Cheapest GPU memory per hour
As of — Every card, cheapest GB first: from $0.005 per GB of memory per hour on the NVIDIA V100 32GB (32 GB at $0.16 on Geodd), 54 cards. Rates are each card’s cheapest live on-demand rate card across the providers we track, read hourly, every one linked to the page it was read from.
When the job is limited by memory - a model that has to fit, a batch that has to be large - the figure to compare is the price per GB of GPU memory per hour: the cheapest on-demand rate for a card divided by its memory. Speed is not in it: a cheaper GB on an older card can take longer to use.
Every card, cheapest GB first
Each card’s cheapest live on-demand rate divided by its memory.
| Card, memory | Cheapest on-demand | Provider, read | Median rate card | $ per GB-hour | Deploy |
|---|---|---|---|---|---|
| NVIDIA V100 32GB · 32 GB | $0.16 |  Geoddcaptured | $1.15 11 providers | $0.005 | Deploy ↗ |
| NVIDIA GB10 (DGX Spark) · 128 GB | $0.79 | GPUwerk captured | $0.82 2 providers | $0.006 | Deploy ↗ |
| NVIDIA RTX PRO 6000 Blackwell · 96 GB | $0.65 from |  Hostingercaptured | $2.20 42 providers | $0.007 | Deploy ↗ |
| NVIDIA RTX A6000 · 48 GB | $0.35 |  Thunder Computecaptured | $0.59 21 providers | $0.007 | Deploy ↗ |
| NVIDIA GeForce RTX 3090 · 24 GB | $0.18 |  OpenRelaycaptured | $0.42 10 providers | $0.007 | Deploy ↗ |
| NVIDIA A40 · 48 GB | $0.39 |  CloudGPUcaptured | $0.64 8 providers | $0.008 | Deploy ↗ |
| AMD Instinct MI325X · 256 GB | $2.25 |  Bentauscaptured | $2.25 3 providers | $0.009 | Deploy ↗ |
| AMD Instinct MI300X · 192 GB | $1.71 from |  TensorWavecaptured | $2.65 7 providers | $0.009 | Deploy ↗ |
| NVIDIA A30 · 24 GB | $0.22 | CloudGPU captured | $0.41 9 providers | $0.009 | Deploy ↗ |
| NVIDIA RTX A4000 · 16 GB | $0.15 |  Hyperstackcaptured | $0.25 11 providers | $0.009 | Deploy ↗ |
| AMD Instinct MI355X · 288 GB | $2.95 from | TensorWave captured | $6.14 4 providers | $0.010 | Deploy ↗ |
| NVIDIA RTX A5000 · 24 GB | $0.27 from |  GPU.aicaptured | $0.58 10 providers | $0.011 | Deploy ↗ |
| Intel Gaudi 2 · 96 GB | $1.10 |  Denvr Dataworkscaptured | $1.16 2 providers | $0.011 | Deploy ↗ |
| NVIDIA A100 80GB · 80 GB | $0.92 |  Yotta Labscaptured | $1.99 63 providers | $0.011 | Deploy ↗ |
| NVIDIA L4 · 24 GB | $0.28 GPU only |  Hikubecaptured | $0.80 23 providers | $0.012 | Deploy ↗ |
| NVIDIA A2 · 16 GB | $0.20 | Infomaniak captured | $0.23 2 providers | $0.012 | Deploy ↗ |
| NVIDIA Quadro RTX 6000 · 24 GB | $0.31 GPU only |  centroncaptured | $0.51 4 providers | $0.013 | Deploy ↗ |
| NVIDIA L40S · 48 GB | $0.63 |  CloudRiftcaptured | $1.42 40 providers | $0.013 | Deploy ↗ |
| NVIDIA RTX 6000 Ada · 48 GB | $0.65 from | GPU.ai captured | $1.03 12 providers | $0.014 | Deploy ↗ |
| AMD Instinct MI350X · 288 GB | $4.00 | CloudRift captured | $5.08 2 providers | $0.014 | Deploy ↗ |
34 more cards qualify, dearer: NVIDIA A10 $0.34, NVIDIA RTX 4000 Ada $0.28, NVIDIA RTX 4090 $0.35, NVIDIA T4 $0.24, NVIDIA RTX 2000 Ada $0.24, NVIDIA V100 $0.24, NVIDIA RTX 5090 $0.50, NVIDIA L40 $0.78, NVIDIA H200 $2.29, NVIDIA RTX PRO 5000 Blackwell $0.80, NVIDIA A100 40GB $0.69, NVIDIA GeForce RTX 4070 Ti SUPER $0.29, NVIDIA RTX PRO 6000 Blackwell Max-Q $1.80, NVIDIA GeForce RTX 4080 $0.31, NVIDIA GeForce RTX 3090 Ti $0.46, NVIDIA B300 $5.89, NVIDIA GeForce GTX 1080 Ti $0.23, NVIDIA GeForce RTX 2080 Ti $0.23, NVIDIA H100 80GB (form factor not stated) $1.72, NVIDIA RTX 4000 SFF Ada $0.44, NVIDIA H100 SXM $1.79, NVIDIA RTX PRO 4500 Blackwell $0.72, NVIDIA GeForce RTX 5080 $0.37, NVIDIA GeForce RTX 3080 $0.23, NVIDIA H200 NVL $3.29, NVIDIA B200 $4.50, NVIDIA GH200 $2.29, NVIDIA H100 PCIe $1.98, NVIDIA RTX 5000 Ada Generation $0.83, NVIDIA H100 NVL $2.92, NVIDIA GB200 NVL72 $6.94, NVIDIA A10G $1.01, NVIDIA GB300 NVL72 $13.30, NVIDIA Quadro RTX 4000 $0.65.
How these figures are read
A card qualifies by its memory alone; speed, interconnect and software support are not ranked here. The rate is the card’s cheapest live on-demand rate card - each provider once, at its cheapest on-demand rate, community capacity and marketplace floors out, read within 48 hours - and the median is taken over the same rate cards, as on every card page. Memory figures are floors from the arithmetic above, not measurements: context length, batch size and the serving stack change them. Methodology.
Other workloads: GPUs for LLM inference · GPUs for fine-tuning · GPUs for training · GPUs for image generation.
Data updated: