What is TapGPU?
TapGPU is an hourly cloud GPU rental platform for AI training, inference, and generative workloads. Instead of buying a card for bursty projects, you rent runtime by the hour.
Homepage: https://www.tapgpu.com/?utm_source=chatgpt.com
Which GPUs can I rent?
Depending on stock: consumer cards such as RTX 3090 / 4090 / 5090, and data-center options including A100 / H100 / H200. Community Cloud and Secure Cloud tiers are available.
How much does it cost per hour?
Public consumer hourly pricing is about ¥0.77–3.12/hour (snapshot; low→high roughly Community→Secure). Data-center SKUs cost more. Always confirm on the live board:
https://www.tapgpu.com/pricing
Who is it for?
Side-project builders, indie hackers, and teams who need GPUs for LoRA / fine-tuning, inference experiments, ComfyUI / image generation, or short training runs — without a long hardware purchase.
How do I start?
Open the homepage, pick a GPU in the console, launch a template if you want, and pay for the hours you actually run. Stop compute and clean unused disks when you finish (storage can still bill after compute stops).
Disclosure
I help operate TapGPU. Prices and stock change; the console is the source of truth.
Top comments (0)