AI pricing pagesGPU cloud pricing
Tracking 742 prices · Checked 2 hours ago
GPU prices by the hour
Every GPU rental price the Index tracks, from on-demand prices to reserved and spot rates, converted to one GPU for one hour and shown beside the provider's own published price.
What each GPU costs by the hour
Prices are for one GPU for one hour, but GPUs are sold in whole units: an H100 server holds 8, a GB300 NVL72 rack holds 72. Each card shows both. The gray bar runs from the cheapest to the priciest on-demand price, on the same scale for every card.
NVIDIA H100: every price, cheapest first
H100 price page →Verda1x H100 SXM5 80GB $1.96+20.2%
Voltage ParkH100 (single GPU, no contract) $1.99
GMI CloudNVIDIA H100 $2.00
Microsoft AzureStandard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_) $2.04
NebiusNVIDIA HGX H100 $2.15
VerdaServerless containers, 1x H100 SXM5 80GB $2.16+20.7%
Microsoft AzureStandard_ND96isr_H100_v5 (ND H100 v5) $2.27
Microsoft AzureStandard_ND96isrf_H100_v5 $2.27
VultrNVIDIA HGX H100, Cloud GPU $2.30
VultrNVIDIA HGX H100, Bare Metal (8 × H100 80 GB SXM) $2.30
Denvr DataworksNVIDIA H100 SXM (8 GPUs, 208 vCPU) $2.45
CoreWeaveNVIDIA HGX H100 $2.46
Cirrascale8X NVIDIA H100 (16-Node+ Clustered) $2.49
Massed ComputeH100 (80 GB) ×1/×2/×4 $2.73
RunpodH100 PCIe / H100 NVL, Pods $2.89
Massed ComputeH100 NVL (94GB) NVLink ×2/×4/×8 $2.92
Impossible CloudH100 SXM 8-GPU nodes (EU, offer 286-H100) $2.95
Massed ComputeH100 NVL ×1/×2/×4/×8 $3.11
Massed ComputeH100 SXM5 (80 GB) ×8 $3.14
Massed ComputeH100 SXM5 (80GB) ×1 $3.14+8.7%
HyperbolicH100 SXM $3.19
Together AIreserved $3.19
DigitalOcean12-month reserved $3.26
LambdaNVIDIA H100 PCIe $3.29
Cirrascale8X NVIDIA H100 (Standalone) $3.42
Astra IntellectusH100 SXM 8-GPU $3.65
NebiusNVIDIA HGX H100 $3.85
Crusoe CloudNVIDIA H100 $3.90
Verda1x H100 SXM5 80GB $3.93+20.9%
LambdaNVIDIA H100 SXM (on-demand instance, 8-GPU config, 208 vCPU/1800 GiB) $3.99
RunpodH100 SXM (80 GB, 20 vCPU), Pods $3.99+14.3%
Together AIH100 (80 GB), 8–256 GPUs $3.99
LambdaNVIDIA H100 SXM (smaller configs) $4.09
CoreWeaveNVIDIA H100 PCIe $4.25
Cirrascale8X H100 (standalone & 16-node+ clustered) $4.28
VerdaServerless containers, 1x H100 SXM5 80GB $4.32+20.7%
DigitalOceanNVIDIA HGX H100 (80 GB) $4.41
CoreWeaveNVIDIA HGX H100 (classic) $4.76
Microsoft AzureStandard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_) $4.86
Google Cloud3-year committed use $4.86
Google Clouda3-highgpu-8g (A3 High) $5.20
Microsoft AzureStandard_ND96isr_H100_v5 (ND H100 v5) $5.40
Microsoft AzureStandard_ND96isrf_H100_v5 $5.40
Google Clouda3-megagpu-8g (A3 Mega) $5.50
Lambda1-Click Clusters, NVIDIA H100 $5.54
AWSp5.48xlarge $5.97+15%
AWSp5.4xlarge $5.97+15%
CoreWeaveNVIDIA HGX H100 $6.16
Google Clouda3-highgpu-8g (A3 High) $6.29−0.3%
BasetenH100 (80 GiB) dedicated deployment / training $6.50
Google Clouda3-megagpu-8g (A3 Mega) $6.64−0.3%
Microsoft AzureStandard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_) $7.08
Google Clouda3-highgpu-8g (A3 High) $7.67
Microsoft AzureStandard_ND96isr_H100_v5 (ND H100 v5) $7.87
Microsoft AzureStandard_ND96isrf_H100_v5 $7.87
Fireworks AIH100 80 GB $8.00+14.3%
Google Clouda3-megagpu-8g (A3 Mega) $8.03
Oracle OCIBM.GPU.H100.8 $10.00
Microsoft AzureStandard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_) $11.06
Google Clouda3-highgpu-8g (A3 High) $11.06
Google Clouda3-megagpu-8g (A3 Mega) $11.68
Microsoft AzureStandard_ND96isr_H100_v5 (ND H100 v5) $12.29
Microsoft AzureStandard_ND96isrf_H100_v5 $12.29
See all 66 H100 offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| Voltage Park | H100 (single GPU, no contract) | on-demand | You can rent one H100 GPU for 1 hour starting at $1.99 without a contract. | $1.99 | No change | On pageOct 10, 2026 | voltagepark.com ↗ |
| GMI Cloud | NVIDIA H100 | on-demand | from $2.00/GPU-hour (AVAILABLE NOW) | $2.00 | No change | RecheckingOct 10, 2026 | gmicloud.ai ↗ |
| Vultr | NVIDIA HGX H100, Cloud GPU | on-demand | $2.300/GPU/hr | $2.30 | No change | BlockedOct 10, 2026 | vultr.com ↗ |
| Denvr Dataworks | NVIDIA H100 SXM (8 GPUs, 208 vCPU) | on-demand | $2.45 / GPU per hour, NVIDIA H100 SXM, 8 GPUs | $2.45 | No change | On pageOct 10, 2026 | denvr.com ↗ |
| Cirrascale | 8X NVIDIA H100 (16-Node+ Clustered) | "from" floor | As low as $2.49 per GPU hour, 8X NVIDIA H100 (16-Node+ Clustered) | $2.49 | No change | On pageOct 10, 2026 | cirrascale.com ↗ |
| Massed Compute | H100 (80 GB) ×1/×2/×4 | on-demand | ×1 $2.73/hr, ×2 $5.46/hr, ×4 $10.92/hr, H100 (80GB), on-demand per instance-hour | $2.73 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Runpod | H100 PCIe / H100 NVL, Pods | on-demand | $2.89/hr, Pods, Secure Cloud | $2.89 | No change | On pageOct 10, 2026 | runpod.io ↗ |
| Massed Compute | H100 NVL (94GB) NVLink ×2/×4/×8 | on-demand | ×2 $5.84/hr, ×4 $11.68/hr, ×8 $23.36/hr, H100 NVL (94GB) NVLink, on-demand per instance-hour | $2.92 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Massed Compute | H100 NVL ×1/×2/×4/×8 | on-demand | ×1 $3.11/hr, ×2 $6.22/hr, ×4 $12.44/hr, ×8 $24.88/hr, H100 NVL, on-demand per instance-hour | $3.11 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Massed Compute | H100 SXM5 (80 GB) ×8 | on-demand | ×8 $25.12/hr, H100 SXM5 (80GB), on-demand per instance-hour | $3.14 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Massed Compute | H100 SXM5 (80GB) ×1 | on-demand | ×1 $3.14/hr, H100 SXM5 (80GB), on-demand per instance-hour | $3.14 | +8.7% | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Hyperbolic | H100 SXM | on-demand | $3.19 / HR, Nvidia H100 SXM, starting from (per GPU hour), refreshed weekly | $3.19 | No change | On pageOct 10, 2026 | hyperbolic.ai ↗ |
| Lambda | NVIDIA H100 PCIe | on-demand | $3.29 per GPU-hour, 1x instance (26 vCPUs), on-demand | $3.29 | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Nebius | NVIDIA HGX H100 | on-demand | $3.85 per GPU-hour, on-demand | $3.85 | No change | On pageOct 10, 2026 | nebius.com ↗ |
| Crusoe Cloud | NVIDIA H100 | on-demand | $3.90/GPU-hr, on-demand | $3.90 | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| Verda | 1x H100 SXM5 80GB | on-demand | $3.93/h on-demand, 1x H100 SXM5 80GB, GPU instances, USD | $3.93 | +20.9% | On pageOct 10, 2026 | verda.com ↗ |
| Lambda | NVIDIA H100 SXM (on-demand instance, 8-GPU config, 208 vCPU/1800 GiB) | on-demand | $3.99 per GPU-hour, 8x instance (208 vCPUs), on-demand | $3.99 | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Runpod | H100 SXM (80 GB, 20 vCPU), Pods | on-demand | $3.99/hr, Pods, Secure Cloud | $3.99 | +14.3% | On pageOct 10, 2026 | runpod.io ↗ |
| Together AI | H100 (80 GB), 8–256 GPUs | on-demand | $3.99/hr per GPU, on-demand | $3.99 | No change | On pageOct 10, 2026 | together.ai ↗ |
| Lambda | NVIDIA H100 SXM (smaller configs) | on-demand | $4.09 per GPU-hour, 4x instance (104 vCPUs), on-demand | $4.09 | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| CoreWeave | NVIDIA H100 PCIe | on-demand | $4.25 per GPU-hour, GPU component cost per hour (classic pricing) | $4.25 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Cirrascale | 8X H100 (standalone & 16-node+ clustered) | $24,999 per server per month, Monthly Term, 8X NVIDIA H100 (Standalone) ($4.28/GPUhr equivalent shown) | $4.28 | No change | On pageOct 10, 2026 | cirrascale.com ↗ | |
