansas City, MO | on_demand | $4.63 via Shadeform. Underlying cloud: nebius. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| GCP | vm | H200 | unknown | 8 | unknown | not stated | 36mo | $4.6511 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | 8 | unknown | eastus | 36mo | $4.8558 eastus retail price. ND96isH100v5. GPU count 8 is numberOfGPUs for STANDARD_ND96IS_H100_V5 on the Azure Machine Learning managed online endpoints VM SKU list. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | 36mo | $4.858 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Datacrunch | serverless | H100 | SXM | 1 | unknown | not stated | on_demand | $4.87 Datacrunch pricing page, now published on Verda. Serverless containers section. The page says auto-scaling containers with pay-per-use. Row: 1x H100 SXM5 80GB, 21 CPUs, 160 GB RAM, 80 GB VRAM. | 2026-10-21T23:10:05Z | high |
| DeepInfra | managed_inference | B200 | unknown | not stated | unknown | not stated | on_demand | $4.89 DeepInfra dedicated GPU price. Form factor is not stated. | 2026-10-21T23:10:28Z | high |
| Azure | vm | H100 | unknown | 8 | NVLink | eastus | unknown | $4.916 5 Years eastus retail price. ND96isrH100v5. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 5 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 43800 hours. 5 Years is not a fixed bucket in the term list, so term is unknown. | 2026-10-21T23:09:07Z | high |
| GCP | vm | B200 | unknown | 8 | unknown | not stated | spot | $4.9542 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | 36mo | $5.0813 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Hyperstack | vm | B200 | unknown | not stated | unknown | not stated | unknown | $5.10 Starting from Hyperstack reservation price. The page does not state the commitment length, so term is unknown. | 2026-10-21T23:09:40Z | low |
| Azure | vm | H100 | PCIe | 1 | unknown | eastus | 12mo | $5.235 eastus retail price. NC40adsH100v5. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | PCIe | 2 | unknown | eastus | 12mo | $5.235 eastus retail price. NC80adisH100v5. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| RunPod | rental_unspecified | H200 | SXM | not stated | unknown | not stated | on_demand | $5.29 RunPod Secure Cloud gpuTypes.securePrice. | 2026-10-21T23:09:28Z | high |
| Azure | vm | H100 | unknown | 8 | NVLink | eastus | 36mo | $5.3953 eastus retail price. ND96isrH100v5. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| Nebius | vm | H200 | unknown | not stated | unknown | not stated | on_demand | $5.40 On-demand column is labelled GPU-hour. Region is not on the row. HGX is not recorded as SXM. | 2026-10-21T23:09:17Z | high |
| Massed Compute | rental_unspecified | B200 | SXM | 8 | unknown | not stated | on_demand | $5.4325 Massed Compute node price divided by the GPU count on the row. | 2026-10-21T23:10:25Z | high |
| Together | rental_unspecified | H100 | unknown | not stated | unknown | not stated | on_demand | $5.49 Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. | 2026-10-21T23:09:17Z | high |
| Crusoe | managed_inference | H100 | unknown | not stated | unknown | not stated | on_demand | $5.50 Crusoe inference price, already per GPU-hour. HGX is not recorded as SXM. | 2026-10-21T23:09:18Z | high |
| nebius via Shadeform | vm | B200 | NVL | 8 | none | not stated | spot | $5.50 via Shadeform. spot_hourly_price in US cents. Underlying cloud: nebius. | 2026-10-21T23:09:43Z | medium |
| Lambda | rental_unspecified | H100 | unknown | 256 | unknown | not stated | unknown | $5.54 2 weeks to 1 year Lambda tab NVIDIA H100. Duration on the page is 2 weeks to 1 year, so term is unknown. | 2026-10-21T23:09:07Z | low |
| Datacrunch | rental_unspecified | H200 | SXM | 1 | unknown | not stated | on_demand | $5.772 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x H200 SXM5 141GB, 44 CPUs, 170 GB RAM, 141 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Lambda | rental_unspecified | H100 | unknown | 64 | unknown | not stated | unknown | $5.85 2 weeks to 1 year Lambda tab NVIDIA H100. Duration on the page is 2 weeks to 1 year, so term is unknown. | 2026-10-21T23:09:07Z | low |
| Paperspace/DigitalOcean | vm | H100 | unknown | 1 | unknown | not stated | on_demand | $5.95 Paperspace machine price. The docs state an H100x8 machine is 8 times the single-GPU hourly price. Form factor is not stated. | 2026-10-21T23:09:38Z | high |
| RunPod | rental_unspecified | B200 | unknown | not stated | unknown | not stated | on_demand | $5.98 RunPod Community Cloud gpuTypes.communityPrice. Host marketplace price on RunPod. | 2026-10-21T23:09:28Z | medium |
