New Zealand · gpu.co.nz
Your Private AI Universe Awaits
Powering unlimited private AI — locally, securely, and without limits. The cosmos of intelligence, yours to explore.
Philosophy & Vision
Modern, cost-effective, plug-and-go private AI machines preloaded with the latest open-source software — ready to chat, prompt, and generate from day one. Always moving with the pace of technology, always shipping with the latest AI tooling. Available for Linux, Windows, and macOS.
Hardware Lineup
Every machine ships pre-installed with the latest AI tooling — ready to run from the moment you power on. No setup headaches, no subscriptions, no cloud dependency.
Education Specialisation
With over two decades in the classroom, from Year 7 to Year 13 across creative and technology subjects, the GPU team deeply understands how AI should work in education — privately, safely, and powerfully.
Students interact with AI locally — no data leaves the classroom. Safe, privacy-compliant, and completely secure for every learner.
Pre-loaded with NZ Curriculum content, NCEA standards and exemplars — AI that speaks the language of New Zealand education from day one.
Image and video generation for digital arts, media studies, and design classes — all running locally with no content filtering surprises.
Educators choose which AI models students access, with full local control and zero reliance on external APIs or subscriptions.
Centralised AI server accessible across the entire school network — one powerful machine serving every classroom simultaneously.
IT administrators maintain full visibility and control over AI usage across all departments with local logging and access management.
The entire AI stack runs on-premises. No outages from third-party services, no data sent to the cloud, no surprise downtime.
Replace endless per-user subscription costs with a single capital investment in hardware that serves your school for years.
Training & Professional Development
Practical AI training grounded in real-world experience — over 20 years in education and 20 years building custom systems, from hands-on hardware to in-depth prompting and deployment strategies for educators and businesses alike.
Available Fridays and weekends for arranged on-site professional development. Staff leave with practical skills and confidence to use AI daily.
Weekend courses include a pre-loaded mini PC to take home and use. Learn with the actual hardware — kept for continued exploration.
Hands-on consulting on how AI can be integrated into business workflows — from automated document handling to private internal chatbots.
Specialised training for creative disciplines — digital media, design, and technology subjects — backed by over two decades of teaching experience.
Continued support as technology evolves — updates, expansions, and guidance to keep every machine running with the latest AI tooling.
Specialised Builds
From a Raspberry Pi tucked in a backpack to a rack-mounted multi-GPU server humming in a server room — if it runs AI, it's been built, experimented with, and refined.
Every build is an experiment. Every experiment becomes a product. That's the GPU way.
Enquire About a Custom BuildReady to launch?
Whether you're an educator exploring AI for your school, a business looking to run AI privately, or an enthusiast wanting total control — GPU is ready to build your private AI universe.
Entry Orbit · Chat & RAG
The smallest footprint, the biggest leap. Private AI that fits in your hand — runs all day on the power of a light bulb.
Performance Profile
Use Cases
What's Included
Apple Orbit · All-in-One
The most powerful small box on earth. Apple silicon makes every AI workload feel effortless — pre-loaded and ready to generate from day one.
Performance Profile
Use Cases
What's Included
GPU Orbit · Intel ARC
Purpose-built for AI acceleration. Intel's ARC GPU delivers serious compute at a fraction of the cost — perfect for large image generation and fast model inference.
Performance Profile
Use Cases
What's Included
Multi-GPU Orbit · Maximum Power
The apex predator of private AI. Three Nvidia GPUs with up to 72GB combined VRAM — run any open-source model on the planet, entirely offline.
Performance Profile
Use Cases
What's Included