| Verda | Serverless containers, 1x H100 SXM5 80GB | serverless | $4.32/h serverless, 1x H100 SXM5 80GB, Serverless containers, USD | $4.32 | +20.7% | On pageOct 10, 2026 | verda.com ↗ |
| DigitalOcean | NVIDIA HGX H100 (80 GB) | on-demand | $4.41/GPU/hour, On-Demand Price | $4.41 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| CoreWeave | NVIDIA HGX H100 (classic) | on-demand | $4.76 per GPU-hour, GPU component cost per hour (classic pricing) | $4.76 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Google Cloud | a3-highgpu-8g (A3 High) | calendar mode | $41.6000 / 1 hour for a3-highgpu-8g (Nvidia H100, GPUs: 8), DWS Calendar Mode price (USD) | $5.20 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Google Cloud | a3-megagpu-8g (A3 Mega) | calendar mode | $44.0000 / 1 hour for a3-megagpu-8g (Nvidia H100, GPUs: 8), DWS Calendar Mode price (USD) | $5.50 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| AWS | p5.48xlarge | capacity block | $47.757 USD ($5.970 USD) per instance-hour (per accelerator), 8 x H100, US East (Atlanta) Local Zone, US East (N. Virginia), US East (Ohio), US West (Oregon), US West (N. California), Asia Pacific (Tokyo), Asia Pacific (Jakarta), Asia Pacific (Mumbai), Australia (Sydney), Europe (London), Europe (Stockholm), South America (Sao Paulo) | $5.97 | +15% | On pageOct 10, 2026 | aws.amazon.com ↗ |
| AWS | p5.4xlarge | capacity block | $5.970 USD ($5.970 USD) per instance-hour (per accelerator), 1 x H100, US East (N. Virginia), US East (Ohio), US West (Oregon), Asia Pacific (Tokyo), Asia Pacific (Mumbai), Australia (Sydney), Europe (London), South America (Sao Paulo) | $5.97 | +15% | On pageOct 10, 2026 | aws.amazon.com ↗ |
| CoreWeave | NVIDIA HGX H100North America | on-demand | $49.24 per hour for 8 GPUs, on-demand, North America | $6.16 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Baseten | H100 (80 GiB) dedicated deployment / training | on-demand | $0.10833 per minute, H100 80 GiB VRAM, Dedicated Deployments | $6.50 | No change | On pageOct 10, 2026 | baseten.co ↗ |
| Fireworks AI | H100 80 GB | on-demand | $8.00 per GPU-hour, on-demand deployments, from Sep 1 | $8.00 | +14.3% | Not foundOct 10, 2026 | fireworks.ai ↗ |
| Oracle OCI | BM.GPU.H100.8 | on-demand | $10.00 | $10.00 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Microsoft Azure | Standard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_)East US 2 | on-demand | 88.488 USD per 1 Hour for 8 GPUs, Linux, East US 2 | $11.06 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Google Cloud | a3-highgpu-8g (A3 High) | on-demand | $88.490000119 / 1 hour for a3-highgpu-8g (Nvidia H100, GPUs: 8), Price (USD) | $11.06 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Google Cloud | a3-megagpu-8g (A3 Mega) | on-demand | $93.400712807 / 1 hour for a3-megagpu-8g (Nvidia H100, GPUs: 8), Price (USD) | $11.68 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Microsoft Azure | Standard_ND96isr_H100_v5 (ND H100 v5)East US | on-demand | 98.32 USD per 1 Hour for 8 GPUs, Linux, East US | $12.29 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND96isrf_H100_v5East US | on-demand | 98.32 USD per 1 Hour for 8 GPUs, Linux, East US | $12.29 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Committed and reserved | |||||||
| Vultr | NVIDIA HGX H100, Bare Metal (8 × H100 80 GB SXM) | pre-paid contract | $2.300/GPU/hour | $2.30 | No change | BlockedOct 10, 2026 | vultr.com ↗ |
| Impossible Cloud | H100 SXM 8-GPU nodes (EU, offer 286-H100)EU | 12-month term | $2.95 per GPU-hour, 12-month term, 12 nodes x 8 GPUs (96 GPUs), EU | $2.95 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Together AI | reserved | reserved | Starting at $3.19/hr per GPU, reserved | $3.19 | No change | On pageOct 10, 2026 | together.ai ↗ |
| DigitalOcean | 12-month reserved | 12-month reserved | $3.26/GPU/hour, 12 Month Reserved Price | $3.26 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Cirrascale | 8X NVIDIA H100 (Standalone) | annual | $19,999 per server per month, Annual Term, 8X NVIDIA H100 (Standalone) ($3.43/GPUhr equivalent shown) | $3.42 | No change | On pageOct 10, 2026 | cirrascale.com ↗ |
| Astra Intellectus | H100 SXM 8-GPUEU #2 | term lease | $3.65/GPU/Hour, H100 SXM 8-GPU, EU #2, 128 nodes, 3-year lease | $3.65 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| Microsoft Azure | Standard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_)East US 2 | 3-year reserved | 1020879 USD for one VM with 8 GPUs, reservation term 3 Years, East US 2 | $4.86 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Google Cloud | 3-year committed use | 3-year committed use | $38.864383195 / 1 hour for a3-highgpu-8g (Nvidia H100, GPUs: 8), Compute Resource CUDs - 3 Year (USD) | $4.86 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Microsoft Azure | Standard_ND96isr_H100_v5 (ND H100 v5)East US | 3-year reserved | 1134310 USD for one VM with 8 GPUs, reservation term 3 Years, East US | $5.40 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND96isrf_H100_v5East US | 3-year reserved | 1134310 USD for one VM with 8 GPUs, reservation term 3 Years, East US | $5.40 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Lambda | 1-Click Clusters, NVIDIA H100 | short-term reserved cluster | $5.54 per GPU-hour for 256 GPUs, 2 weeks – 1 year, 1-Click Clusters | $5.54 | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Microsoft Azure | Standard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_)East US 2 | 1-year reserved | 496099 USD for one VM with 8 GPUs, reservation term 1 Year, East US 2 | $7.08 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Google Cloud | a3-highgpu-8g (A3 High) | 1-year committed use | $61.383674231 / 1 hour for a3-highgpu-8g (Nvidia H100, GPUs: 8), Compute Resource CUDs - 1 Year (USD) | $7.67 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Microsoft Azure | Standard_ND96isr_H100_v5 (ND H100 v5)East US | 1-year reserved | 551221 USD for one VM with 8 GPUs, reservation term 1 Year, East US | $7.87 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND96isrf_H100_v5East US | 1-year reserved | 551221 USD for one VM with 8 GPUs, reservation term 1 Year, East US | $7.87 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Google Cloud | a3-megagpu-8g (A3 Mega) | 1-year committed use | $64.206558727 / 1 hour for a3-megagpu-8g (Nvidia H100, GPUs: 8), Compute Resource CUDs - 1 Year (USD) | $8.03 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Spot and preemptible | |||||||
| Verda | 1x H100 SXM5 80GB | spot | $1.96/h spot, 1x H100 SXM5 80GB, GPU instances, USD | $1.96 | +20.2% | On pageOct 10, 2026 | verda.com ↗ |
| Microsoft Azure | Standard_ND96is_H100_v5 (no InfiniBand variants incl. _noIB_, _flex_)East US 2 | spot | 16.352582 USD per 1 Hour for 8 GPUs, Linux Spot, East US 2 | $2.04 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Nebius | NVIDIA HGX H100 | spot | $2.15 per GPU-hour, preemptible | $2.15 | No change | Not foundOct 10, 2026 | nebius.com ↗ |
| Verda | Serverless containers, 1x H100 SXM5 80GB | spot | $2.16/h spot, 1x H100 SXM5 80GB, Serverless containers, USD | $2.16 | +20.7% | On pageOct 10, 2026 | verda.com ↗ |