| paperspace via Shadeform | vm | H100 | PCIe | 1 | none | US, New York, NY | on_demand | $5.99 via Shadeform. Underlying cloud: paperspace. Region: US, New York, NY. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | spot | $5.9948 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Crusoe | managed_inference | H200 | unknown | not stated | unknown | not stated | on_demand | $6.00 Crusoe inference price, already per GPU-hour. HGX is not recorded as SXM. | 2026-10-21T23:09:18Z | high |
| Hyperstack | vm | B200 | unknown | not stated | unknown | not stated | on_demand | $6.00 Hyperstack on-demand table, price per GPU per hour. Node size is not a single number on the row. | 2026-10-21T23:09:40Z | high |
| CoreWeave | bare_metal | H100 | unknown | 8 | unknown | NORTH AMERICA | on_demand | $6.155 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| CoreWeave | bare_metal | H100 | unknown | 8 | unknown | EUROPE | on_demand | $6.155 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| Lambda | rental_unspecified | H100 | unknown | 16 | unknown | not stated | unknown | $6.16 2 weeks to 1 year Lambda tab NVIDIA H100. Duration on the page is 2 weeks to 1 year, so term is unknown. | 2026-10-21T23:09:07Z | low |
| OVHcloud | rental_unspecified | H200 | unknown | 8 | NVLink | not stated | on_demand | $6.195 OVHcloud says From. Provider marks the figure as estimated for 8 GPUs, so the node price was divided by 8. | 2026-10-21T23:10:28Z | low |
| Modal | serverless | B200 | unknown | not stated | unknown | not stated | on_demand | $6.2496 Modal GPU task price. Converted from per second to per hour. Node size is not stated. | 2026-10-21T23:09:29Z | high |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | spot | $6.2739 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| CoreWeave | bare_metal | H200 | unknown | 8 | unknown | NORTH AMERICA | on_demand | $6.305 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| CoreWeave | bare_metal | H200 | unknown | 8 | unknown | EUROPE | on_demand | $6.305 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| verda via Shadeform | vm | B200 | SXM | 8 | NVLink | FI, Central | on_demand | $6.305 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Datacrunch | serverless | H200 | SXM | 1 | unknown | not stated | on_demand | $6.35 Datacrunch pricing page, now published on Verda. Serverless containers section. The page says auto-scaling containers with pay-per-use. Row: 1x H200 SXM5 141GB, 21 CPUs, 160 GB RAM, 141 GB VRAM. | 2026-10-21T23:10:05Z | high |
| GCP | vm | H200 | unknown | 8 | unknown | not stated | spot | $6.3542 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| verda via Shadeform | vm | B200 | SXM | 4 | NVLink | FI, Central | on_demand | $6.4175 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Baseten | managed_inference | H100 | unknown | not stated | unknown | not stated | on_demand | $6.50 Baseten pricePerHour. Form factor is not in the record. MIG slices are skipped. | 2026-10-21T23:09:28Z | high |
| verda via Shadeform | vm | B200 | SXM | 1 | NVLink | FI, Central | on_demand | $6.52 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| verda via Shadeform | vm | B200 | SXM | 2 | NVLink | FI, Central | on_demand | $6.52 via Shadeform. Underlying cloud: verda. Region: FI, Central. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Lambda | vm | B200 | SXM | 8 | unknown | not stated | on_demand | $6.69 Lambda tab 8x. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | B200 | SXM | 8 | none | AU, Sydney | on_demand | $6.69 via Shadeform. Underlying cloud: lambdalabs. Region: AU, Sydney. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | B200 | SXM | 8 | none | US, Dulles, VA | on_demand | $6.69 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dulles, VA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Lambda | vm | B200 | SXM | 4 | unknown | not stated | on_demand | $6.79 Lambda tab 4x. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | B200 | SXM | 4 | none | AU, Sydney | on_demand | $6.79 via Shadeform. Underlying cloud: lambdalabs. Region: AU, Sydney. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | B200 | SXM | 4 | none | US, Dulles, VA | on_demand | $6.79 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dulles, VA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| AWS | vm | H100 | unknown | 1 | NVLink | us-east-1 | on_demand | $6.88 Linux on-demand in US East (N. Virginia). GPU count 1 read from the AWS instance type page. Price file does not state a region code; this file is the US East (N. Virginia) Linux map, recorded as us-east-1. | 2026-10-21T23:09:07Z | high |
