GPU Cloud Pricing — VoltageGPU

Transparent pricing for GPU compute, AI inference, fine-tuning and confidential compute. Per-second billing, no hidden fees.

GPU Compute Pricing

• NVIDIA RTX 4090 (24 GB VRAM): from $0.37/hour
• NVIDIA A100 80GB (80 GB HBM2e): from $2.02/hour
• NVIDIA H100 80GB (80 GB HBM3): from $2.77/hour
• NVIDIA H200 141GB (141 GB HBM3e): from $4.07/hour
• NVIDIA B200 192GB (192 GB HBM3e): from $7.50/hour

AI Inference Pricing

• DeepSeek-R1: $0.46/M input tokens, $1.85/M output tokens
• Qwen3-32B: $0.15/M input, $0.44/M output
• Llama 3.3 70B: $0.35/M input, $0.40/M output
• 140+ models available via OpenAI-compatible API

Fine-Tuning Pricing

• Small models (1-3B params): from $18.50/hour
• Medium models (7-13B): from $27.75/hour
• Large models (30-70B): from $46.25/hour
• Image LoRA (FLUX/SDXL): from $18.50/hour

Confidential Compute

Intel TDX encrypted GPU instances. Prices vary by configuration. Available GPUs: H100, H200, B200, RTX 4090.

Frequently Asked Questions

How does VoltageGPU billing work?

VoltageGPU uses per-second billing. You only pay for the exact time your GPU is running. Stop your pod and billing stops instantly.

Is there a minimum commitment?

No. No contracts, no reserved instances. Deploy and stop anytime. Minimum top-up is $5.

How do VoltageGPU prices compare to AWS?

VoltageGPU is typically 50-85% cheaper than AWS for equivalent GPU instances. For example, an H100 costs $2.77/hr on VoltageGPU vs $4.30/hr on AWS.