| Microsoft Azure | Standard_ND96isr_H100_v5 (ND H100 v5)East US | spot | 18.169536 USD per 1 Hour for 8 GPUs, Linux Spot, East US | $2.27 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND96isrf_H100_v5East US | spot | 18.169536 USD per 1 Hour for 8 GPUs, Linux Spot, East US | $2.27 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| CoreWeave | NVIDIA HGX H100North America | spot | $19.71 per hour for 8 GPUs, spot, North America | $2.46 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Google Cloud | a3-highgpu-8g (A3 High) | spot | $50.315338521 / 1 hour for a3-highgpu-8g (Nvidia H100, GPUs: 8), Current Spot pricing (USD) | $6.29 | −0.3% | On pageOct 10, 2026 | cloud.google.com ↗ |
| Google Cloud | a3-megagpu-8g (A3 Mega) | spot | $53.099242521 / 1 hour for a3-megagpu-8g (Nvidia H100, GPUs: 8), Current Spot pricing (USD) | $6.64 | −0.3% | On pageOct 10, 2026 | cloud.google.com ↗ |
| No public price | |||||||
| Fluidstack | H100 | not published | , | Not published | No change | ResearchSep 10, 2026 | fluidstack.io ↗ |
| Lightning AI | H100 | not published | Cost/GPU/hr* | Not published | No change | ResearchSep 10, 2026 | lightning.ai ↗ |
| Runpod | Clusters, H100 SXM | not published | , | Not published | No change | On pageOct 10, 2026 | runpod.io ↗ |
NVIDIA H200: every price, cheapest first
H200 price page →NebiusNVIDIA HGX H200 $2.45
Verda1x H200 SXM5 141GB $2.56+28%
GMI CloudNVIDIA H200 $2.60
CoreWeaveNVIDIA HGX H200 $2.62
VerdaServerless containers, 1x H200 SXM5 141GB $2.82+28.2%
Impossible CloudH200 SXM 8-GPU nodes (EU, offer 287-H200) $3.11
Impossible CloudH200 SXM 8-GPU nodes (APAC, offer 294-H200) $3.16
TheAI Cloud1-year reserved $3.20
DigitalOcean12-month reserved $3.40
TheAI Cloud6-month reserved $3.45
Massed ComputeH200 NVL (141 GB) ×8 $3.62
Cirrascale8X NVIDIA H200 $3.63
Astra IntellectusH200 SXM 8-GPU $3.65
Astra IntellectusH200 SXM 8-GPU (TH #1, 128 nodes) $3.65
TheAI Cloud3-month reserved $3.70
Massed ComputeH200 NVL NVLink ×8 $3.82
HyperbolicH200 $3.99
Together AIreserved $3.99
Crusoe CloudNVIDIA H200 $4.29
RunpodClusters, H200 SXM $4.31
DigitalOceanNVIDIA HGX H200 (141 GB) $4.47
NebiusNVIDIA HGX H200 $4.50
Cirrascale8X H200 $4.54
Google Cloud3-year committed use $4.65
Verda1x H200 SXM5 141GB $5.12+28%
VerdaInstant clusters, H200 SXM5 141GB $5.12+28%
Microsoft AzureStandard_ND96isr_H200_v5 (ND H200 v5) $5.28
RunpodH200 (141 GB, 24 vCPU), Pods $5.29+15.3%
VerdaServerless containers, 1x H200 SXM5 141GB $5.64+28.2%
Microsoft AzureStandard_ND96isr_H200_v5 (ND H200 v5) $5.82
Together AIH200 (140 GB), 256–1,000 GPUs $5.99
CoreWeaveNVIDIA HGX H200 $6.31
Google Clouda3-ultragpu-8g (A3 Ultra) $6.36+3.8%
AWSp5e.48xlarge (extra, for H200 comparison) $6.87+15%
Google Clouda3-ultragpu-8g (A3 Ultra) $7.31
Google Clouda3-ultragpu-8g (A3 Ultra) $7.42
AWSp5en.48xlarge $7.90+15%
Fireworks AIH200 141 GB $8.00+14.3%
Oracle OCIBM.GPU.H200.8 $10.00
Microsoft AzureStandard_ND96isr_H200_v5 (ND H200 v5) $10.60
Google Clouda3-ultragpu-8g (A3 Ultra) $10.60
See all 46 H200 offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| GMI Cloud | NVIDIA H200 | on-demand | from $2.60/GPU-hour (AVAILABLE NOW) | $2.60 | No change | RecheckingOct 10, 2026 | gmicloud.ai ↗ |
| Massed Compute | H200 NVL (141 GB) ×8 | on-demand | ×8 $28.96/hr, H200 NVL (141GB), on-demand per instance-hour | $3.62 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Massed Compute | H200 NVL NVLink ×8 | on-demand | ×8 $30.56/hr, H200 NVL (141GB) NVLink, on-demand per instance-hour | $3.82 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Hyperbolic | H200 | on-demand | $3.99 / HR, Nvidia H200, starting from (per GPU hour), refreshed weekly | $3.99 | No change | On pageOct 10, 2026 | hyperbolic.ai ↗ |
| Crusoe Cloud | NVIDIA H200 | on-demand | $4.29/GPU-hr, on-demand | $4.29 | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| Runpod | Clusters, H200 SXM | on-demand | $4.31/hr, Clusters | $4.31 | No change | On pageOct 10, 2026 | runpod.io ↗ |
| DigitalOcean | NVIDIA HGX H200 (141 GB) | on-demand | $4.47/GPU/hour, On-Demand Price | $4.47 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Nebius | NVIDIA HGX H200 | on-demand | $4.50 per GPU-hour, on-demand | $4.50 | No change | On pageOct 10, 2026 | nebius.com ↗ |
| Cirrascale | 8X H200 | $26,499 per server per month, Monthly Term, 8X NVIDIA H200 ($4.54/GPUhr equivalent shown) | $4.54 | No change | On pageOct 10, 2026 | cirrascale.com ↗ | |
| Verda | 1x H200 SXM5 141GB | on-demand | $5.12/h on-demand, 1x H200 SXM5 141GB, GPU instances, USD | $5.12 | +28% | On pageOct 10, 2026 | verda.com ↗ |
| Verda | Instant clusters, H200 SXM5 141GB | on-demand | $5.12/h per GPU, H200 SXM5 141GB, Instant clusters, USD | $5.12 | +28% | On pageOct 10, 2026 | verda.com ↗ |
| Runpod | H200 (141 GB, 24 vCPU), Pods | on-demand | $5.29/hr, Pods, Secure Cloud | $5.29 | +15.3% | On pageOct 10, 2026 | runpod.io ↗ |
| Verda | Serverless containers, 1x H200 SXM5 141GB | serverless | $5.64/h serverless, 1x H200 SXM5 141GB, Serverless containers, USD | $5.64 | +28.2% | On pageOct 10, 2026 | verda.com ↗ |
| Together AI | H200 (140 GB), 256–1,000 GPUs | on-demand | $5.99/hr per GPU, on-demand | $5.99 | No change | On pageOct 10, 2026 | together.ai ↗ |
| CoreWeave | NVIDIA HGX H200North America | on-demand | $50.44 per hour for 8 GPUs, on-demand, North America | $6.31 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| AWS | p5e.48xlarge (extra, for H200 comparison) | capacity block | $54.924 USD ($6.866 USD) per instance-hour (per accelerator), 8 x H200, US East (Ohio), US West (N. California), US West (Oregon), US West (Phoenix) Local Zone, Europe (Stockholm), Europe (London), Australia (Sydney), Asia Pacific (Jakarta), Asia Pacific (Mumbai), Asia Pacific (Tokyo), South America (Sao Paulo) | $6.87 | +15% | On pageOct 10, 2026 | aws.amazon.com ↗ |
| Google Cloud | a3-ultragpu-8g (A3 Ultra) | calendar mode | $59.3600 / 1 hour for a3-ultragpu-8g (Nvidia H200, GPUs: 8), DWS Calendar Mode price (USD) | $7.42 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| AWS | p5en.48xlarge | capacity block | $63.158 USD ($7.895 USD) per instance-hour (per accelerator), 8 x H200, US East (Ohio), US East (N. Virginia), US West (N. California), US West (Oregon), Europe (Stockholm), Europe (London), Europe (Spain), Asia Pacific (Jakarta), Asia Pacific (Mumbai), Asia Pacific (Tokyo), Asia Pacific (Seoul), US East (Atlanta) Local Zone | $7.90 | +15% | On pageOct 10, 2026 | aws.amazon.com ↗ |
| Fireworks AI | H200 141 GB | on-demand | $8.00 per GPU-hour, on-demand deployments, from Sep 1 | $8.00 | +14.3% | Not foundOct 10, 2026 | fireworks.ai ↗ |
| Oracle OCI | BM.GPU.H200.8 | on-demand | $10.00 | $10.00 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Microsoft Azure | Standard_ND96isr_H200_v5 (ND H200 v5)East US 2 | on-demand | 84.8 USD per 1 Hour for 8 GPUs, Linux, East US 2 | $10.60 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Google Cloud | a3-ultragpu-8g (A3 Ultra) | on-demand | $84.806908493 / 1 hour for a3-ultragpu-8g (Nvidia H200, GPUs: 8), Price (USD) | $10.60 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Committed and reserved | |||||||