| AWS | vm | H100 | unknown | 8 | NVLink | us-east-1 | on_demand | $6.88 Linux on-demand in US East (N. Virginia). GPU count 8 read from the AWS instance type page. Price file does not state a region code; this file is the US East (N. Virginia) Linux map, recorded as us-east-1. | 2026-10-21T23:09:07Z | high |
| Lambda | vm | B200 | SXM | 2 | unknown | not stated | on_demand | $6.89 Lambda tab 2x. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | B200 | SXM | 2 | none | US, Dulles, VA | on_demand | $6.89 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dulles, VA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | B200 | SXM | 2 | none | AU, Sydney | on_demand | $6.89 via Shadeform. Underlying cloud: lambdalabs. Region: AU, Sydney. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Azure | vm | H100 | PCIe | 1 | unknown | eastus | on_demand | $6.98 eastus retail price. NC40adsH100v5. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | PCIe | 2 | unknown | eastus | on_demand | $6.98 eastus retail price. NC80adisH100v5. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| Lambda | vm | B200 | SXM | 1 | unknown | not stated | on_demand | $6.99 Lambda tab 1x. | 2026-10-21T23:09:07Z | high |
| lambdalabs via Shadeform | vm | B200 | SXM | 1 | none | US, Dulles, VA | on_demand | $6.99 via Shadeform. Underlying cloud: lambdalabs. Region: US, Dulles, VA. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| lambdalabs via Shadeform | vm | B200 | SXM | 1 | none | AU, Sydney | on_demand | $6.99 via Shadeform. Underlying cloud: lambdalabs. Region: AU, Sydney. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Azure | vm | H100 | unknown | 8 | unknown | eastus | 12mo | $7.079 eastus retail price. ND96isH100v5. GPU count 8 is numberOfGPUs for STANDARD_ND96IS_H100_V5 on the Azure Machine Learning managed online endpoints VM SKU list. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| GCP | vm | B200 | unknown | 8 | unknown | not stated | 36mo | $7.0884 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| GCP | vm | H200 | unknown | 8 | unknown | not stated | 12mo | $7.309 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | 12mo | $7.673 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | 8 | NVLink | eastus | 12mo | $7.8656 eastus retail price. ND96isrH100v5. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| AWS | vm | H200 | unknown | 8 | NVLink | us-east-1 | on_demand | $7.912 Linux on-demand in US East (N. Virginia). GPU count 8 read from the AWS instance type page. Price file does not state a region code; this file is the US East (N. Virginia) Linux map, recorded as us-east-1. | 2026-10-21T23:09:07Z | high |
| RunPod | rental_unspecified | B200 | unknown | not stated | unknown | not stated | on_demand | $7.99 RunPod Secure Cloud gpuTypes.securePrice. | 2026-10-21T23:09:28Z | high |
| Fireworks | managed_inference | H100 | unknown | not stated | unknown | not stated | on_demand | $8.00 Fireworks on-demand deployments table, per-hour column. Form factor is not stated. | 2026-10-21T23:09:18Z | high |
| Fireworks | managed_inference | H200 | unknown | not stated | unknown | not stated | on_demand | $8.00 Fireworks on-demand deployments table, per-hour column. Form factor is not stated. | 2026-10-21T23:09:18Z | high |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | 12mo | $8.0258 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Datacrunch | rental_unspecified | B200 | SXM | 1 | unknown | not stated | on_demand | $8.19 Datacrunch pricing page, now published on Verda. GPU instances section. The page says 1x, 2x, 4x, or 8x NVIDIA GPU instances. It does not say virtual machine or bare metal. Row: 1x B200 SXM6 180GB, 30 CPUs, 170 GB RAM, 180 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Datacrunch | rental_unspecified | B200 | SXM | 1 | unknown | not stated | on_demand | $8.355 Datacrunch pricing page, now published on Verda. Confidential computing section. The page says data and the model stay encrypted in memory, in transit, and at rest. It does not say virtual machine or bare metal. Row: 1x B200 CC SXM6 180GB, 30 CPUs, 170 GB RAM, 180 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Nebius | vm | B200 | unknown | not stated | unknown | not stated | on_demand | $8.50 On-demand column is labelled GPU-hour. Region is not on the row. HGX is not recorded as SXM. | 2026-10-21T23:09:17Z | high |
