What does a PetaFLOP cost right now?
Live GPU rental prices from 32 providers, normalized per GPU-hour and ranked by what you actually get: dense FLOPS, VRAM or memory bandwidth per dollar.
Find the sweet spot
Cheapest offer for each GPU: price vs. dense FP16 / BF16 throughput. Up and to the left is better; the line connects GPUs nothing else beats on both.
Best value by GPU
Ranked by $ / PFLOP·h at each GPU's cheapest matching offer. Click a column header to re-sort, or a GPU to compare every provider.
- 1
Instinct MI355X$1.03/PFLOP·h$2.59/GPU·h at Vultr 🇺🇸 · 2.52 PFBest - 2
RTX 6000 Ada$1.36/PFLOP·h$0.49/GPU·h at Vast.ai 🇺🇸 · 0.364 PFGood +32% - 3A100 SXM 40GB$1.41/PFLOP·h$0.44/GPU·h at Vast.ai 🇺🇸 · 0.312 PFGood +37%
- 4Instinct MI300X$1.42/PFLOP·h$1.85/GPU·h at Vultr 🇺🇸 · 1.31 PFGood +37%
- 5B200 SXM$1.42/PFLOP·h$3.20/GPU·h at Vultr 🇺🇸 · 2.25 PFGood +38%
- 6RTX A5000$1.44/PFLOP·h$0.16/GPU·h at RunPod 🇺🇸 · 0.111 PFGood +40%
- 7V100 SXM2 16GB$1.52/PFLOP·h$0.19/GPU·h at Verda (DataCrunch) 🇫🇮 · 0.125 PFGood +48%
- 8Instinct MI325X$1.53/PFLOP·h$2.00/GPU·h at Vultr 🇺🇸 · 1.31 PFGood +49%
- 9RTX 4070 SUPER$1.54/PFLOP·h$0.11/GPU·h at Vast.ai 🇺🇸 · 0.071 PFGood +49%
- 10V100 PCIe$1.70/PFLOP·h$0.19/GPU·h at RunPod 🇺🇸 · 0.112 PFGood +65%
- 11H100 SXM$1.83/PFLOP·h$1.82/GPU·h at Cudo Compute 🇬🇧 · 0.989 PFGood +78%
- 12RTX 4070 Ti$1.85/PFLOP·h$0.15/GPU·h at Vast.ai 🇺🇸 · 0.0802 PFGood +79%
- 13RTX 5000 Ada$1.88/PFLOP·h$0.49/GPU·h at RunPod 🇺🇸 · 0.261 PFGood +82%
- 14RTX A4000$1.96/PFLOP·h$0.15/GPU·h at Hyperstack 🇬🇧 · 0.0767 PFGood +90%
- 15A30$2.00/PFLOP·h$0.33/GPU·h at Seeweb 🇮🇹 · 0.165 PFGood +94%
| # | |||||||
|---|---|---|---|---|---|---|---|
| 1 | 288 GB | 8 TB/s | 2.52 PF |  $2.59/GPU·hat Vultr 🇺🇸 median $5.60 across 2 | $1.03/PFLOP·hBest | 2 → | |
| 2 | 48 GB | 1 TB/s | 0.364 PF |  $0.49/GPU·hat Vast.ai 🇺🇸 median $0.77 across 6 | $1.36/PFLOP·hGood +32% | 6 → | |
| 3 | 40 GB | 1.6 TB/s | 0.312 PF | $0.44/GPU·hat Vast.ai 🇺🇸 median $1.69 across 10 | $1.41/PFLOP·hGood +37% | 10 → | |
| 4 | 192 GB | 5.3 TB/s | 1.31 PF | $1.85/GPU·hat Vultr 🇺🇸 median $2.83 across 8 | $1.42/PFLOP·hGood +37% | 8 → | |
| 5 | 180 GB | 7.7 TB/s | 2.25 PF | $3.20/GPU·hat Vultr 🇺🇸 median $6.79 across 11 | $1.42/PFLOP·hGood +38% | 11 → | |
| 6 | 24 GB | 0.8 TB/s | 0.111 PF |  $0.16/GPU·hat RunPod 🇺🇸 median $0.27 across 5 | $1.44/PFLOP·hGood +40% | 5 → | |
| 7 | 16 GB | 0.9 TB/s | 0.125 PF |  $0.19/GPU·hat Verda (DataCrunch) 🇫🇮 median $0.79 across 7 | $1.52/PFLOP·hGood +48% | 7 → | |
| 8 | 256 GB | 6 TB/s | 1.31 PF | $2.00/GPU·hat Vultr 🇺🇸 median $2.90 across 2 | $1.53/PFLOP·hGood +49% | 2 → | |