| Impossible Cloud | H200 SXM 8-GPU nodes (EU, offer 287-H200)EU | 12-month term | $3.11 per GPU-hour, 12-month term, 2 nodes x 8 GPUs (16 GPUs), EU | $3.11 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Impossible Cloud | H200 SXM 8-GPU nodes (APAC, offer 294-H200)APAC | 6-month term | $3.16 per GPU-hour, 6-month term, 8 nodes x 8 GPUs (64 GPUs), APAC, available 8 Oct 2026 | $3.16 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| TheAI Cloud | 1-year reserved | 1-year reserved | $3.20/GPU·hr on a 1-year term, NVIDIA H200, available now, reserved whole node | $3.20 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| DigitalOcean | 12-month reserved | 12-month reserved | $3.40/GPU/hour, 12 Month Reserved Price | $3.40 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| TheAI Cloud | 6-month reserved | 6-month reserved | $3.45/GPU·hr on a 6-month term, NVIDIA H200, available now, reserved whole node | $3.45 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Cirrascale | 8X NVIDIA H200 | annual | $21,199 per server per month, Annual Term, 8X NVIDIA H200 ($3.63/GPUhr equivalent shown) | $3.63 | No change | On pageOct 10, 2026 | cirrascale.com ↗ |
| Astra Intellectus | H200 SXM 8-GPUUS #2 | 3-year lease | $3.65/GPU/Hour, H200 SXM 8-GPU, US #2, 128 nodes, 3-year lease | $3.65 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| Astra Intellectus | H200 SXM 8-GPU (TH #1, 128 nodes)TH #1 | 3-year lease | $3.65/GPU/Hour, H200 SXM 8-GPU, TH #1, 128 nodes, 3-year lease | $3.65 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| TheAI Cloud | 3-month reserved | 3-month reserved | $3.70/GPU·hr on a 3-month term, NVIDIA H200, available now, reserved whole node | $3.70 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Together AI | reserved | reserved | Starting at $3.99/hr per GPU, reserved | $3.99 | No change | On pageOct 10, 2026 | together.ai ↗ |
| Google Cloud | 3-year committed use | 3-year committed use | $37.208420822 / 1 hour for a3-ultragpu-8g (Nvidia H200, GPUs: 8), Compute Resource CUDs - 3 Year (USD) | $4.65 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Microsoft Azure | Standard_ND96isr_H200_v5 (ND H200 v5)East US 2 | 3-year reserved | 1109592 USD for one VM with 8 GPUs, reservation term 3 Years, East US 2 | $5.28 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND96isr_H200_v5 (ND H200 v5)East US 2 | 1-year reserved | 407526 USD for one VM with 8 GPUs, reservation term 1 Year, East US 2 | $5.82 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Google Cloud | a3-ultragpu-8g (A3 Ultra) | 1-year committed use | $58.471933151 / 1 hour for a3-ultragpu-8g (Nvidia H200, GPUs: 8), Compute Resource CUDs - 1 Year (USD) | $7.31 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Spot and preemptible | |||||||
| Nebius | NVIDIA HGX H200 | spot | $2.45 per GPU-hour, preemptible | $2.45 | No change | Not foundOct 10, 2026 | nebius.com ↗ |
| Verda | 1x H200 SXM5 141GB | spot | $2.56/h spot, 1x H200 SXM5 141GB, GPU instances, USD | $2.56 | +28% | On pageOct 10, 2026 | verda.com ↗ |
| CoreWeave | NVIDIA HGX H200North America | spot | $20.93 per hour for 8 GPUs, spot, North America | $2.62 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Verda | Serverless containers, 1x H200 SXM5 141GB | spot | $2.82/h spot, 1x H200 SXM5 141GB, Serverless containers, USD | $2.82 | +28.2% | On pageOct 10, 2026 | verda.com ↗ |
| Google Cloud | a3-ultragpu-8g (A3 Ultra) | spot | $50.834997041 / 1 hour for a3-ultragpu-8g (Nvidia H200, GPUs: 8), Current Spot pricing (USD) | $6.36 | +3.8% | On pageOct 10, 2026 | cloud.google.com ↗ |
| No public price | |||||||
| Denvr Dataworks | NVIDIA H200 SXM (8 GPUs) | not published | , | Not published | No change | On pageOct 10, 2026 | denvr.com ↗ |
| Fluidstack | H200 | not published | , | Not published | No change | ResearchSep 10, 2026 | fluidstack.io ↗ |
| Lambda | 1-Click Clusters, H200 | not published | n.a. | Not published | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Lightning AI | H200 | not published | Cost/GPU/hr* | Not published | No change | ResearchSep 10, 2026 | lightning.ai ↗ |
| Vultr | H200 | not published | Vultr | Not published | No change | ResearchSep 10, 2026 | vultr.com ↗ |
NVIDIA B200: every price, cheapest first
B200 price page →VultrNVIDIA HGX B200, Cloud GPU $3.50
VultrNVIDIA HGX B200, Bare Metal (8 × B200 192 GB) $3.50
Verda1x B200 SXM6 180GB $3.71+21.2%
NebiusNVIDIA HGX B200 $3.95
GMI CloudNVIDIA B200 $4.00
VerdaServerless containers, 1x B200 SXM6 180GB $4.08+21.4%
CoreWeaveNVIDIA HGX B200 $4.26
TheAI Cloud24-month reserved $4.50
Cirrascale8X NVIDIA B200 $4.79
TheAI Cloud1-year reserved $4.85
Google Clouda4-highgpu-8g (A4) $4.95
TheAI Cloud6-month reserved $5.35
Massed ComputeB200 SXM6 ×8 $5.43
TheAI Cloud3-month reserved $5.55
TheAI Cloud1-month reserved $5.75
HyperbolicB200 $5.99
Cirrascale8X B200 $5.99
LambdaNVIDIA B200 SXM6 $6.69
Together AIreserved $6.79
Google Cloud3-year committed use $7.09
NebiusNVIDIA HGX B200 $7.15
Verda1x B200 SXM6 180GB $7.42+21.4%
VerdaInstant clusters, B200 SXM6 180GB $7.42+21.4%
RunpodB200 (180 GB, 28 vCPU), Pods $7.99+17.7%
Google Clouda4-highgpu-8g (A4) $8.06
VerdaServerless containers, 1x B200 SXM6 180GB $8.16+21.4%
Together AIHGX B200 (180 GB) $8.19
CoreWeaveNVIDIA HGX B200 $8.60
Lambda1-Click Clusters, NVIDIA HGX B200 $8.87
BasetenB200 (180 GiB) $9.98
Google Clouda4-highgpu-8g (A4) $11.12
Google Clouda4-highgpu-8g (A4) $11.28
AWSp6-b200.48xlarge $12.36
Fireworks AIB200 180 GB $13.00+30%
Oracle OCIBM.GPU.B200.8 $14.00
See all 42 B200 offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| Vultr | NVIDIA HGX B200, Cloud GPU | on-demand | $3.500/GPU/hr | $3.50 | No change | BlockedOct 10, 2026 | vultr.com ↗ |
| GMI Cloud | NVIDIA B200 | on-demand | from $4.00/GPU-hour (Limited Availability) | $4.00 | No change | RecheckingOct 10, 2026 | gmicloud.ai ↗ |
| Massed Compute | B200 SXM6 ×8 | on-demand | ×8 $43.46/hr, B200 SXM6, on-demand per instance-hour | $5.43 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| Hyperbolic | B200 | on-demand | $5.99 / HR, Nvidia B200, starting from (per GPU hour), refreshed weekly | $5.99 | No change | On pageOct 10, 2026 | hyperbolic.ai ↗ |
| Cirrascale | 8X B200 | $34,999 per server per month, Monthly Term, 8X NVIDIA B200 ($5.99/GPUhr equivalent shown) | $5.99 | No change | On pageOct 10, 2026 | cirrascale.com ↗ | |
| Lambda | NVIDIA B200 SXM6 | on-demand | $6.69 per GPU-hour, 8x instance (208 vCPUs), on-demand | $6.69 | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Nebius | NVIDIA HGX B200 | on-demand | $7.15 per GPU-hour, on-demand | $7.15 | No change | On pageOct 10, 2026 | nebius.com ↗ |
| Verda | 1x B200 SXM6 180GB | on-demand | $7.42/h on-demand, 1x B200 SXM6 180GB, GPU instances, USD | $7.42 | +21.4% | On pageOct 10, 2026 | verda.com ↗ |
| Verda | Instant clusters, B200 SXM6 180GB | on-demand | $7.42/h per GPU, B200 SXM6 180GB, Instant clusters, USD | $7.42 | +21.4% | On pageOct 10, 2026 | verda.com ↗ |
| Runpod | B200 (180 GB, 28 vCPU), Pods | on-demand | $7.99/hr, Pods, Secure Cloud | $7.99 | +17.7% | On pageOct 10, 2026 | runpod.io ↗ |
| Google Cloud | a4-highgpu-8g (A4) | flex-start | $64.4400 / 1 hour for a4-highgpu-8g (Nvidia B200, GPUs: 8), DWS Flex-start price (USD) | $8.06 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Verda | Serverless containers, 1x B200 SXM6 180GB | serverless | $8.16/h serverless, 1x B200 SXM6 180GB, Serverless containers, USD | $8.16 | +21.4% | On pageOct 10, 2026 | verda.com ↗ |
| Together AI | HGX B200 (180 GB) | on-demand | $8.19/hr per GPU, on-demand | $8.19 | No change | On pageOct 10, 2026 | together.ai ↗ |
| CoreWeave | NVIDIA HGX B200North America | on-demand | $68.80 per hour for 8 GPUs, on-demand, North America | $8.60 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Baseten | B200 (180 GiB) | on-demand | $0.16633 per minute, B200 180 GiB VRAM, Dedicated Deployments | $9.98 | No change | On pageOct 10, 2026 | baseten.co ↗ |
