Supreme Blue

Powering what’s next.

驅動下一場 AI 進化:台灣在地的 AI 基礎設施

Supreme Blue builds the GPUs, model training and app hosting that teams in Taiwan build on, with every byte kept on the island.

One stack, three ways in

Put an app online, train a model on your own data, or take the GPU itself. Each layer runs on the one below it.

  1. Apps

    RUNtw

    台灣在地的網站與應用平台

    Put websites, apps and databases online from Taiwan. Describe a site and let AI build it, or deploy your own code from GitHub, a Docker image or a folder.

    • Preview first, then publish, with one-click rollback
    • Postgres, MySQL, Redis and MongoDB, with backups and custom domains
    • Built-in AI models behind an OpenAI-compatible API, with prompts never used for training
    • Monthly plans in NT$, with e-invoices that carry your 統編
    Start on RUNtw
  2. Models

    LLMtw

    訓練與部署你的專業語言模型

    Train a language model on your own documents without ML expertise. Upload files, describe the job, and get a tuned, evaluated model to download or deploy.

    • Reads PDF, Word, PowerPoint, Excel and scanned pages
    • Fine-tunes Qwen and Gemma models, then compares them with the base model on held-out data
    • Exports to safetensors and GGUF, or serves the model through an OpenAI-compatible endpoint
    • Built for hospitals and companies whose data must stay in Taiwan
    Train a model
  3. GPUs

    GPUtw

    專屬 GPU 雲端

    Dedicated bare-metal GPUs ready in under two minutes and billed by the second, from an RTX 4090 up to H200 and B200.

    • No virtualization overhead, with SSH and JupyterLab on every instance
    • Vault persistent storage, with resumable uploads and server-side Hugging Face downloads
    • REST API, an MCP server and skills for coding agents
    • Prices in NT$, with quotes and invoices for schools and companies
    Rent a GPU

About us

Taiwan makes the world’s most advanced chips and trains world-class researchers, yet it had no affordable GPU cloud of its own. We started Supreme Blue to build one.

GPUtw came first, so researchers and engineers could rent a GPU in minutes instead of waiting weeks or paying overseas prices. Running it meant building our own data centers in Hsinchu and Tainan.

On that base we built RUNtw for the shops, clinics and offices that need a website more than a GPU, and LLMtw for hospitals and companies that want a model trained on their own documents. All three run on the same machines, on the island.

Data residency
Compute, storage, backups and AI inference all stay in Taiwan.
Language
Every product works in Traditional Chinese and English.
Billing
Prices in NT$, ECPay top-ups, and e-invoices with your 統編.
Company
Operated by 青金岩科技股份有限公司, tax ID (統一編號) 61992153.
Hsinchu 新竹Tainan 台南
Multiple data centers in Taiwan

Trusted by researchers and teams at

  • National Taiwan University
  • National Tsing Hua University
  • National Yang Ming Chiao Tung University
  • National Cheng Kung University
  • National Chung Hsing University
  • Academia Sinica
  • ITRI Material and Chemical Research Laboratories
  • Taipei Medical University
  • National Taipei University of Technology
  • MediaTek
  • TSMC
  • Inventec

Leadership

Luke Lin

Luke Lin 林澈

Co-founder & CTO

Works on vision language models, physical AI and high-performance computing infrastructure. Leads KSAC Lab (Knowledge System and Applied Computing), working with Taiwan’s leading universities and industry partners.

Tell us what you’re building.

Training runs, a private model for your hospital, or GPUs for your lab. Tell us what you need and we’ll reply.

Email [email protected]