| CoreWeave | bare_metal | B200 | unknown | 8 | unknown | NORTH AMERICA | on_demand | $8.60 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| CoreWeave | bare_metal | B200 | unknown | 8 | unknown | EUROPE | on_demand | $8.60 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| Vast.ai | rental_unspecified | B200 | unknown | not stated | unknown | not stated | on_demand | $8.61 Read from vast.ai/pricing. The public JSON feed was not used because its license restricts index redistribution. | 2026-10-21T23:09:31Z | medium |
| Lambda | rental_unspecified | B200 | unknown | not stated | unknown | not stated | unknown | $8.87 2 weeks to 1 year Lambda tab NVIDIA HGX B200. Duration on the page is 2 weeks to 1 year, so term is unknown. | 2026-10-21T23:09:07Z | low |
| Together | rental_unspecified | B200 | unknown | not stated | unknown | not stated | on_demand | $8.99 Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. | 2026-10-21T23:09:17Z | high |
| Datacrunch | serverless | B200 | SXM | 1 | unknown | not stated | on_demand | $9.01 Datacrunch pricing page, now published on Verda. Serverless containers section. The page says auto-scaling containers with pay-per-use. Row: 1x B200 SXM6 180GB, 28 CPUs, 165 GB RAM, 180 GB VRAM. | 2026-10-21T23:10:05Z | high |
| Lambda | rental_unspecified | B200 | unknown | 64 | unknown | not stated | unknown | $9.36 2 weeks to 1 year Lambda tab NVIDIA HGX B200. Duration on the page is 2 weeks to 1 year, so term is unknown. | 2026-10-21T23:09:07Z | low |
| nebius via Shadeform | vm | B200 | NVL | 8 | none | US, Kansas City, MO | on_demand | $9.50 via Shadeform. Underlying cloud: nebius. Region: US, Kansas City, MO. hourly_price is US cents per instance-hour. | 2026-10-21T23:09:43Z | medium |
| Crusoe | managed_inference | B200 | unknown | not stated | unknown | not stated | on_demand | $9.65 Crusoe inference price, already per GPU-hour. HGX is not recorded as SXM. | 2026-10-21T23:09:18Z | high |
| Lambda | rental_unspecified | B200 | unknown | 16 | unknown | not stated | unknown | $9.86 2 weeks to 1 year Lambda tab NVIDIA HGX B200. Duration on the page is 2 weeks to 1 year, so term is unknown. | 2026-10-21T23:09:07Z | low |
| Baseten | managed_inference | B200 | unknown | not stated | unknown | not stated | on_demand | $9.98 Baseten pricePerHour. Form factor is not in the record. MIG slices are skipped. | 2026-10-21T23:09:28Z | high |
| Oracle | vm | H100 | unknown | not stated | unknown | not stated | on_demand | $10.00 Oracle cetools price list, PAY_AS_YOU_GO, metric GPU Per Hour. Region is not in the API record. | 2026-10-21T23:09:07Z | high |
| Oracle | vm | H200 | unknown | not stated | unknown | not stated | on_demand | $10.00 Oracle cetools price list, PAY_AS_YOU_GO, metric GPU Per Hour. Region is not in the API record. | 2026-10-21T23:09:07Z | high |
| CoreWeave | bare_metal | GB200 | NVL | 4 | NVLink | NORTH AMERICA | on_demand | $10.50 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| CoreWeave | bare_metal | GB200 | NVL | 4 | NVLink | EUROPE | on_demand | $10.50 Divided by the GPU count in the CoreWeave pricing row. GH200 is not H200 and is skipped. | 2026-10-21T23:09:07Z | high |
| GCP | vm | H200 | unknown | 8 | unknown | not stated | on_demand | $10.6009 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Oracle | vm | H100 | unknown | not stated | unknown | not stated | on_demand | $10.75 Oracle cetools price list, PAY_AS_YOU_GO, metric GPU Per Hour. Region is not in the API record. SKU name is H100T. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | 8 | unknown | eastus | on_demand | $11.061 eastus retail price. ND96isH100v5. GPU count 8 is numberOfGPUs for STANDARD_ND96IS_H100_V5 on the Azure Machine Learning managed online endpoints VM SKU list. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | on_demand | $11.0613 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| GCP | vm | B200 | unknown | 8 | unknown | not stated | 12mo | $11.1159 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| GCP | vm | H100 | unknown | 8 | unknown | not stated | on_demand | $11.6751 Google Cloud accelerator-optimized price sheet. 1 Year and 3 Year columns are resource CUDs, stored as 12mo and 36mo. Region is not stated on the row. Form factor is not stated. | 2026-10-21T23:09:07Z | high |
| Azure | vm | GB200 | NVL | 4 | NVLink | eastus | 36mo | $11.8976 eastus retail price. ND128isrNDRGB200v6. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | 8 | NVLink | eastus | on_demand | $12.29 eastus retail price. ND96isrH100v5. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| Fireworks | managed_inference | B200 | unknown | not stated | unknown | not stated | on_demand | $13.00 Fireworks on-demand deployments table, per-hour column. Form factor is not stated. | 2026-10-21T23:09:18Z | high |