| Google Cloud | a4-highgpu-8g (A4) | calendar mode | $90.22 / 1 hour for a4-highgpu-8g (Nvidia B200, GPUs: 8), DWS Calendar Mode price (USD) | $11.28 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| AWS | p6-b200.48xlarge | capacity block | $98.84 USD ($12.355 USD) per instance-hour (per accelerator), 8 x B200, US East (Ohio), US East (N. Virginia), US West (Oregon), Asia Pacific (Mumbai), Asia Pacific (Hyderabad), Europe (Stockholm), Europe (Paris); $102.960 USD ($12.870 USD) in 2 other regions | $12.36 | No change | Not foundOct 10, 2026 | aws.amazon.com ↗ |
| Fireworks AI | B200 180 GB | on-demand | $13.00 per GPU-hour, on-demand deployments, from Sep 1 | $13.00 | +30% | Not foundOct 10, 2026 | fireworks.ai ↗ |
| Oracle OCI | BM.GPU.B200.8 | on-demand | $14.00 | $14.00 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Committed and reserved | |||||||
| Vultr | NVIDIA HGX B200, Bare Metal (8 × B200 192 GB) | pre-paid contract | $3.500/GPU/hour | $3.50 | No change | BlockedOct 10, 2026 | vultr.com ↗ |
| TheAI Cloud | 24-month reserved | 24-month reserved | $4.50/GPU·hr on a 24-month term, NVIDIA B200, December 2026, reserved whole node | $4.50 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Cirrascale | 8X NVIDIA B200 | annual | $27,999 per server per month, Annual Term, 8X NVIDIA B200 ($4.79/GPUhr equivalent shown) | $4.79 | No change | On pageOct 10, 2026 | cirrascale.com ↗ |
| TheAI Cloud | 1-year reserved | 1-year reserved | $4.85/GPU·hr on a 1-year term, NVIDIA B200, December 2026, reserved whole node | $4.85 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| TheAI Cloud | 6-month reserved | 6-month reserved | $5.35/GPU·hr on a 6-month term, NVIDIA B200, December 2026, reserved whole node | $5.35 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| TheAI Cloud | 3-month reserved | 3-month reserved | $5.55/GPU·hr on a 3-month term, NVIDIA B200, December 2026, reserved whole node | $5.55 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| TheAI Cloud | 1-month reserved | 1-month reserved | $5.75/GPU·hr on a 1-month term, NVIDIA B200, December 2026, reserved whole node | $5.75 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Together AI | reserved | reserved | Starting at $6.79/hr per GPU, reserved | $6.79 | No change | On pageOct 10, 2026 | together.ai ↗ |
| Google Cloud | 3-year committed use | 3-year committed use | $56.7072 / 1 hour for a4-highgpu-8g (Nvidia B200, GPUs: 8), Compute Resource CUDs - 3 Year (USD) | $7.09 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Lambda | 1-Click Clusters, NVIDIA HGX B200 | short-term reserved cluster | $8.87 per GPU-hour for 256+ GPUs, 2 weeks – 1 year, 1-Click Clusters | $8.87 | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Google Cloud | a4-highgpu-8g (A4) | 1-year committed use | $88.9272 / 1 hour for a4-highgpu-8g (Nvidia B200, GPUs: 8), Compute Resource CUDs - 1 Year (USD) | $11.12 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| Spot and preemptible | |||||||
| Verda | 1x B200 SXM6 180GB | spot | $3.71/h spot, 1x B200 SXM6 180GB, GPU instances, USD | $3.71 | +21.2% | On pageOct 10, 2026 | verda.com ↗ |
| Nebius | NVIDIA HGX B200 | spot | $3.95 per GPU-hour, preemptible | $3.95 | No change | Not foundOct 10, 2026 | nebius.com ↗ |
| Verda | Serverless containers, 1x B200 SXM6 180GB | spot | $4.08/h spot, 1x B200 SXM6 180GB, Serverless containers, USD | $4.08 | +21.4% | On pageOct 10, 2026 | verda.com ↗ |
| CoreWeave | NVIDIA HGX B200North America | spot | $34.11 per hour for 8 GPUs, spot, North America | $4.26 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Google Cloud | a4-highgpu-8g (A4) | spot | $39.6336 / 1 hour for a4-highgpu-8g (Nvidia B200, GPUs: 8), Current Spot pricing (USD) | $4.95 | No change | On pageOct 10, 2026 | cloud.google.com ↗ |
| No public price | |||||||
| Crusoe Cloud | NVIDIA B200 | not published | /GPU-hr column | Not published | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| Denvr Dataworks | NVIDIA B200 SXM (8) | not published | , | Not published | No change | On pageOct 10, 2026 | denvr.com ↗ |
| DigitalOcean | B200 | not published | , | Not published | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Fluidstack | B200 | not published | , | Not published | No change | ResearchSep 10, 2026 | fluidstack.io ↗ |
| Google Cloud | a4x-highgpu-4g (A4X) | not published | GPU count from Compute Engine docs | Not published | No change | On pageOct 10, 2026 | docs.cloud.google.com ↗ |
| Lightning AI | B200 | not published | Cost/GPU/hr* | Not published | No change | ResearchSep 10, 2026 | lightning.ai ↗ |
| Runpod | Clusters, B200 | not published | , | Not published | No change | On pageOct 10, 2026 | runpod.io ↗ |
NVIDIA GB200 NVL72: every price, cheapest first
GB200 NVL72 price page →Microsoft AzureStandard_ND128isrf_NDR_GB200_v6 $6.22
GMI CloudNVIDIA GB200 $8.00
CoreWeaveNVIDIA GB200 NVL72 $10.50
AWSu-p6e-gb200x36 (UltraServer) $10.58
AWSu-p6e-gb200x72 (UltraServer) $10.58
Microsoft AzureStandard_ND128isr_NDR_GB200_v6 (ND GB200 v6) $11.90
Microsoft AzureStandard_ND128isrf_NDR_GB200_v6 $13.68
Oracle OCIBM.GPU.GB200.4 $16.00
Microsoft AzureStandard_ND128isr_NDR_GB200_v6 (ND GB200 v6) $17.31
Microsoft AzureStandard_ND128isrf_NDR_GB200_v6 $19.90
Microsoft AzureStandard_ND128isr_NDR_GB200_v6 (ND GB200 v6) $27.04
Microsoft AzureStandard_ND128isrf_NDR_GB200_v6 $31.10
See all 20 GB200 NVL72 offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| GMI Cloud | NVIDIA GB200 | on-demand | from $8.00/GPU-hour (AVAILABLE NOW) | $8.00 | No change | On pageOct 10, 2026 | gmicloud.ai ↗ |
| CoreWeave | NVIDIA GB200 NVL72North America | on-demand | $42.00 per hour for 4 GPUs, on-demand, North America | $10.50 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| AWS | u-p6e-gb200x36 (UltraServer) | capacity block | $380.952 ($10.582 USD) per UltraServer-hour (per accelerator), 36 x B200, US East (Dallas) Local Zone | $10.58 | No change | On pageOct 10, 2026 | aws.amazon.com ↗ |
| AWS | u-p6e-gb200x72 (UltraServer) | capacity block | $761.904 ($10.582 USD) per UltraServer-hour (per accelerator), 72 x B200, US East (Dallas) Local Zone | $10.58 | No change | On pageOct 10, 2026 | aws.amazon.com ↗ |
| Oracle OCI | BM.GPU.GB200.4 | on-demand | $16.00 | $16.00 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Microsoft Azure | Standard_ND128isr_NDR_GB200_v6 (ND GB200 v6)East US | on-demand | 108.16 USD per 1 Hour for 4 GPUs, Linux, East US | $27.04 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND128isrf_NDR_GB200_v6Korea South | on-demand | 124.4 USD per 1 Hour for 4 GPUs, Linux, Korea South | $31.10 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Committed and reserved | |||||||
| Microsoft Azure | Standard_ND128isr_NDR_GB200_v6 (ND GB200 v6)East US | 3-year reserved | 1250676 USD for one VM with 4 GPUs, reservation term 3 Years, East US | $11.90 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND128isrf_NDR_GB200_v6Korea South | 3-year reserved | 1438462 USD for one VM with 4 GPUs, reservation term 3 Years, Korea South | $13.68 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND128isr_NDR_GB200_v6 (ND GB200 v6)East US | 1-year reserved | 606388 USD for one VM with 4 GPUs, reservation term 1 Year, East US | $17.31 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Microsoft Azure | Standard_ND128isrf_NDR_GB200_v6Korea South | 1-year reserved | 697436 USD for one VM with 4 GPUs, reservation term 1 Year, Korea South | $19.90 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Spot and preemptible | |||||||
