Compute, available the moment you need it
iGPU.tw is operated by 網絡商數科技. We've been building computers, systems and software since 2002, and in recent years our focus has shifted to AI — on-premise language models, enterprise knowledge bases, workstation and storage architecture. iGPU.tw is the next step on that path: letting people who need compute get it without buying a whole machine first.
Rent the compute, skip the hardware
Training a model, running inference, rendering — the need is usually "GPU-heavy for a few days, then idle." Buying a flagship card means paying for the chassis, power supply, cooling, rack space, electricity and depreciation too. Once the project ships, the machine mostly gathers dust.
iGPU.tw turns that into a service: billed by the hour, one whole machine to yourself, stop when you're done. You pay for the hours you actually use, instead of taking on the capital cost of a machine for a few days of peak demand.
A range of hardware
From NVIDIA RTX 3090/5090 and RTX PRO 6000 to DGX Spark, Apple Mac Studio M5 and AMD Radeon AI PRO — different cards for different workloads.
Environments preinstalled
JupyterLab, PyTorch, Ollama + Open WebUI, ComfyUI, vLLM, llama.cpp and more can be set up in advance — connect and start working.
A machine to yourself
One lease equals one whole machine — no sharing a GPU with strangers. Wiped when the lease ends; data is kept 2 hours for download, then deleted.
Based in Taiwan
Hosted in a Taiwan-based data center; your data never leaves Taiwan. A Taiwanese company issuing Taiwanese e-invoices, with support and contracts in your timezone.
Why “iGPU”
The “i” in front of the name is deliberate, just like iPhone, iMac and iPad. It turns a generic GPU into compute that is dedicated, personal and always within reach. We read that “i” two ways:
Internet — connected
Cloud and network-coordinated computing. The GPU doesn’t have to sit under your desk; it’s there over the network whenever you need it. Graphics and AI compute designed for the internet era.
Individual — personal
“My GPU”: a whole machine that is yours alone, a local AI compute core. More and more people run AI models on their own workstations and want compute they control. iGPU.tw gives you that independence without buying the hardware first.
Put the two together and you have what iGPU.tw is for: compute you reach over the network, and that is yours alone while you use it.
Two decades of systems and software
網絡商數科技 (網絡商數科技有限公司) was founded in 2002 and launched Taiwan's first online workstation configurator — build your own spec, transparent pricing, one-stop service and repair.
The team knows AI, Windows, Linux and macOS tuning inside out, and designs Mac, NAS and AI storage setups as well as cross-platform workgroup and server architecture. From picking parts to system integration to optimizing the whole workflow, that's two decades of experience you can't shortcut — and it's why we're comfortable handing customers an entire machine.
These days we're also a software house and an AI adoption consultancy: deploying language models on-premise so confidential data never goes to a public cloud, using RAG so AI can read a company's own knowledge base, and wiring it into ERP/CRM, inventory, order and invoicing systems. We work with GPUs, models and storage every day — we know exactly how much it hurts when compute runs short.
Before "cloud" was a buzzword
In 2005 the team launched JUMBO, a personal cloud brand giving every user their own address like YourName.mac.to, bundling a personal site, blog, photo albums and cloud storage. We were doing it before most people had heard the word "cloud."
We built the Chinese edition of international fashion title Marie Claire, and for Apacer we handled planning, design and development end to end, delivering an official site spanning thirteen languages.
From personal cloud to international media to multilingual corporate sites, that hands-on work across content, systems and users gave us a more than theoretical understanding of how digital products behave in the real world.
The same standard, whatever the size of the job
Our focus stays on what matters most to customers: stability, delivering on time, and sticking around for the long run. Big client or small, everyone gets the same standard — expertise, innovation, quality, peace of mind, service. We exist because of our customers, and their trust is what we're proudest of.