Rent GPU Compute Online

A Taiwan data center, a whole machine to yourself, activated on the spot. Billed for the time you use, stop anytime.

iGPU.tw
A machine to yourself
Never sharing a GPU with strangers
Taiwan data center
Data stays in Taiwan
Billed by time used
Stop anytime, no contract

GPU Infrastructure

From entry-level to flagship — pick the card that fits the workload. Every lease is one whole machine.

Full

RTX PRO 6000

96 GB GDDR7 ECC
TWD60/hr
Full

RTX 5090

32 GB GDDR7
TWD40/hr
Full

RTX 5070 Ti

16 GB GDDR7
TWD20/hr
Full

AMD AI PRO R9700

32 GB GDDR6
TWD30/hr
Full

Mac Studio M5 Ultra

256 GB Unified Memory
TWD60/hr
Full

Mac Studio M5 Max

128 GB Unified Memory
TWD40/hr

Popular Environments, Ready to Deploy

These common AI frameworks and tools can be pre-installed when your machine is provisioned — no manual setup needed.

Jupyter Lab
Notebook

A lightweight CUDA + JupyterLab base image with no framework preinstalled — add PyTorch/TensorFlow yourself as needed.

Ollama + Open WebUI
Recommended

Run open-source LLMs locally with a ChatGPT-style web chat interface.

ComfyUI
Generative AI

A node-based Stable Diffusion workflow interface — modular and flexible.

Ubuntu 24.04 Base
OS Base

A clean Ubuntu 22.04 base image with Docker and GPU drivers, ready for you to build your own environment.

llama.cpp Server
Recommended

A lightweight GGUF inference engine with a built-in OpenAI-compatible API — quantized models run fine even on small GPUs.

PyTorch 2.x + JupyterLab
Notebook

A CUDA PyTorch dev environment with JupyterLab, torchvision, torchaudio and common packages.

vLLM Inference Server
Recommended

A high-throughput LLM inference framework with an OpenAI-compatible API endpoint, built for production deployment.

Unsloth Fine-Tuning
Notebook

A JupyterLab environment preloaded with an Unsloth fine-tuning example — great for learning and quick experiments.

How do I start?

Four steps from signing up to running your workload.

1. Join

Log in with email or phone + a verification code. You can check "remember this device".

2. Buy hours

Pay online by card — credited and ready to use as soon as payment clears.

First purchase 12% off

3. Pick a spec & activate

SSH and Jupyter access details the moment payment clears. You're not billed while the environment initializes.

4. Done? Stop.

Billed hourly; data kept for 2 hours after the lease ends for download.

Based in Taiwan

The data center, the GPU machines, the company, the invoices and the support are all in Taiwan — so you can use it with confidence.

Taiwan Company

網絡商數科技有限公司 — a professional server services provider, serving customers in Taiwan for over 20 years. Our data center and support are both here.

Taiwan Invoicing

All prices include tax. An e-invoice is issued automatically with every purchase, and you can opt for a company invoice with your tax ID.

Taiwan Data Center

Servers housed in a Taiwan data center. A dedicated machine to yourself — your data never leaves the country.

Local Support

Professional support in both Chinese and English — no timezone headaches.

Ready to deploy?

Signing up is free. Billed by the hour — you only pay for the time you actually use, and you can stop anytime. No contract.