| Microsoft Azure | Standard_ND128isrf_NDR_GB200_v6Korea South | spot | 24.88 USD per 1 Hour for 4 GPUs, Linux Spot, Korea South | $6.22 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| No public price | |||||||
| Crusoe Cloud | NVIDIA GB200 | not published | /GPU-hr column | Not published | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| Denvr Dataworks | GB200 (2) | not published | , | Not published | No change | On pageOct 10, 2026 | denvr.com ↗ |
| DigitalOcean | GB200 | not published | , | Not published | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Google Cloud | a4x-highgpu-4g (A4X) | not published | GPU count from Compute Engine docs | Not published | No change | On pageOct 10, 2026 | docs.cloud.google.com ↗ |
| Lambda | 1-Click Clusters, GB200 | not published | n.a. | Not published | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Nebius | NVIDIA GB200 NVL72 | not published | On-demand, GPU-hour" header, no value | Not published | No change | On pageOct 10, 2026 | nebius.com ↗ |
| Together AI | GB200 NVL72 (186 GB), 512–1,000+ GPUs | not published | , | Not published | No change | On pageOct 10, 2026 | together.ai ↗ |
| Vultr | GB200 | not published | Vultr | Not published | No change | ResearchSep 10, 2026 | vultr.com ↗ |
NVIDIA B300: every price, cheapest first
B300 price page →Astra IntellectusB300 SXM 8-GPU (US #1, 128 nodes) $3.65
Astra IntellectusB300 SXM 8-GPU (US #3, 128 nodes) $3.65
Astra IntellectusB300 SXM 8-GPU $3.85
NebiusNVIDIA HGX B300 $4.30
Impossible CloudB300 SXM 8-GPU nodes (North America, offer 290-B300) $4.32
CoreWeaveNVIDIA HGX B300 $4.48
Verda1x B300 SXM6 268GB $4.67+24.5%
Impossible CloudB300 SXM 8-GPU nodes (EU, offer 293-B300) $4.74
Impossible CloudB300 8-GPU nodes (US, offer 289-B300) $4.84
Impossible CloudB300 SXM 8-GPU nodes (EU, offer 291-B300) $4.95
VerdaServerless containers, 1x B300 SXM6 268GB $5.14+24.5%
Impossible CloudB300 SXM 8-GPU nodes (US, offer 292-B300) $5.21
TheAI Cloud24-month reserved $5.30+8.2%
Impossible CloudB300 SXM 8-GPU nodes (North America, offer 290-B300) $5.37
Astra IntellectusB300 SXM 8-GPU (MY #1, 128 nodes) $5.45
Impossible CloudB300 8-GPU nodes (US, offer 289-B300) $5.47
Astra IntellectusB300 SXM 8-GPU (US #3, 128 nodes) $5.65
Astra IntellectusB300 SXM 8-GPU (MY #2, 64 nodes) $5.75
Astra IntellectusB300 SXM 8-GPU (EU #3, 64 nodes) $6.00
TheAI Cloud1-year reserved $6.05+14.2%
TheAI Cloud6-month reserved $6.50+14%
Impossible CloudB300 SXM 8-GPU nodes (EU, offer 293-B300) $6.53
Massed ComputeB300 SXM6 ×8 (280 vCPU, 2800 GB) $6.60
TheAI Cloud3-month reserved $6.75+3.1%
TheAI CloudB300 (8× B300 SXM6 · 288GB … whole node) $6.90
TheAI Cloud1-month reserved $7.50+11.1%
NebiusNVIDIA HGX B300 $7.85
Impossible CloudB300 SXM 8-GPU nodes (EU, offer 288-B300) $7.89
DigitalOceanNVIDIA HGX B300 $7.94
RunpodB300 (288 GB HBM3e, 32 vCPU), Pods $8.99+13.9%
Verda1x B300 SXM6 268GB $9.34+24.5%
VerdaInstant clusters, B300 SXM6 268GB $9.34+24.5%
RunpodServerless, B300 $9.98
Together AIHGX B300 (270 GB) $9.99
VerdaServerless containers, B300 $10.27+24.5%
AWSp6-b300.48xlarge $14.04
Fireworks AIB300 288 GB $15.00+25%
Oracle OCIBM.GPU.B300.8 $15.00
See all 43 B300 offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| Massed Compute | B300 SXM6 ×8 (280 vCPU, 2800 GB) | on-demand | ×8 $52.80/hr, B300 SXM6, on-demand per instance-hour | $6.60 | No change | On pageOct 10, 2026 | vm.massedcompute.com ↗ |
| TheAI Cloud | B300 (8× B300 SXM6 · 288GB … whole node) | on-demand | $6.90/GPU·HR on-demand ("PAY AS YOU GO", per-minute); reserved $6.75 (1 mo), $6.55 (3 mo), $5.50 (6 mo), $5.30/GPU·hr (1 yr) | $6.90 | No change | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Nebius | NVIDIA HGX B300 | on-demand | $7.85 per GPU-hour, on-demand | $7.85 | No change | On pageOct 10, 2026 | nebius.com ↗ |
| Runpod | B300 (288 GB HBM3e, 32 vCPU), Pods | on-demand | $8.99/hr, Pods, Secure Cloud | $8.99 | +13.9% | On pageOct 10, 2026 | runpod.io ↗ |
| Verda | 1x B300 SXM6 268GB | on-demand | $9.34/h on-demand, 1x B300 SXM6 268GB, GPU instances, USD | $9.34 | +24.5% | On pageOct 10, 2026 | verda.com ↗ |
| Verda | Instant clusters, B300 SXM6 268GB | on-demand | $9.34/h per GPU, B300 SXM6 268GB, Instant clusters, USD | $9.34 | +24.5% | On pageOct 10, 2026 | verda.com ↗ |
| Runpod | Serverless, B300 | serverless | $9.98/hr, Serverless | $9.98 | No change | On pageOct 10, 2026 | runpod.io ↗ |
| Together AI | HGX B300 (270 GB) | $9.99/hr per GPU, on-demand | $9.99 | No change | On pageOct 10, 2026 | together.ai ↗ | |
| Verda | Serverless containers, B300 | serverless | $10.27/h serverless, 1x B300 SXM6 268GB, Serverless containers, USD | $10.27 | +24.5% | On pageOct 10, 2026 | verda.com ↗ |
| AWS | p6-b300.48xlarge | capacity block | $112.32 USD ($14.04 USD) per instance-hour (per accelerator), 8 x B300, US West (Oregon), US East (N. Virginia), US East (Atlanta) Local Zone, Asia Pacific (Seoul), Asia Pacific (Hyderabad), Asia Pacific (Jakarta); $117.00 USD ($14.625 USD) in 1 other region | $14.04 | No change | Not foundOct 10, 2026 | aws.amazon.com ↗ |
| Fireworks AI | B300 288 GB | on-demand | $15.00 per GPU-hour, on-demand deployments, from Sep 1 | $15.00 | +25% | Not foundOct 10, 2026 | fireworks.ai ↗ |
| Oracle OCI | BM.GPU.B300.8 | on-demand | $15.00 | $15.00 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Committed and reserved | |||||||
| Astra Intellectus | B300 SXM 8-GPU (US #1, 128 nodes)US #1 | 3-year lease | $3.65/GPU/Hour, B300 SXM 8-GPU, US #1, 128 nodes, 3-year lease | $3.65 | No change | RecheckingOct 10, 2026 | astra-intellectus.com ↗ |
| Astra Intellectus | B300 SXM 8-GPU (US #3, 128 nodes)US #3 | 3-year lease | $3.65/GPU/Hour, B300 SXM 8-GPU, US #3, 128 nodes, 3-year lease | $3.65 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| Astra Intellectus | B300 SXM 8-GPUEU #3 | 3-year lease | $3.85/GPU/Hour, B300 SXM 8-GPU, EU #3, 64 nodes, 3-year lease | $3.85 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| Impossible Cloud | B300 SXM 8-GPU nodes (North America, offer 290-B300)North America | 60-month term | $4.32 per GPU-hour, 60-month term, 4 nodes x 8 GPUs (32 GPUs), North America | $4.32 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Impossible Cloud | B300 SXM 8-GPU nodes (EU, offer 293-B300)EU | 36-month term | $4.74 per GPU-hour, 36-month term, 1 node x 8 GPUs (8 GPUs), EU, available 1 Jan 2027 | $4.74 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Impossible Cloud | B300 8-GPU nodes (US, offer 289-B300)US | 60-month term | $4.84 per GPU-hour, 60-month term, 16 nodes x 8 GPUs (128 GPUs), US, available 2026 Dec (RFS), 8-12 weeks after MSA | $4.84 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Impossible Cloud | B300 SXM 8-GPU nodes (EU, offer 291-B300)EU | 36 to 60-month term | $4.95 per GPU-hour, 36 to 60-month term, 64 nodes x 8 GPUs (512 GPUs), EU, available 1 Mar 2027 | $4.95 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Impossible Cloud | B300 SXM 8-GPU nodes (US, offer 292-B300)US | 36-month term | $5.21 per GPU-hour, 36-month term, 24 nodes x 8 GPUs (192 GPUs), US | $5.21 