| Oracle | vm | B200 | unknown | not stated | unknown | not stated | on_demand | $14.00 Oracle cetools price list, PAY_AS_YOU_GO, metric GPU Per Hour. Region is not in the API record. | 2026-10-21T23:09:07Z | high |
| AWS | vm | B200 | unknown | 8 | unknown | us-east-1 | on_demand | $14.2416 Linux on-demand in US East (N. Virginia). GPU count 8 read from the AWS instance type page. Price file does not state a region code; this file is the US East (N. Virginia) Linux map, recorded as us-east-1. | 2026-10-21T23:09:07Z | high |
| Oracle | vm | GB200 | unknown | not stated | unknown | not stated | on_demand | $16.00 Oracle cetools price list, PAY_AS_YOU_GO, metric GPU Per Hour. Region is not in the API record. | 2026-10-21T23:09:07Z | high |
| Azure | vm | GB200 | NVL | 4 | NVLink | eastus | 12mo | $17.3056 eastus retail price. ND128isrNDRGB200v6. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | GB200 | NVL | 4 | NVLink | eastus | spot | $27.04 eastus retail price. ND128isrNDRGB200v6 Spot. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| Azure | vm | GB200 | NVL | 4 | NVLink | eastus | on_demand | $27.04 eastus retail price. ND128isrNDRGB200v6. Page also states InfiniBand for scale-out. Windows, Low Priority and DevTestConsumption meters skipped. | 2026-10-21T23:09:07Z | high |
| AWS | vm | GB200 | unknown | 4 | unknown | us-east-1 | on_demand | not listed Instance is on the AWS spec page but has no on-demand price in the us-east-1 Linux price file. | 2026-10-21T23:09:07Z | high |
| AWS | vm | H200 | unknown | 8 | NVLink | us-east-1 | on_demand | not listed Instance is on the AWS spec page but has no on-demand price in the us-east-1 Linux price file. | 2026-10-21T23:09:07Z | high |
| Azure | managed_inference | H100 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter NCadisH100v5 Capacity Unit. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | managed_inference | H100 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter NCadsH100 v5 Capacity Unit. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | spot | not listed Retail meter ND96isflH100v5 Spot. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | unknown | H100 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter Mngd H100_80GB Gl. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter ND96isnIBH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | 12mo | not listed Retail meter ND96isnIBH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | 36mo | not listed Retail meter ND96isnIBH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | unknown | not listed 5 Years Retail meter ND96isnIBH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 5 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 43800 hours. 5 Years is not a fixed bucket in the term list, so term is unknown. | 2026-10-21T23:09:07Z | high |
| Azure | managed_inference | H100 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter NDisrH100v5 Capacity Unit. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter ND96isrfH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | 12mo | not listed Retail meter ND96isrfH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | 36mo | not listed Retail meter ND96isrfH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | unknown | not listed 5 Years Retail meter ND96isrfH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 5 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 43800 hours. 5 Years is not a fixed bucket in the term list, so term is unknown. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | spot | not listed Retail meter ND96isnIBH100v5 Spot. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter ND96isflH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | 12mo | not listed Retail meter ND96isflH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 1 Year: the API unit is 1 Hour but the retail price is the upfront total, divided by 8760 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | 36mo | not listed Retail meter ND96isflH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 3 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 26280 hours. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | unknown | not listed 5 Years Retail meter ND96isflH100v5. GPU count was not on the fetched Learn size pages, so the instance price was not divided. Azure reservationTerm 5 Years: the API unit is 1 Hour but the retail price is the upfront total, divided by 43800 hours. 5 Years is not a fixed bucket in the term list, so term is unknown. | 2026-10-21T23:09:07Z | high |
