Cloud GPU Pricing — Cost Intelligence for H100, A100 & B200
GPU pricing stopped being a stable lookup table. Supply constraints and new hardware releases mean prices move meaningfully quarter to quarter. We track them independently — across 10 providers, every major GPU, and all pricing tiers — so your team has a neutral view before committing budget.
Cloud GPU Pricing — Jul 2026
Compare Cloud GPU Costs
Independent cost intelligence across major GPU cloud providers. Prices are regularly tracked.
$2.00
Cheapest H100 /GPU/hr
Nebius · 71% below AWS
$0.12
Best Spot Deal
Hyperstack A4000 16GB
10
Providers Tracked
Major cloud platforms
10+
GPU Types
B300, H200, H100, MI300X, TPU…
Tools
Reserved vs On-Demand break-even
See how many months until your reserved commitment pays off versus on-demand pricing.
Commitment Analyzer
Reserved vs on-demand break-even
Model the full-term cost of a GPU commitment versus on-demand at your expected utilization, across three market-movement scenarios.
Total cost over 12 months
On-demand total
$90,753.60
Reserved total
$105,120.00
On-demand saves $14,366.40
On-demand total
$113,442.00
Reserved total
$105,120.00
Reserved saves $8,322.00
On-demand total
$136,130.40
Reserved total
$105,120.00
Reserved saves $31,010.40
Cost vs utilization (flat scenario)
Monthly cost at each utilization level
Verdict
At 70% utilization, a 12-month commitment saves $8,322.00 — but if market prices fall 20%, on-demand wins by $14,366.40.
Estimates only. Reserved rates assume locked-in pricing for full commitment term. Confirm current rates on the provider's pricing page before procurement.
| Provider | GPU | VRAM | On-Demand $/GPU-hr ↑ | Spot $/GPU-hr | 1Y Reserved $/GPU-hr | vs AWS | Hist | Watch | Deploy |
|---|---|---|---|---|---|---|---|---|---|
| Hyperstack | A4000 16GB | 16 GB | $0.15 | $0.1220% off | $0.11 | — | Deploy → | ||
| Hyperstack | A6000 48GB | 48 GB | $0.50 | $0.4020% off | $0.35 | — | Deploy → | ||
| GCP | TPU v5e | 128 GB HBM | $0.60 | — | $0.38 | — | Deploy → | ||
| Hyperstack | L40 48GB | 48 GB | $1.00 | $0.8020% off | $0.70 | — | Deploy → | ||
| Hyperstack | A100 PCIe 80GB | 80 GB | $1.40 | $1.0823% off | $0.95 | — | Deploy → | ||
| Lambda | L40S 48GB | 384 GB (8×48) | $1.50 | — | — | −25% | Deploy → | ||
| CoreWeave | L40S 48GB | 48 GB | $1.50 | — | — | −25% | Deploy → | ||
| GCP | TPU v6e | 256 GB HBM | $1.56 | — | $0.98 | — | Deploy → | ||
| GCP | TPU v4 | 128 GB HBM | $1.60 | — | $0.96 | — | Deploy → | ||
| Hyperstack | A100 SXM 80GB | 80 GB | $1.60 | $1.2820% off | $1.36 | — | Deploy → | ||
| RunPod | A100 80GB | 640 GB (8×80) | $1.64 | $0.8150% off | — | — | Deploy → | ||
| Lambda | A100 80GB | 640 GB (8×80) | $1.79 | — | — | — | Deploy → | ||
| Hyperstack | RTX Pro 6000 SE 96GB | 96 GB | $1.85 | $1.4422% off | $1.30 | — | Deploy → | ||
| AWS | L40S 48GB | 384 GB (8×48) | $2.00 | $0.6966% off | $1.25 | baseline | Deploy → | ||
| Nebius | H100 80GB | 640 GB (8×80) | $2.00↓12% vs Jun '25 | — | $1.32 | −71% | Deploy → | ||
| CoreWeave | A100 80GB | 80 GB | $2.21 | — | — | — | Deploy → | ||
| Nebius | H200 141GB | 1128 GB (8×141) | $2.30 | — | $1.65 | −82% | Deploy → | ||
| Crusoe | H100 80GB | 640 GB (8×80) | $2.50 | $1.5040% off | — | −64% | Deploy → | ||
| Hyperstack | H100 PCIe 80GB | 80 GB | $2.50 | $1.5239% off | $1.75 | −64% | Deploy → | ||
| Hyperstack | H100 NVLink 80GB | 80 GB | $2.60 | $1.5640% off | $1.82 | −62% | Deploy → | ||
| GCP | TPU v5p | 384 GB HBM | $2.64 | — | $1.66 | — | Deploy → | ||
| RunPod | H100 80GB | 640 GB (8×80) | $2.69↓23% vs Jun '25 | $1.2554% off | — | −61% | Deploy → | ||
| Lambda | MI300X 192GB | 1536 GB (8×192) | $3.06 | — | — | — | Deploy → | ||
| Lambda | H100 80GB | 640 GB (8×80) | $3.14↓8% vs Jun '25 | — | $1.89 | −54% | Deploy → | ||
| Hyperstack | H100 SXM 80GB | 80 GB | $3.20 | $1.9240% off | $2.72 | −53% | Deploy → | ||
| Crusoe | MI300X 192GB | 1536 GB (8×192) | $3.45 | — | — | — | Deploy → | ||