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| TheAI Cloud | 24-month reserved | 24-month reserved | $5.30/GPU·hr on a 24-month term, NVIDIA B300, reserved whole node | $5.30 | +8.2% | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Impossible Cloud | B300 SXM 8-GPU nodes (North America, offer 290-B300)North America | 36-month term | $5.37 per GPU-hour, 36-month term, 4 nodes x 8 GPUs (32 GPUs), North America | $5.37 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Astra Intellectus | B300 SXM 8-GPU (MY #1, 128 nodes)MY #1 | 5-year lease | $5.45/GPU/Hour, B300 SXM 8-GPU, MY #1, 128 nodes, 5-year lease | $5.45 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| Impossible Cloud | B300 8-GPU nodes (US, offer 289-B300)US | 36-month term | $5.47 per GPU-hour, 36-month term, 16 nodes x 8 GPUs (128 GPUs), US, available 2026 Dec (RFS), 8-12 weeks after MSA | $5.47 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| Astra Intellectus | B300 SXM 8-GPU (US #3, 128 nodes)US #3 | 5-year lease | $5.65/GPU/Hour, B300 SXM 8-GPU, US #3, 128 nodes, 5-year lease | $5.65 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| Astra Intellectus | B300 SXM 8-GPU (MY #2, 64 nodes)MY #2 | 3-year lease | $5.75/GPU/Hour, B300 SXM 8-GPU, MY #2, 64 nodes, 3-year lease | $5.75 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| Astra Intellectus | B300 SXM 8-GPU (EU #3, 64 nodes)EU #3 | 5-year lease | $6.00/GPU/Hour, B300 SXM 8-GPU, EU #3, 64 nodes, 5-year lease | $6.00 | No change | On pageOct 10, 2026 | astra-intellectus.com ↗ |
| TheAI Cloud | 1-year reserved | 1-year reserved | $6.05/GPU·hr on a 1-year term, NVIDIA B300, reserved whole node | $6.05 | +14.2% | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| TheAI Cloud | 6-month reserved | 6-month reserved | $6.50/GPU·hr on a 6-month term, NVIDIA B300, reserved whole node | $6.50 | +14% | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Impossible Cloud | B300 SXM 8-GPU nodes (EU, offer 293-B300)EU | 24-month term | $6.53 per GPU-hour, 24-month term, 1 node x 8 GPUs (8 GPUs), EU, available 1 Jan 2027 | $6.53 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| TheAI Cloud | 3-month reserved | 3-month reserved | $6.75/GPU·hr on a 3-month term, NVIDIA B300, reserved whole node | $6.75 | +3.1% | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| TheAI Cloud | 1-month reserved | 1-month reserved | $7.50/GPU·hr on a 1-month term, NVIDIA B300, reserved whole node | $7.50 | +11.1% | Not foundOct 10, 2026 | cloud.theai.com ↗ |
| Impossible Cloud | B300 SXM 8-GPU nodes (EU, offer 288-B300)EU | 12-month term | $7.89 per GPU-hour, 12-month term, 60 nodes x 8 GPUs (480 GPUs), EU | $7.89 | No change | FeedOct 11, 2026 | console.impossiblecloud.com ↗ |
| DigitalOcean | NVIDIA HGX B300 | 12-month reserved | $7.94/GPU/hour, 12 Month Reserved Price | $7.94 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Spot and preemptible | |||||||
| Nebius | NVIDIA HGX B300 | spot | $4.30 per GPU-hour, preemptible | $4.30 | No change | Not foundOct 10, 2026 | nebius.com ↗ |
| CoreWeave | NVIDIA HGX B300North America | spot | $35.84 per hour for 8 GPUs, spot, North America | $4.48 | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Verda | 1x B300 SXM6 268GB | spot | $4.67/h spot, 1x B300 SXM6 268GB, GPU instances, USD | $4.67 | +24.5% | On pageOct 10, 2026 | verda.com ↗ |
| Verda | Serverless containers, 1x B300 SXM6 268GB | spot | $5.14/h spot, 1x B300 SXM6 268GB, Serverless containers, USD | $5.14 | +24.5% | On pageOct 10, 2026 | verda.com ↗ |
| No public price | |||||||
| AWS | p6e-b300 | not published | n.a. | Not published | No change | On pageOct 10, 2026 | aws.amazon.com ↗ |
| Crusoe Cloud | B300 | not published | , | Not published | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| DigitalOcean | B300 (non-reserved) | not published | , | Not published | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Lambda | 1-Click Clusters, B300 | not published | n.a. | Not published | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Vultr | NVIDIA HGX B300, Cloud GPU | not published | listed with GPU count 8 | Not published | No change | ResearchSep 10, 2026 | vultr.com ↗ |
NVIDIA GB300 NVL72: every price, cheapest first
GB300 NVL72 price page →Astra IntellectusGB300 NVL72 (EU #2, 30 racks) $3.70
Verda1x GB300 SXM6 288GB $5.37+24.6%
Verda24-month $8.06+24.6%
Verda1x GB300 SXM6 288GB $10.74+24.6%
VerdaInstant Clusters, GB300 $10.74+24.6%
Oracle OCIBM.GPU.GB300.4 $18.00
Fireworks AIGB300 288 GB $20.00+11.1%
See all 16 GB300 NVL72 offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| Verda | 1x GB300 SXM6 288GB | on-demand | $10.74/h on-demand, 1x GB300 SXM6 288GB, GPU instances, USD | $10.74 | +24.6% | On pageOct 10, 2026 | verda.com ↗ |
| Verda | Instant Clusters, GB300 | on-demand | $10.74/h per GPU, GB300 SXM6 288GB, Instant clusters, USD | $10.74 | +24.6% | On pageOct 10, 2026 | verda.com ↗ |
| Oracle OCI | BM.GPU.GB300.4 | on-demand | $18.00 | $18.00 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Fireworks AI | GB300 288 GB | on-demand | $20.00 per GPU-hour, on-demand deployments, from Sep 1 | $20.00 | +11.1% | Not foundOct 10, 2026 | fireworks.ai ↗ |
| Committed and reserved | |||||||
| Astra Intellectus | GB300 NVL72 (EU #2, 30 racks)EU #2 | 3-year lease | $3.70/GPU/Hour, GB300 NVL72, EU #2, 30 racks, 3-year lease | $3.70 | No change | RecheckingOct 10, 2026 | astra-intellectus.com ↗ |
| Verda | 24-month | 24-month | 2 years: 25% off on-demand ($10.74/h on-demand, 1x GB300 SXM6 288GB), USD | $8.06 | +24.6% | On pageOct 10, 2026 | verda.com ↗ |
| Spot and preemptible | |||||||
| Verda | 1x GB300 SXM6 288GB | spot | $5.37/h spot, 1x GB300 SXM6 288GB, GPU instances, USD | $5.37 | +24.6% | On pageOct 10, 2026 | verda.com ↗ |
| No public price | |||||||
| CoreWeave | NVIDIA GB300 NVL72North America | not published | GPU Count = 4 | Not published | No change | On pageOct 10, 2026 | coreweave.com ↗ |
| Crusoe Cloud | GB300 | not published | , | Not published | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| DigitalOcean | GB300 | not published | , | Not published | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| GMI Cloud | NVIDIA GB300 | not published | GPU-hour | Not published | No change | On pageOct 10, 2026 | gmicloud.ai ↗ |
| Lambda | 1-Click Clusters, GB300 | not published | n.a. | Not published | No change | On pageOct 10, 2026 | lambda.ai ↗ |
| Microsoft Azure | Standard_ND128isr_GB300_v6 / Standard_ND128isrG5_GB300_v6 (ND GB300 v6) | not published | size doc lists "4 GPUs" per VM | Not published | No change | FeedOct 11, 2026 | learn.microsoft.com ↗ |
| Nebius | NVIDIA GB300 NVL72 | not published | same | Not published | No change | On pageOct 10, 2026 | nebius.com ↗ |
| Together AI | GB300 NVL72 (288 GB) | not published | , | Not published | No change | On pageOct 10, 2026 | together.ai ↗ |
| Vultr | NVIDIA GB300 NVL72, Cloud GPU | not published | 4 GPUs per compute node; 18 nodes = 72 GPUs per rack | Not published | No change | ResearchSep 10, 2026 | vultr.com ↗ |
One Vera Rubin GPU-hour, sold as tokens
Anthropic pays Nscale about $5.83 for an hour of one Vera Rubin GPU. SAVRN calculates that hour writes about 854,067 tokens and reads about 6,832,536, from SemiAnalysis's published cost and price per GPU-hour. At Claude Fable 5.1 list prices, those tokens bill $111.03: 19.0x what Anthropic pays for the hour.
