Scrutica
CHEAPEST ON-DEMAND
CHEAPEST SPOT
ON-PREM (3YR AMORT.)
CHEAPEST ON-DEMAND
CHEAPEST SPOT
ON-PREM (3YR AMORT.)
| Provider | Instance | Region | $/hr | $/SCU | Source |
|---|---|---|---|---|---|
| AWS | p5.48xlarge | US East (N. Virginia) | ↓ $43.28$55.04 | $166.87 | |
| AWS | p5.48xlarge | US West (Oregon) | ↓ $43.28$55.04 | $166.87 | |
| AWS | p5.48xlarge | Asia Pacific (Tokyo) | $123.19 | $373.51 | |
| Azure | Standard_ND96isr_H100_v5 | East US | ↑ $4.42$102.74 | $311.48 | |
| Azure | Standard_ND96isr_H100_v5 | centralus | $120.93 | $366.65 | |
| Azure | Standard_ND96isr_H100_v5 | West US 2 | ↑ $4.42$102.74 | $311.48 | |
| Azure | Standard_ND96isr_H100_v5 | canadacentral | $122.40 | $371.10 | |
| Azure | Standard_ND96isr_H100_v5 | usgovarizona | $122.90 | $372.61 | |
| Azure | Standard_ND96isr_H100_v5 | japanwest | $156.81 | $475.43 | |
| Azure | Standard_ND96isr_H100_v5 | westus3 | $102.74 | $311.48 | |
| Azure | Standard_ND96isr_H100_v5 | australiaeast | $142.60 | $432.34 | |
| Azure | Standard_ND96isr_H100_v5 | East US 2 | ↑ $4.42$102.74 | $311.48 | |
| Azure | Standard_ND96isr_H100_v5 | southafricanorth | $173.09 | $524.77 | |
| Azure | Standard_ND96isr_H100_v5 | West Europe | $127.82 | $387.52 | |
| Azure | Standard_ND96isr_H100_v5 | switzerlandnorth | $145.01 | $439.66 | |
| Azure | Standard_ND96isr_H100_v5 | spaincentral | $127.82 | $387.52 | |
| Azure | Standard_ND96isr_H100_v5 | japaneast | $142.56 | $432.23 | |
| Azure | Standard_ND96isr_H100_v5 | uksouth | $122.90 | $372.61 | |
| Azure | Standard_ND96isr_H100_v5 | swedencentral | $127.82 | $387.52 | |
| Azure | Standard_ND96isr_H100_v5 | westus | $127.82 | $387.52 | |
| Azure | Standard_ND96isr_H100_v5 | polandcentral | $127.82 | $387.52 | |
| Azure | Standard_ND96isr_H100_v5 | westcentralus | $117.98 | $357.71 | |
| Azure | Standard_ND96isr_H100_v5 | southafricawest | $186.61 | $565.76 | |
| Azure | Standard_ND96isr_H100_v5 | koreacentral | $137.15 | $415.81 | |
| Azure | Standard_ND96isr_H100_v5 | northcentralus | $122.40 | $371.10 | |
| Azure | Standard_ND96isr_H100_v5 | southcentralus | $117.97 | $357.66 | |
| Azure | Standard_ND96isr_H100_v5 | uaenorth | $140.60 | $426.27 | |
| CoreWeave | HGX-H100_SXM5_80GB | US East (N. Virginia) | $49.24 | $149.29 | |
| GCP | a3-highgpu-8g | US Central (Iowa) | ↓ $9.66$88.49 | $268.29 | |
| Lambda | H100_SXM5_80GB-8gpu | US (Texas) | ↑ $14.40$34.32 | $104.05 | |
| Oracle Cloud | OCI-H100_SXM5_80GB | global | $80.00 | $242.55 |
Unit. $/SCU values use BF16 densethroughput (989.5 TFLOP/s per H100 SXM, 312 per A100, 2,250 per B200). NVIDIA marketing quotes sparsity-inclusive rates (1,979 / 624 / 4,500); a reader deriving $/SCU from provider $/hour against those sparsity-inclusive numbers will read values ~2× lower than the index here. See the Methodology section below for the unit choice.
Coverage. B200 (Blackwell) cloud instance pricing not yet included for all providers. MI300X instance pricing (Azure ND MI300X v5) not yet included. Cloud coverage strongest for H100, H200, A100; Lambda re-scraped 2026-04-01 against lambda.ai after lambdalabs.com domain migration.
Methodology
Scrutica Compute Unit (SCU) = 1 petaFLOP-day of BF16 training compute = 8.64 × 10¹&sup9; FLOP. At 40% MFU against BF16 dense throughput, one H100 SXM produces ~0.40 SCU per day (989.5 TFLOP/s × 0.40 = 395.8 TFLOP/s = 0.396 PFLOP/s).
Scrutica Compute Unit (SCU) = one petaflop-day of training compute (8.64 × 10¹&sup9; FLOP, i.e., a petaflop sustained for 24 hours). The unit normalises compute cost across providers so the rate cards compare like-for-like.
Cloud pricing: $/SCU = hourly instance price × 24 / (num_gpus × bf16_tflops / 1000). BF16 TFLOP/s dense (no 2:4 structured sparsity) from Epoch AI ML Hardware dataset: H100 = 989.5, A100 = 312, B200 = 2,250. NVIDIA marketing sheets quote sparsity-inclusive figures (1,979 / 624 / 4,500); Scrutica uses dense for consistency with the FLOP Capacity Engine and real-world training throughput.
On-premise: $/SCU = (hardware amortization + lifetime power cost) / (daily PFLOP output × useful life in days). Includes GPU, networking ($5K/GPU), and facility share ($3K/GPU) amortized over useful life.