| 9 | 12 GB | 0.5 TB/s | 0.071 PF | $0.11/GPU·hat Vast.ai 🇺🇸 | $1.54/PFLOP·hGood +49% | 1 → | |
| 10 | 16 GB | 0.9 TB/s | 0.112 PF | $0.19/GPU·hat RunPod 🇺🇸 median $2.34 across 3 | $1.70/PFLOP·hGood +65% | 3 → | |
| 11 | 80 GB | 3.4 TB/s | 0.989 PF |  $1.82/GPU·hat Cudo Compute 🇬🇧 median $3.58 across 22 | $1.83/PFLOP·hGood +78% | 22 → | |
| 12 | 12 GB | 0.5 TB/s | 0.0802 PF | $0.15/GPU·hat Vast.ai 🇺🇸 median $0.17 across 2 | $1.85/PFLOP·hGood +79% | 2 → | |
| 13 | 32 GB | 0.6 TB/s | 0.261 PF | $0.49/GPU·hat RunPod 🇺🇸 | $1.88/PFLOP·hGood +82% | 1 → | |
| 14 | 16 GB | 0.4 TB/s | 0.0767 PF |  $0.15/GPU·hat Hyperstack 🇬🇧 median $0.17 across 3 | $1.96/PFLOP·hGood +90% | 3 → | |
| 15 | 24 GB | 0.9 TB/s | 0.165 PF |  $0.33/GPU·hat Seeweb 🇮🇹 median $0.34 across 2 | $2.00/PFLOP·hGood +94% | 2 → |
Longer, greener bars mean more compute per dollar. The label shows how much more each GPU costs per unit than the best one. Click a GPU to compare every provider.
Data sources · 29/29 healthy
- Vast.ai229 offers
- RunPod67 offers
- Verda92 offers
- GPU.ai32 offers
- Shadeform187 offers
- Vultr15 offers
- AWS pricing113 offers
- Lambda28 offers
- Nebius7 offers
- Hyperstack30 offers
- DigitalOcean30 offers
- Together AI11 offers
- Scaleway15 offers
- OVHcloud37 offers
- Cudo Compute3 offers
- Hyperbolic7 offers
- Oracle Cloud15 offers
- Crusoe6 offers
- Hot Aisle4 offers
- TensorWave6 offers
- Cirrascale48 offers
- Seeweb40 offers
- Azure Retail API74 offers
- Massed Compute74 offers
- Latitude9 offers
- SkyPilot: aws143 offers
- SkyPilot: gcp76 offers
- SkyPilot: nebius52 offers
- SkyPilot: seeweb27 offers
Browse by provider
See which GPUs each cloud rents out and how its prices compare.
- RunPod 🇺🇸42 GPUs · $0.12
- Vast.ai 🇺🇸21 GPUs · $0.10
GPU.ai 🇺🇸20 GPUs · $0.27
AWS 🇺🇸16 GPUs · $0.53
Massed Compute 🇺🇸16 GPUs · $0.35
Oracle Cloud 🇺🇸15 GPUs · $1.28
Google Cloud 🇺🇸12 GPUs · $0.54
OVHcloud 🇫🇷12 GPUs · $0.41- Verda (DataCrunch) 🇫🇮12 GPUs · $0.19
Azure 🇺🇸11 GPUs · $0.53
Cirrascale 🇺🇸11 GPUs · $0.27- Hyperstack 🇬🇧11 GPUs · $0.15
Lambda 🇺🇸11 GPUs · $0.69
DigitalOcean 🇺🇸10 GPUs · $0.76- Seeweb 🇮🇹10 GPUs · $0.33
- Vultr 🇺🇸10 GPUs · $0.47
Paperspace 🇺🇸7 GPUs · $0.80
Crusoe 🇺🇸6 GPUs · $1.50
Nebius 🇳🇱6 GPUs · $1.55
Scaleway 🇫🇷6 GPUs · $0.90
Denvr 🇨🇦4 GPUs · $1.10
Latitude.sh 🇺🇸4 GPUs · $0.74
Together AI 🇺🇸4 GPUs · $3.99- Cudo Compute 🇬🇧3 GPUs · $0.88
IMWT 3 GPUs · $0.49
TensorWave 🇺🇸3 GPUs · $1.71- Horizon 2 GPUs · $1.95
Hyperbolic 🇺🇸2 GPUs · $3.99- Phyntec 2 GPUs · $0.40
- Amaya 1 GPU · $2.25
Hot Aisle 🇺🇸1 GPU · $2.99
Voltage Park 🇺🇸1 GPU · $1.99
Browse GPU prices
Every GPU we track right now, with its lowest on-demand price per hour.