1 of 72GPUs in one Vera Rubin NVL72 rack. Every bar is that one GPU for one hour.
SAVRN also trains the model on the customer's own data there, which a frontier lab's API cannot do, making it a better model for that customer's use than the frontier model it is compared against here.
- Tokens written854,067SAVRN's calculation, not a published figure: the $3.57 cost per GPU-hour ÷ $4.18 per million output tokens at 300 tokens a second per user, both published
- Tokens read6,832,5368 input tokens for every output token: the test sends 8,000 in and gets 1,000 back
- Output billed$42.70854,067 × $50.00 per million
- Input billed$68.336,832,536 × $10.00 per million
- One GPU-hour$111.03× 72 GPUs × 8,760 hours = $70.0M a rack-year
- The deployed rack paid for in2.1 monthsthe $12.5M all-in deployed estimate ÷ $70.0M of tokens a year
Tokens per hour from CoreWeave's measurements of DeepSeek R1 on Vera Rubin NVL72, costed by SemiAnalysis, and SAVRN's own arithmetic on those two published figures. Prices from each lab's published price list, tracked on What frontier models charge. NVIDIA has published no Vera Rubin list price. $12.5M all in deployed is SAVRN estimate, not a published figure. The hardware alone has been reported at: Morgan Stanley, via Yahoo Finance, $7.80M, May 22, 2026; Bernstein, via TrendForce, Up to $9.1M, Jul 28, 2026. SAVRN's own estimate, not a published figure: $8.0M to $9.0M.
AMD Instinct MI300X: every price, cheapest first
MI300X price page →Microsoft AzureStandard_ND96isr_MI300X_v5 $1.11
VultrAMD MI300X, Cloud GPU / Bare Metal $1.85
DigitalOcean12-month reserved $1.91
DigitalOceanAMD MI300X (192 GB) $2.59
Microsoft AzureStandard_ND96isr_MI300X_v5 $2.63
Cirrascale8X AMD MI300X $3.08
Crusoe CloudAMD MI300X $3.45
Microsoft AzureStandard_ND96isr_MI300X_v5 $3.84
Cirrascale8X MI300X $3.85
Microsoft AzureStandard_ND96isr_MI300X_v5 $6.00
Oracle OCIBM.GPU.MI300X.8 $6.00
See all 13 MI300X offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| Vultr | AMD MI300X, Cloud GPU / Bare Metal | on-demand | $1.850/GPU/hr (cloud); $1.750/GPU/hour (bare metal) | $1.85 | No change | BlockedOct 10, 2026 | vultr.com ↗ |
| DigitalOcean | AMD MI300X (192 GB) | on-demand | $2.59/GPU/hour, On-Demand Price | $2.59 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Crusoe Cloud | AMD MI300X | on-demand | $3.45/GPU-hr, on-demand | $3.45 | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| Cirrascale | 8X MI300X | $22,499 per server per month, Monthly Term, 8X AMD MI300X ($3.85/GPUhr equivalent shown) | $3.85 | No change | On pageOct 10, 2026 | cirrascale.com ↗ | |
| Microsoft Azure | Standard_ND96isr_MI300X_v5East US 2 | on-demand | 48 USD per 1 Hour for 8 GPUs, Linux, East US 2 | $6.00 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Oracle OCI | BM.GPU.MI300X.8 | on-demand | $6.00 | $6.00 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Committed and reserved | |||||||
| DigitalOcean | 12-month reserved | 12-month reserved | $1.91/GPU/hour, 12 Month Reserved Price | $1.91 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Microsoft Azure | Standard_ND96isr_MI300X_v5East US 2 | 3-year reserved | 553772 USD for one VM with 8 GPUs, reservation term 3 Years, East US 2 | $2.63 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Cirrascale | 8X AMD MI300X | annual | $17,999 per server per month, Annual Term, 8X AMD MI300X ($3.08/GPUhr equivalent shown) | $3.08 | No change | On pageOct 10, 2026 | cirrascale.com ↗ |
| Microsoft Azure | Standard_ND96isr_MI300X_v5East US 2 | 1-year reserved | 269107 USD for one VM with 8 GPUs, reservation term 1 Year, East US 2 | $3.84 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| Spot and preemptible | |||||||
| Microsoft Azure | Standard_ND96isr_MI300X_v5East US 2 | spot | 8.8704 USD per 1 Hour for 8 GPUs, Linux Spot, East US 2 | $1.11 | No change | FeedOct 11, 2026 | prices.azure.com ↗ |
| No public price | |||||||
| Denvr Dataworks | AMD MI300X (8) | not published | , | Not published | No change | On pageOct 10, 2026 | denvr.com ↗ |
| TensorWave | MI300X | not published | , | Not published | No change | ResearchSep 10, 2026 | tensorwave.com ↗ |
AMD Instinct MI325X: every price, cheapest first
MI325X price page →VultrAMD MI325X, Cloud GPU / Bare Metal $2.00
DigitalOcean12-month reserved $2.88
DigitalOceanAMD MI325X (256 GB) $3.80
See all 5 MI325X offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| Vultr | AMD MI325X, Cloud GPU / Bare Metal | on-demand | $2.000/GPU/hr; $2.000/GPU/hour | $2.00 | No change | BlockedOct 10, 2026 | vultr.com ↗ |
| DigitalOcean | AMD MI325X (256 GB) | on-demand | $3.80/GPU/hour, On-Demand Price | $3.80 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| Committed and reserved | |||||||
| DigitalOcean | 12-month reserved | 12-month reserved | $2.88/GPU/hour, 12 Month Reserved Price | $2.88 | No change | On pageOct 10, 2026 | digitalocean.com ↗ |
| No public price | |||||||
| Crusoe Cloud | MI325X | not published | , | Not published | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| TensorWave | MI325X | not published | , | Not published | No change | ResearchSep 10, 2026 | tensorwave.com ↗ |
AMD Instinct MI355X: every price, cheapest first
MI355X price page →VultrAMD MI355X, Cloud GPU / Bare Metal $2.59
DigitalOceanAMD MI355X $2.97−34%
Oracle OCIBM.GPU.MI355X.8 $8.60
See all 5 MI355X offers as a table
| Provider | Offer | Terms | Published price | Per GPU-hour | Change | Checked | Source |
|---|---|---|---|---|---|---|---|
| List prices | |||||||
| Vultr | AMD MI355X, Cloud GPU / Bare Metal | on-demand | $2.590/GPU/hr; $2.650/GPU/hour | $2.59 | No change | BlockedOct 10, 2026 | vultr.com ↗ |
| Oracle OCI | BM.GPU.MI355X.8 | on-demand | $8.60 | $8.60 | No change | ResearchSep 10, 2026 | oracle.com ↗ |
| Spot and preemptible | |||||||
| DigitalOcean | AMD MI355X | spot | $2.97/GPU/hour, Current Spot Price | $2.97 | −34% | RecheckingOct 10, 2026 | digitalocean.com ↗ |
| No public price | |||||||
| Crusoe Cloud | AMD MI355X | not published | /GPU-hr column | Not published | No change | On pageOct 10, 2026 | crusoe.ai ↗ |
| TensorWave | MI355X | not published | , | Not published | No change | ResearchSep 10, 2026 | tensorwave.com ↗ |
Other accelerators
NVIDIA GH200
LambdaNVIDIA GH200 $2.29
Denvr DataworksGH200 $3.87
CoreWeaveNVIDIA GH200 $6.50
AMD Instinct MI350X
DigitalOceanAMD MI350X $4.76
Seven references, one system
Capital, delay, policy, water, grid, permits and scarcity are one system seen from seven sides. Each reference is primary-sourced, refreshed on a schedule, and links into the others, so a project can be followed from the money committed to the docket that stopped it.




