ABS AITPC Powered by PHISON aiDAPTIV+
ABS AITPC (AI Training PC) is an affordable, all-in-one hardware solution designed for AI model inference and training. Unlike typical AI PCs that can only run existing models (inference), AITPC empowers users to train and fine-tune models—even using your own private data—directly on-premises.
It’s built by ABS and powered by their PHISON aiDAPTIV+ platform, which is engineered for scalable, domain-specific AI training at a fraction of the cost of traditional AI servers.
Key Benefits of AITPC
Dual Functionality: Inference + Fine-tuning
- Perform real AI model training, not just inference.
- Train models like Llama-2-13B, Whisper, and Stable Diffusion SDXL directly on the device.
Cost-effective AI Training
- At around $4,500 USD, it’s vastly more affordable than traditional AI servers (which require 10–100 GPUs).
- Delivers hands-on AI training power to schools, researchers, and developers.
Edge + Privacy Focused
- Keeps your data and models on-prem, protecting intellectual property and privacy.
- Ideal for schools, research institutions, and enterprises with sensitive datasets.
Scalable GPU Architecture
- Flexible GPU configurations (NVIDIA, AMD, Intel).
- Breaks the memory bottleneck of HBM by using NAND-based extension—more GPU RAM at lower cost.
Who Needs AITPC?
Education and AI Talent Development
- Helps students learn by doing: not just using AI tools but also training their own models.
- Already adopted by National Taiwan University, National Yang Ming Chiao Tung University, National Cheng Kung University, and others, and have already established AITPC facilities for student education and professional training for the public.
- All finetuning can be referred to as Full Parameter Model Training, which actually emphasizes the need for a large amount of memory. This highlights that our system is capable of handling large-scale training and supports high-capacity model training effectively.
Research Institutions
- Supports full-parameter training for models up to 13B+ parameters—ideal for universities and labs.
- Covers both Generative AI theory and hands-on practice.
Businesses and Enterprises
- Enables domain-specific fine-tuning for verticals like medical, finance, retail, or specific languages.
- Drives competitive edge through customized, in-house trained AI.
Education Program Overview
What Students Learn
- Model architecture
- Inference and deployment
- RAG (Retrieval-Augmented Generation)
- Fine-tuning techniques
- Real-world AI applications
Teaching Format
- 3-weekend hybrid course (16 hours total)
- Lab access via remote use of AITPC
- Instructor certification program also available
Affordable LLM Training On-Site
AITPC bridges the gap between theory and practice in AI education and makes local fine-tuning and custom model training possible for everyone—from students to researchers to industry professionals.
It’s cost-effective, scalable, and practical, making it the perfect tool to train the next generation of AI talent.
“Not just learn how to use LLMs—learn how to train your own.”
