GPU
On-demand and marketplace prices per GPU-hour, normalized to USD across every cloud we track.
H100 SXM
$2.89Median
Cheapest $1.00
B200
$7.99Median
Cheapest $2.69
A100 80GB
$1.79Median
Cheapest $0.665
GeForce RTX 4090
$0.691Median
Cheapest $0.280
Datacenter
| GPU | Cheapest | Median | VRAM | Architecture | 25th pct | Providers | Offers | Price history |
|---|---|---|---|---|---|---|---|---|
| GB300 (Blackwell Ultra)NVIDIA | $20.00 | $20.00 | 288 GB | Blackwell Ultra · 2025 | $20.00 | 1 | 1 | |
| B300 (Blackwell Ultra)NVIDIA | $2.93 | $9.50 | 288 GB | Blackwell Ultra · 2025 | $7.47 | 12 | 13 | |
| GB200 (Grace Blackwell)NVIDIA | $8.00 | $9.99 | 186 GB | Blackwell · 2024 | $9.00 | 3 | 3 | |
| B200NVIDIA | $2.69 | $7.99 | 180 GB | Blackwell · 2024 | $6.25 | 21 | 33 | |
| H200NVIDIA | $1.00 | $4.10 | 141 GB | Hopper · 2024 | $2.60 | 33 | 60 | |
| H200 NVLNVIDIA | $5.04 | $5.04 | 141 GB | Hopper · 2024 | $5.04 | 1 | 1 | |
| GH200 (Grace Hopper)NVIDIA | $1.99 | $2.14 | 96 GB | Hopper · 2023 | $1.99 | 3 | 4 | |
| H100 SXMNVIDIA | $1.00 | $2.89 | 80 GB | Hopper · 2022 | $1.98 | 44 | 108 | |
| H100 NVLNVIDIA | $1.98 | $3.19 | 94 GB | Hopper · 2023 | $3.19 | 5 | 14 | |
| H100 PCIeNVIDIA | $1.68 | $2.18 | 80 GB | Hopper · 2022 | $1.90 | 8 | 12 | |
| L40SNVIDIA | $0.600 | $1.22 | 48 GB | Ada Lovelace · 2023 | $1.00 | 23 | 75 | |
| L40NVIDIA | $0.790 | $0.995 | 48 GB | Ada Lovelace · 2022 | $0.820 | 6 | 28 | |
| L4NVIDIA | $0.380 | $0.633 | 24 GB | Ada Lovelace · 2023 | $0.590 | 13 | 34 | |
| A100 80GBNVIDIA | $0.665 | $1.79 | 80 GB | Ampere · 2020 | $1.40 | 42 | 109 | |
| A100 40GBNVIDIA | $0.721 | $1.99 | 40 GB | Ampere · 2020 | $1.40 | 9 | 17 | |
| A40NVIDIA | $0.590 | $0.590 | 48 GB | Ampere · 2020 | $0.590 | 2 | 11 | |
| A30NVIDIA | $0.290 | $0.832 | 24 GB | Ampere · 2021 | $0.561 | 2 | 2 | |
| A10NVIDIA | $0.435 | $2.00 | 24 GB | Ampere · 2021 | $1.29 | 4 | 9 | |
| A10GNVIDIA | $1.01 | $1.83 | 24 GB | Ampere · 2021 | $1.37 | 1 | 8 | |
| T4NVIDIA | $0.110 | $0.978 | 16 GB | Turing · 2018 | $0.605 | 7 | 22 | |
| V100 32GBNVIDIA | $0.175 | $2.19 | 32 GB | Volta · 2018 | $0.883 | 7 | 21 | |
| V100 16GBNVIDIA | $0.138 | $1.97 | 16 GB | Volta · 2017 | $0.660 | 11 | 27 | |
| P100NVIDIA | $1.24 | $1.28 | 16 GB | Pascal · 2016 | $1.26 | 2 | 3 | |
| P40NVIDIA | $0.150 | $0.150 | 24 GB | Pascal · 2016 | $0.150 | 1 | 1 | |
| P4NVIDIA | $0.0600 | $0.330 | 8 GB | Pascal · 2016 | $0.195 | 2 | 2 | |
| Tesla M60NVIDIA | $1.14 | $1.14 | 16 GB | Maxwell · 2015 | $1.14 | 1 | 3 | |
| Instinct MI355XAMD | $2.59 | $2.77 | 288 GB | CDNA 4 · 2025 | $2.68 | 2 | 2 | |
| Instinct MI350XAMD | $5.49 | $5.49 | 288 GB | CDNA 4 · 2025 | $5.49 | 1 | 6 | |
| Instinct MI325XAMD | $2.25 | $3.03 | 256 GB | CDNA 3 · 2024 | $2.64 | 2 | 2 | |
| Instinct MI300XAMD | $2.39 | $2.39 | 192 GB | CDNA 3 · 2023 | $2.39 | 6 | 14 |
Workstation
| GPU | Cheapest | Median | VRAM | Architecture | 25th pct | Providers | Offers | Price history |
