GPU Compute — Rent NVIDIA GPUs On Demand | VoltageGPU

GPU Pricing

• NVIDIA RTX 4090 — $0.37/hr, 24 GB VRAM, best value for inference and fine-tuning
• NVIDIA A100 — $2.02/hr, 80 GB VRAM, ideal for training and research
• NVIDIA H100 — $2.77/hr, 80 GB VRAM, transformer engine with FP8 and NVLink
• NVIDIA H200 — $4.07/hr, 141 GB HBM3e, large model training
• NVIDIA B200 — $7.50/hr, 192 GB VRAM, frontier AI and FP8

Features

• Per-second billing — pay only for what you use, no minimum commitment
• SSH access — full root access to your GPU instance
• Custom Docker templates — PyTorch, TensorFlow, Jupyter, Stable Diffusion pre-installed
• Persistent volumes — NVMe storage that persists across pod restarts
• Multi-GPU support — up to 8x H100 with NVLink for distributed training

Use Cases

• AI training — train deep learning models, distributed training across multi-GPU nodes
• Fine-tuning — LoRA, QLoRA, full fine-tuning of LLMs on consumer and enterprise GPUs
• Inference — deploy models for real-time inference with vLLM, TGI, Triton
• Rendering — 3D rendering, video encoding, VFX with Blender, Maya, Unreal Engine
• Research — academic and scientific computing with GPU acceleration

How It Works

1. Browse available GPUs — filter by model, VRAM, price, or availability
2. Deploy your pod — select a template and launch in 30 seconds
3. Connect via SSH — full root access with CUDA, drivers, and frameworks installed
4. Pay per second — no minimum commitment, stop anytime