Clean energy in.
Useful intelligence out.
A small solar/battery-powered computer in a home does AI work for someone who needs it. The homeowner earns. The buyer saves. The sun is not wasted.
Rooftop solar/battery
Spare midday sun
Home computer
A small quiet box
SunStack network
Routes the work
AI for a buyer
Real results
Too much sun. Too little AI.
The worst trade in the country. Every sunny day.
So many Australian businesses send their data offshore to save money.
Two problems. Never meeting.
This maths only works in Australia.
Most solar homes % on Earth
About 1 in 3 Australian homes has rooftop solar. In some suburbs, more than half.
The widest price gap
Cheap at noon, dear at night — the biggest spread in the developed world.
Batteries at scale
~350,000 homes added a battery in the scheme’s first ten months. 2M+ targeted by 2030.
Ready-made rails
The energy market already accepts home-based “virtual power plants.”
Three parties. One story.
Lends the sun
Spare solar/battery energy and a corner of the garage. Earns a share of every job the box runs.
Runs the market
Builds the box, routes each job, handles billing, security, and the energy market.
Gets the work done
Sends AI work, gets results. Never learns whose garage ran it.
The box dials out to us, so it works behind any home internet — no special setup, no fixed address.
Many small homes.
One big machine.
Thousands of homes plug into SunStack, which routes each job, meters the energy, and keeps it secure. If a storm cuts the sun in Sydney, work shifts elsewhere — or waits.
Unreliable in any one place. Reliable everywhere at once.
A quiet box on the wall.
One or more small computers — say a DGX Spark and a Mac — linked together by a fast connection (our PAIR router) and sealed into a single quiet box, about the size of a large desktop. Its own circuit. Home Internet. It works only on spare solar/battery energy, and steps aside when the house needs power.
No upfront cost
Financed from what it earns.
One sealed node
The linked units act as one node, isolated from your home network.
You’re in control
Pause it any time from your phone.
It already runs.
A buyer signs in, picks a model, and chats. Real text. Real images. Metered and billed, end to end.
# The same OpenAI client — pointed at SunStack from openai import OpenAI client = OpenAI( base_url="https://api.sunstack.ai/v1", api_key="sk-sun-…", ) resp = client.chat.completions.create( model="qwen3.5-0.8b", messages=[{"role": "user", "content": "Write a short product blurb."}], ) print(resp.choices[0].message.content)
The same API you already use.
One console runs the marketplace.
See every home. Watch each job go to the least-busy one.
One console tracks the energy.
Track usage, energy, and billing across the fleet.
Three parties. Everyone better off.
The economics are simple: cheap solar energy in, useful AI work out, and three groups share what's left.
A new income
Spare solar runs the financed box — no upfront cost. About A$2,500/yr on a single DGX Spark node.
Lower prices
Australian-hosted AI ~20% under cloud rates — saving ~A$1,150/yr — and data never leaves home soil.
A defensible business
Recurring compute revenue no overseas rival can replicate — ~A$780/yr margin per node, growing with the fleet.
Every number in this model is sourced and auditable — open the interactive calculator to change the hardware, model, energy and split, and watch all three parties move.
The margin is in the solution, not the token.
Raw inference is a commodity — priced to the cent and racing to the floor. SunStack sits one layer up: we turn cheap, local tokens into managed AI that does a real job, and price it on the value delivered.
Raw tokens
A pay-as-you-go API that undercuts the big clouds. High volume, thin margin — the floor everyone competes on.
Managed models
Private, fine-tuned models run for a team — with service levels, monitoring and support. Priced for the service, not the token.
Vertical solutions
An assistant wired into a small business’s ERP, accounts and documents — automating real workflows. Sold as an outcome, at many times the token price.
Same solar-powered tokens underneath — sold as a solution, sovereign and local.
For homeowners.
A new income
Earn a share of every job your box runs, from spare rooftop solar/battery energy.
Nothing to pay upfront
The box is financed and pays for itself.
Full control
Pause any time. Remove it any time.
For the people who use the AI.
Lower prices
Powered by the cheapest energy in the country.
Australian & sovereign
Data stays on home soil.
Familiar & trusted
The same tools you already use — with proof of where each job runs, coming soon.
For SunStack, the university, and the country.
For SunStack & the university
A defensible new business
The first integrated operator of a national compute network — recurring revenue from compute and grid services.
Real research & IP
Energy-aware scheduling, attestation, virtual power plants, tenant isolation. The output: papers, patents, and trained engineers.
A uniquely Australian story
National infrastructure no other country can build the same way.
For the country
Sovereign AI
Designed so every job runs in an Australian home.
A calmer grid
Soaks up midday surplus right where it’s made.
No new land
Uses rooftops people already own.
A big prize — from both sides.
Generative AI could add up to A$115 billion a year to the Australian economy by 2030.1 SunStack sits between fast-growing demand for AI and the largest pool of idle green energy on Earth.
Demand · the buyers
Supply · the homes
A small share is a big business.
We don’t need the whole market — just a sliver of the roofs. Here is the node economics from the calculator, scaled.
Bottom-up · the build
Top-down · the headroom
Even the 50,000-node case barely registers against the demand pool:
The limit is how fast we sign homes — not demand. Room to grow the fleet 10–20× before we touch 1% of cloud spend.
- 1 SecurityBrief AU (Gartner), 2026
- 2 Gartner, 2026
- † Per-node economics from the SunStack revenue calculator (operator A$780/yr, homeowner A$2,505/yr net, buyer saves A$1,164/yr).
Where we take it from here.
Greener, cheaper energy
Solar/battery-aware scheduling that gets more useful work out of every kilowatt-hour.
Stronger security
Hardware attestation, signed receipts, and tenant isolation between jobs.
Faster inference
Quicker responses and bigger models on the same home hardware.
Deployment & trials
Real garages onboarded, with university and industry pilots — and grid-service trials.
Start small. Prove each step. Scale what works.