|---|---|---|---|---|---|---|---|---|
| RTX PRO 6000 BlackwellNVIDIA | $0.930 | $2.45 | 96 GB | Blackwell · 2025 | $1.48 | 18 | 48 | |
| RTX PRO 5000 BlackwellNVIDIA | $0.960 | $0.960 | 48 GB | Blackwell · 2025 | $0.960 | 1 | 1 | |
| RTX PRO 4500 BlackwellNVIDIA | $0.720 | $3.04 | 32 GB | Blackwell · 2025 | $0.720 | 2 | 9 | |
| RTX 6000 AdaNVIDIA | $0.520 | $0.990 | 48 GB | Ada Lovelace · 2022 | $0.970 | 7 | 17 | |
| RTX 5880 AdaNVIDIA | $0.350 | $0.350 | 48 GB | Ada Lovelace · 2024 | $0.350 | 1 | 1 | |
| RTX 4000 AdaNVIDIA | $0.280 | $0.520 | 20 GB | Ada Lovelace · 2023 | $0.280 | 4 | 20 | |
| RTX 2000 AdaNVIDIA | $0.240 | $0.240 | 16 GB | Ada Lovelace · 2024 | $0.240 | 1 | 8 | |
| RTX A6000NVIDIA | $0.350 | $0.570 | 48 GB | Ampere · 2020 | $0.500 | 9 | 45 | |
| RTX A5000NVIDIA | $0.200 | $0.270 | 24 GB | Ampere · 2021 | $0.270 | 4 | 16 | |
| RTX A4500NVIDIA | $0.250 | $0.250 | 20 GB | Ampere · 2021 | $0.250 | 1 | 4 | |
| RTX A4000NVIDIA | $0.150 | $0.250 | 16 GB | Ampere · 2021 | $0.150 | 4 | 26 | |
| Quadro RTX 6000NVIDIA | $0.290 | $1.50 | 24 GB | Turing · 2018 | $1.10 | 3 | 7 |
Consumer
| GPU | Cheapest | Median | VRAM | Architecture | 25th pct | Providers | Offers | Price history |
|---|---|---|---|---|---|---|---|---|
| GeForce RTX 5090NVIDIA | $0.400 | $0.650 | 32 GB | Blackwell · 2025 | $0.650 | 11 | 38 | |
| GeForce RTX 5070NVIDIA | $0.0925 | $0.0925 | 12 GB | Blackwell · 2025 | $0.0925 | 1 | 1 | |
| GeForce RTX 4090NVIDIA | $0.280 | $0.691 | 24 GB | Ada Lovelace · 2022 | $0.378 | 11 | 23 | |
| GeForce RTX 4080 / SuperNVIDIA | $0.150 | $0.150 | 16 GB | Ada Lovelace · 2022 | $0.150 | 1 | 1 | |
| GeForce RTX 4070 / SuperNVIDIA | $0.193 | $0.193 | 12 GB | Ada Lovelace · 2023 | $0.193 | 1 | 1 | |
| GeForce RTX 4060 TiNVIDIA | $0.0725 | $0.0725 | 16 GB | Ada Lovelace · 2023 | $0.0725 | 1 | 1 | |
| GeForce RTX 3090 / TiNVIDIA | $0.0900 | $0.500 | 24 GB | Ampere · 2020 | $0.229 | 5 | 16 | |
| GeForce RTX 3070 / TiNVIDIA | $0.980 | $0.980 | 8 GB | Ampere · 2020 | $0.980 | 1 | 1 | |
| GeForce RTX 3060 / TiNVIDIA | $0.0525 | $0.140 | 12 GB | Ampere · 2021 | $0.0963 | 2 | 3 |
Accelerators
| GPU | Cheapest | Median | VRAM | Architecture | 25th pct | Providers | Offers | Price history |
|---|---|---|---|---|---|---|---|---|
| TPU v6e (Trillium)Google | $2.70 | $2.70 | 32 GB | TPU · 2024 | $2.70 | 1 | 8 | |
| TPU v5pGoogle | $4.20 | $4.20 | 95 GB | TPU · 2023 | $4.20 | 1 | 94 | |
| TPU v5eGoogle | $1.20 | $1.20 | 16 GB | TPU · 2023 | $1.20 | 1 | 8 | |
| TPU v4Google | $3.22 | $3.22 | 32 GB | TPU · 2021 | $3.22 | 1 | 52 | |
| TrainiumAWS | $1.34 | $1.34 | 32 GB | Trainium · 2022 | $1.34 | 1 | 3 | |
| Inferentia2AWS | $0.758 | $1.08 | 32 GB | Inferentia · 2023 | $1.00 | 1 | 4 | |
| InferentiaAWS | $0.228 | $0.295 | 8 GB | Inferentia · 2019 | $0.278 | 1 | 4 |
Prices are per GPU-hour in USD-equivalent at today's exchange rates; multi-GPU nodes are divided by GPU count. Fractional (vGPU) plans are excluded from the index.