Blackwell UltraCapacity request
NVIDIA B300
288 GB HBM3e
EU West · Paris · EU Central · Frankfurt
Frontier reasoning and very long contexts
Specs and rental guideGPU directory
Rent cloud GPUs by the hour.
Compare 19 NVIDIA and AMD GPUs by VRAM, workload fit, hourly price and European request region. Open a model guide, calculate the full rental cost, then configure the selected resource from your workspace.
Calculate GPU rental cost Create a workspace19 priced accelerators
NVIDIA and AMD, from 16 to 288 GB.
Catalog rates and requestable status last reviewed September 1, 2026. The final provisioning state remains visible in the workspace.
CDNA 3Capacity request
AMD Instinct MI300X
192 GB HBM3
EU West · Paris · EU Central · Frankfurt
LLM training, inference and HPC under ROCm
Specs and rental guideBlackwellCapacity request
NVIDIA B200
180 GB HBM3e
EU West · Paris · EU Central · Frankfurt
Frontier training and inference
Specs and rental guideHopperCapacity request
NVIDIA H200 SXM
141 GB HBM3e
EU West · Paris · EU Central · Frankfurt
Large-model inference and HPC
Specs and rental guideBlackwellCapacity request
NVIDIA RTX PRO 6000 Blackwell
96 GB GDDR7 ECC
EU West · Paris · EU Central · Frankfurt
Agentic AI, science and rendering
Specs and rental guideHopperCapacity request
NVIDIA H100 NVL
94 GB HBM3
EU West · Paris · EU Central · Frankfurt
Large-model inference
Specs and rental guideHopperCapacity request
NVIDIA H100 SXM
80 GB HBM3
EU West · Paris · EU Central · Frankfurt
Intensive training and FP8 inference
Specs and rental guideAmpereCapacity request
NVIDIA A100 80GB
80 GB HBM2e
EU West · Paris · EU Central · Frankfurt
Training, fine-tuning and HPC
Specs and rental guideAdaCapacity request
NVIDIA RTX 6000 Ada
48 GB GDDR6 ECC
EU West · Paris · EU Central · Frankfurt
Professional AI, rendering and simulation
AdaCapacity request
NVIDIA L40S
48 GB GDDR6 ECC
EU West · Paris · EU Central · Frankfurt
Production inference, fine-tuning and video
Specs and rental guideAdaCapacity request
NVIDIA L40
48 GB GDDR6 ECC
EU West · Paris · EU Central · Frankfurt
Generative AI and visualization
AmpereCapacity request
NVIDIA RTX A6000
48 GB GDDR6 ECC
EU West · Paris · EU Central · Frankfurt
LoRA, visualization and larger models
AmpereCapacity request
NVIDIA A40
48 GB GDDR6 ECC
EU West · Paris · EU Central · Frankfurt
Inference, rendering and fine-tuning
BlackwellCapacity request
NVIDIA RTX 5090
32 GB GDDR7
EU West · Paris · EU Central · Frankfurt
Fast single-GPU inference and media
Specs and rental guideAdaCapacity request
NVIDIA L4
24 GB GDDR6
EU West · Paris · EU Central · Frankfurt
Efficient inference, video and endpoints
AdaCapacity request
NVIDIA RTX 4090
24 GB GDDR6X
EU West · Paris · EU Central · Frankfurt
Image, video and mid-size inference
Specs and rental guideAmpereCapacity request
NVIDIA RTX 3090
24 GB GDDR6X
EU West · Paris · EU Central · Frankfurt
Cost-efficient image generation
AmpereCapacity request
NVIDIA RTX A5000
24 GB GDDR6 ECC
EU West · Paris · EU Central · Frankfurt
General ML, Stable Diffusion and LoRA
AmpereCapacity request
NVIDIA RTX A4000
16 GB GDDR6 ECC
EU West · Paris · EU Central · Frankfurt
Development and lightweight inference
Choose by workload
VRAM first. Then throughput. Then price.
LLM training
Start with model size, optimizer state and batch target. H100, H200 and B200 fit the heaviest jobs.
Open workload guideAI inference
Compare input shape, batch, p95 latency and cost per accepted prediction for vision, audio and embeddings.
Open workload guideImage & video
VRAM and media accelerators matter more than peak tensor FLOPS. L40S, RTX 5090 and L4 are strong fits.
Open workload guideSelection checklist
Avoid the expensive mismatch.
Confirm the model fits in VRAM
Check region-level stock
Compare PCIe, SXM and NVLink
Include storage and stopped-state cost
Test throughput with your own batch
Plan a compatible fallback GPU
Review license and data residency
Set a budget alert before launch
From hardware to resource
Need the container, storage and region workflow?
Ready to build?