| Azure | vm | H100 | unknown | not stated | unknown | eastus | spot | not listed Retail meter ND96isrfH100v5 Spot. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Azure | unknown | H200 | unknown | not stated | unknown | eastus | on_demand | not listed Retail meter Mngd H200_141GB Gl. GPU count was not on the fetched Learn size pages, so the instance price was not divided. | 2026-10-21T23:09:07Z | high |
| Crusoe | vm | B200 | unknown | not stated | unknown | not stated | on_demand | not listed Crusoe instance row says Contact sales. HGX is not recorded as SXM. | 2026-10-21T23:09:18Z | high |
| Crusoe | vm | GB200 | NVL | not stated | unknown | not stated | on_demand | not listed Crusoe instance row says Contact sales. HGX is not recorded as SXM. | 2026-10-21T23:09:18Z | high |
| GCP | vm | B200 | unknown | 8 | unknown | not stated | on_demand | not listed Accelerator-optimized price sheet lists N/A. Region is not on the row. | 2026-10-21T23:09:07Z | high |
| Lambda | rental_unspecified | B200 | unknown | not stated | unknown | not stated | unknown | not listed 1 year+ Lambda tab NVIDIA HGX B200. 1 year+ has no fixed term and no numeric price. Price not listed. | 2026-10-21T23:09:07Z | low |
| Lambda | rental_unspecified | H100 | unknown | not stated | unknown | not stated | unknown | not listed 1 year+ Lambda tab NVIDIA H100. 1 year+ has no fixed term and no numeric price. Price not listed. | 2026-10-21T23:09:07Z | low |
| Nebius | vm | GB200 | NVL | not stated | unknown | not stated | on_demand | not listed Nebius on-demand column says Contact us. vCPU and RAM columns are not a GPU count. | 2026-10-21T23:09:17Z | high |
| Together | rental_unspecified | B200 | unknown | not stated | unknown | not stated | unknown | not listed Reserved Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. The reserved column does not state a duration, so term is unknown. | 2026-10-21T23:09:17Z | high |
| Together | rental_unspecified | GB200 | NVL | not stated | unknown | not stated | on_demand | not listed Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. On-demand cell is contact. | 2026-10-21T23:09:17Z | high |
| Together | rental_unspecified | GB200 | NVL | not stated | unknown | not stated | unknown | not listed Reserved Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. The reserved column does not state a duration, so term is unknown. | 2026-10-21T23:09:17Z | high |
| Together | rental_unspecified | H100 | unknown | not stated | unknown | not stated | unknown | not listed Reserved Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. The reserved column does not state a duration, so term is unknown. | 2026-10-21T23:09:17Z | high |
| Together | rental_unspecified | H200 | unknown | not stated | unknown | not stated | on_demand | not listed Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. On-demand cell is contact. | 2026-10-21T23:09:17Z | high |
| Together | rental_unspecified | H200 | unknown | not stated | unknown | not stated | unknown | not listed Reserved Together hardware table says prices are per GPU per hour. HGX is not recorded as SXM. The reserved column does not state a duration, so term is unknown. | 2026-10-21T23:09:17Z | high |
Stale tier-1 providers
- Vultr: stale. Last row: none.
- SF Compute: stale. Last row: none.
- Fluidstack: stale. Last row: none.
- Hot Aisle: stale. Last row: none.
- Ori: stale. Last row: none.
Failures this run
- Vultr: http_error Headless Chrome loaded https://www.vultr.com/pricing/, which robots.txt allows, and Cloudflare returned a challenge page (Just a moment) instead of the pricing table. api.vultr.com/robots.txt disallows the whole host, so the plans API was not called. The challenge was not bypassed. No price was stored.. Last successful price: none.
- SF Compute: login_gated sfcompute.com/pricing is a sign-in page. An unauthenticated request to the orderbook quote returned HTTP 401 (missing authentication token). The docs require a dashboard bearer token. No price was stored. Homepage calculator examples were not stored.. Last successful price: none.
- Fluidstack: login_gated fluidstack.io does not list an H100, H200, B200 or GB200 price. fluidstack.io/pricing returned HTTP 404. console.fluidstack.io is a sign-in page. No price was stored.. Last successful price: none.
- Hot Aisle: zero_prices hotaisle.xyz/pricing does not list an H100, H200, B200 or GB200 price. The page lists AMD MI300X or MI355X prices, which are outside this index.. Last successful price: none.
- Ori: zero_prices https://radiant.co/ori is Ori's own page after the merger with Radiant. It does not list an H100, H200, B200 or GB200 rental price. https://radiant.co/gpus lists GPU hardware and no H100, H200, B200 or GB200 rental price.. Last successful price: none.
Held for review
None.
Page views: …