NVIDIA61
- H100 SXM$1.82
- RTX PRO 6000 Blackwell$1.25
- A100 SXM 80GB$1.05
- H200 SXM$2.25
- L40S$0.74
- RTX A6000$0.33
- A100 SXM 40GB$0.44
- B200 SXM$3.20
- A100 PCIe 80GB$1.13
- B300 SXM$6.60
- L4$0.43
- V100 SXM2 16GB$0.19
- H100 PCIe$1.66
- RTX 4090$0.34
- RTX 6000 Ada$0.49
- RTX 5090$0.43
- T4$0.53
- RTX A5000$0.16
- H100 NVL$2.59
- A10$1.00
- L40$0.69
- RTX A4000$0.15
- H200 NVL$3.29
- V100 PCIe$0.19
- A30$0.33
- RTX 5080$0.25
- RTX 4060 Ti$0.10
- RTX 5060 Ti$0.12
- GB300$9.72
- A100 PCIe 40GB$1.99
- RTX 3090$0.22
- A10G$1.01
- RTX PRO 4500 Blackwell$0.72
- RTX 5070 Ti$0.20
- Quadro RTX 6000$0.33
- GB200$16.0
- RTX 4080 SUPER$0.28
- RTX 4070$0.14
- A16$0.47
- V100S PCIe$0.88
- RTX 4000 Ada$0.20
- RTX 4070 Ti$0.15
- RTX 2080 Ti$0.11
- RTX 4070 SUPER$0.11
- GH200$1.99
- A40$0.49
- V100 SXM2 32GB$3.90
- P100$1.28
- RTX A4500$0.19
- RTX PRO 5000 Blackwell$0.82
- RTX 5000 Ada$0.49
- RTX PRO 4000 Blackwell$0.57
- RTX 4000 SFF Ada$0.18
- RTX 2000 Ada$0.24
- RTX A2000$0.12
- Quadro RTX 8000$0.27
- RTX 4080$0.27
- RTX 3090 Ti$0.27
- RTX 3080 Ti$0.18
- RTX 3080$0.17
- RTX 3070$0.13
AMD5
Google TPUAWS Trainium3
Intel1
Popular comparisons
- H100 SXM vs H200 SXM
- H100 SXM vs A100 SXM 80GB
- H100 SXM vs B200 SXM
- H200 SXM vs B200 SXM
- B200 SXM vs B300 SXM
- GB200 vs B200 SXM
- H100 SXM vs Instinct MI300X
- H200 SXM vs Instinct MI300X
- Instinct MI300X vs Instinct MI325X
- B200 SXM vs Instinct MI355X
- H100 PCIe vs H100 SXM
- H100 NVL vs H100 SXM
- GH200 vs H100 SXM
- A100 SXM 80GB vs A100 PCIe 80GB
- A100 SXM 40GB vs A100 SXM 80GB
- L40S vs A100 SXM 80GB
- All comparisons →
GPU pricing questions
What is $ per PFLOP-hour?
It is the rental price of one GPU for one hour divided by its peak compute in PetaFLOPS at the precision you choose. It shows how much compute each dollar buys, so a cheap slow GPU and an expensive fast one can be compared directly. Lower is better.
Which GPU gives the most compute per dollar right now?
At dense FP16/BF16, the Instinct MI355X currently leads at $1.03 per PFLOP-hour ($2.59 per GPU-hour at Vultr). Rankings change with prices; the table above updates every 2 hours.
How much does an H100 cost per hour?
On-demand H100 SXM prices currently start at $1.82 per GPU-hour, with a median of $3.58 across 22 providers. Spot capacity starts at $1.26.
Why dense FLOPS instead of the numbers on NVIDIA's spec sheets?
Headline tensor numbers often assume 2:4 structured sparsity, which doubles them but is rarely used in practice. PetaPrice always uses dense throughput so GPUs and vendors are compared fairly.
Where does the price data come from?
Directly from the providers wherever they publish prices (public APIs, official price lists and pricing pages), with Shadeform's multi-cloud API and SkyPilot's catalogs filling gaps: 891 offers from 32 providers, refreshed about every 2 hours. Prices are indicative list prices; confirm with the provider before booking.
What's the difference between on-demand, spot and reserved?
On-demand is billed by the hour with no commitment. Spot (interruptible) capacity is cheaper but can be reclaimed at any time. Reserved prices require a commitment, typically 1 to 3 years. PetaPrice never mixes them; the default view is on-demand.