| Together AI | H100 80GB | 640 GB (8×80) | $3.49 | — | $2.25 | −49% | Deploy → | ||
| GCP | TPU v7 | 384 GB HBM3e | $3.56 | — | $2.31 | — | Deploy → | ||
| Lambda | H200 141GB | 1128 GB (8×141) | $3.75↓15% vs Jun '25 | — | — | −71% | Deploy → | ||
| Hyperstack | H200 141GB | 141 GB | $3.99 | $2.8030% off | $2.79 | −69% | Deploy → | ||
| Azure | A100 80GB | 640 GB (8×80) | $4.10 | $1.2370% off | $2.55 | — | Deploy → | ||
| CoreWeave | MI300X 192GB | 192 GB | $4.10 | — | — | — | Deploy → | ||
| CoreWeave | H100 80GB | 80 GB | $4.76 | — | — | −31% | Deploy → | ||
| GCP | A100 80GB | 640 GB (8×80) | $5.03 | $1.5170% off | $3.17 | — | Deploy → | ||
| CoreWeave | H200 141GB | 141 GB | $5.20 | — | — | −60% | Deploy → | ||
| RunPod | H200 141GB | 1128 GB (8×141) | $5.50 | $2.7550% off | — | −58% | Deploy → | ||
| Hyperstack | B200 192GB | 192 GB | $6.00 | $4.8020% off | $5.10 | −58% | Deploy → | ||
| Lambda | B300 288GB | 2304 GB (8×288) | $6.50 | — | — | −65% | Deploy → | ||
| CoreWeave | B200 192GB | 192 GB | $6.50 | — | — | −55% | Deploy → | ||
| AWS | H100 80GB | 640 GB (8×80) | $6.88↓44% vs Jun '25 | $2.4864% off | $4.81 | baseline | Deploy → | ||
| CoreWeave | B300 288GB | 288 GB | $8.50 | — | — | −54% | Deploy → | ||
| Azure | MI250X 128GB | 512 GB (4×128) | $9.00 | $2.7070% off | $5.50 | — | Deploy → | ||
| GCP | H100 80GB | 640 GB (8×80) | $10.98↑11% vs Jun '25 | $3.3070% off | $6.90 | +60% | Deploy → | ||
| Azure | MI300A 128GB | 512 GB (4×128) | $11.25 | $3.3870% off | $7.00 | — | Deploy → | ||
| Azure | MI300X 192GB | 1536 GB (8×192) | $11.56 | $3.4770% off | $7.26 | — | Deploy → | ||
| Azure | MI325X 288GB | 2304 GB (8×288) | $12.25 | $3.6770% off | $7.75 | — | Deploy → | ||
| Azure | H100 80GB | 640 GB (8×80) | $12.29↓5% vs Jun '25 | $3.6970% off | $7.62 | +79% | Deploy → | ||
| AWS | H200 141GB | 1128 GB (8×141) | $13.00↓6% vs Jun '25 | $4.7563% off | $8.50 | baseline | Deploy → | ||
| GCP | H200 141GB | 1128 GB (8×141) | $13.13 | $3.8870% off | $8.50 | +1% | Deploy → | ||
| Azure | MI355X 288GB | 2304 GB (8×288) | $13.50 | $4.0570% off | $8.75 | — | Deploy → | ||
| GCP | B200 192GB | 1536 GB (8×192) | $13.75 | $4.1370% off | $9.00 | −4% | Deploy → | ||
| Azure | B200 192GB | 1536 GB (8×192) | $14.00 | $4.2070% off | $9.13 | −3% | Deploy → | ||
| AWS | B200 192GB | 1536 GB (8×192) | $14.38 | $5.2563% off | $9.38 | baseline | Deploy → | ||
| GCP | B300 288GB | 2304 GB (8×288) | $17.75 | $5.3370% off | $11.50 | −4% | Deploy → | ||
| Azure | B300 288GB | 2304 GB (8×288) | $18.13 | $5.4470% off | $11.75 | −2% | Deploy → | ||
| AWS | B300 288GB | 2304 GB (8×288) | $18.50 | $6.5065% off | $12.00 | baseline | Deploy → |
Use Spot for Training
Save 60–70% on training runs with checkpointing. GCP Spot offers up to 70% discount on A100/H100 instances.
Best for: Fault-tolerant training with checkpoints
Reserved for Inference
1-year commitments save 35–40% for always-on inference endpoints. Azure and AWS offer the deepest reserved discounts.
Best for: Production inference workloads
Lambda/Nebius for Value
GPU-specialist clouds offer 2–4× lower per-GPU pricing than hyperscalers. Nebius leads on committed pricing; Lambda leads on on-demand.
Best for: Pure GPU compute without cloud services
Provider Intelligence
Cost · Reliability · Ecosystem rated 1–5Most complete cloud. Premium pricing. Best SLA.
Best TPU access. JAX-native. Strong SLA.
Best AMD GPU selection. Enterprise-ready.
Best $/GPU for pure compute. No ecosystem overhead.
Hyperscaler-grade infra at specialist pricing.
Cheapest H100 on-demand. Community cloud model.
Best committed pricing. EU data centers.
Inference API + raw GPU in one platform.
Zero egress fees. OpenAI Stargate infrastructure partner.
Zero-egress European cloud with strong price-performance on H-series.
Prices are approximate and vary by region and availability. Pricing reflects Jul 2026 estimates — always verify with provider pricing pages before procurement.