Spot vs on-demand GPU prices
Compare observed Spot discounts within matching configurations. Cheaper interruptible capacity comes with service continuity risks; a discount is not a reliability or availability guarantee.
Current means eligible observations in this build’s dated database snapshot, not a new availability check or a market-wide minimum. Whole-GPU USD quotes only; standard and unspecified regions, tax excluded or unknown. Spot stays separate. Stale, fractional, advertised starting prices and GPU-only components do not enter the calculations. Source, tax, marketplace, FX and billing qualifiers remain on the offer cards.
Price observations dated: 2026-11-18
Match the exact model, provider, SKU, region, GPU count, allocation fraction, scope, currency, tax basis, CPU, RAM and initial billing duration. For each matched pair compute the percentage reduction from on-demand to Spot, then take the median of those reductions per popular model. Negative values mean Spot was dearer. The two price medians describe the matched cohort; their ratio need not equal the median discount. No cross-provider or cross-region pairing; regional from-price floors are excluded. Sample counts measure matched configurations, not capacity or interruption probability.
AWS interruption notices are best effort and hibernation can begin immediately. Azure may evict Spot VMs when capacity is needed or the configured maximum price is exceeded. Google Cloud can reclaim Spot VMs and does not provide the same availability commitments as standard VMs. Plan checkpointing, retries and recovery, and check the official provider documentation linked below. These observations do not estimate interruption rates.
| GPU model | Matched pairs | On-demand median (USD/GPU-hour) | Spot median (USD/GPU-hour) | Median paired discount |
|---|---|---|---|---|
| H100 NVL 94GB | 54 | $8.72 | $1.84 | 79.28% |
| H100 SXM 80GB | 27 | $15.12 | $2.79 | 81.52% |
| H200 141GB | 30 | $13.52 | $13.52 | 0.00% |
| H200 SXM5 141GB | 4 | $6.90 | $3.45 | 50.00% |
| B200 CC SXM6 180GB | 6 | $9.99 | $5.00 | 50.00% |
| B200 SXM6 180GB | 4 | $9.80 | $4.90 | 50.00% |
| B300 CC SXM6 268GB | 4 | $13.05 | $6.52 | 50.01% |
| B300 SXM6 268GB | 4 | $12.79 | $6.39 | 50.00% |
| GB300 SXM6 288GB | 3 | $14.71 | $7.36 | 50.00% |
| A100 40GB | 5 | $4.08 | $0.885 | 78.29% |
| A100 80GB | 26 | $5.12 | $1.38 | 74.24% |
| A100 PCIe 80GB | 1 | $0.354 | $0.250 | 29.41% |
| A100 SXM4 40GB | 4 | $1.37 | $0.808 | 38.15% |
| A100 SXM4 80GB | 4 | $2.42 | $1.21 | 50.00% |
| L40S 48GB | 4 | $2.08 | $1.04 | 50.00% |
| RTX 5090 32GB | 118 | $0.542 | $0.531 | 0.00% |
| RTX 4090 24GB | 56 | $0.417 | $0.417 | 0.00% |
| RTX 6000 Ada 48GB | 4 | $1.65 | $0.827 | 50.00% |
| RTX PRO 6000 96GB | 4 | $2.92 | $1.46 | 50.00% |
| RTX PRO 6000 Blackwell MaxQ Works 96GB | 1 | $1.88 | $1.88 | 0.00% |
| RTX PRO 6000 Blackwell Workstation 96GB | 3 | $1.67 | $1.67 | 0.00% |
| RTX PRO 6000 CC 96GB | 1 | $2.98 | $1.49 | 50.00% |
| RTX A6000 48GB | 4 | $0.915 | $0.458 | 50.00% |
| MI300X 192GB | 17 | $7.50 | $1.39 | 81.52% |
| A10 24GB | 105 | $4.52 | $0.845 | 81.52% |
| L4 24GB | 1 | $0.104 | $0.083 | 20.00% |
Live-data examples and quote conditions
H100 NVL 94GB 2×
US
USD (stablecoin) • platform fees may apply
$0.458 per GPU-hour
Minimum billable instance cost: USD $0.917/hr · Minimum billing duration unknown
Details & source
SKU: 122309 · Availability: Available · A · Official API
Reported GPU name: NVIDIA H100 NVL
Reported VRAM: 94 GB
Hourly amount in source scope: USD $0.917/hr · Instance
Source billing unit: Instance-hour
72/72
512 GB
94 GB
Listed whole-server daily price divided by 24 and GPU count. Platform service fees may apply; tax treatment unknown. Stablecoin-denominated USD price (USD-Blockchain).
API record pageH100 NVL 94GB 2×
US
USD (stablecoin) • platform fees may apply
$0.583 per GPU-hour
Minimum billable instance cost: USD $1.17/hr · Minimum billing duration unknown
Details & source
SKU: 115374 · Availability: Available · A · Official API
Reported GPU name: NVIDIA H100 NVL
Reported VRAM: 94 GB
Hourly amount in source scope: USD $1.17/hr · Instance
Source billing unit: Instance-hour
80/80
630 GB
94 GB
Listed whole-server daily price divided by 24 and GPU count. Platform service fees may apply; tax treatment unknown. Stablecoin-denominated USD price (USD-Blockchain).
API record pageH100 NVL 94GB 2×
US
USD (stablecoin) • platform fees may apply
$0.833 per GPU-hour
Minimum billable instance cost: USD $1.67/hr · Minimum billing duration unknown
Details & source
SKU: 122309 · Availability: Available · A · Official API
Reported GPU name: NVIDIA H100 NVL
Reported VRAM: 94 GB
Hourly amount in source scope: USD $1.67/hr · Instance
Source billing unit: Instance-hour
72/72
512 GB
94 GB
Listed whole-server daily price divided by 24 and GPU count. Platform service fees may apply; tax treatment unknown. Stablecoin-denominated USD price (USD-Blockchain).
API record pageH100 NVL 94GB 2×
US
USD (stablecoin) • platform fees may apply
$0.938 per GPU-hour
Minimum billable instance cost: USD $1.88/hr · Minimum billing duration unknown
Details & source
SKU: 115374 · Availability: Available · A · Official API
Reported GPU name: NVIDIA H100 NVL
Reported VRAM: 94 GB
Hourly amount in source scope: USD $1.88/hr · Instance
Source billing unit: Instance-hour
80/80
630 GB
94 GB
Listed whole-server daily price divided by 24 and GPU count. Platform service fees may apply; tax treatment unknown. Stablecoin-denominated USD price (USD-Blockchain).
API record page$1.42 per GPU-hour
Minimum billable instance cost: USD $1.42/hr · Minimum billing duration unknown
Details & source
SKU: Standard_NC40ads_H100_v5 · Availability: Unknown · A · Official API
Reported GPU name: H100 NVL 94GB
Hourly amount in source scope: USD $1.42/hr · Instance
Source billing unit: Instance-hour
$6.98 per GPU-hour
Minimum billable instance cost: USD $6.98/hr · Minimum billing duration unknown
Details & source
SKU: Standard_NC40ads_H100_v5 · Availability: Unknown · A · Official API
Reported GPU name: H100 NVL 94GB
Hourly amount in source scope: USD $6.98/hr · Instance
Source billing unit: Instance-hour
Official sources
Know the full job cost before you rent: GPU Budget Kit ($12)