Every layer of SUN 1.0, taken apart.
Scenario
Layers >>top to bottom
Box-in-box hydrogen vault
Electrolyzer cell
A data center that powers itself from sun, air and water.
Sunlight is the energy. Air and rain are the feedstock. Hydrogen and batteries carry the day into the night. No grid connection request, no fuel, no smokestack.
Upgrades from the patent baseline to SUN 1.0
Every patent component stays. These changes use parts that already ship today, and the numbers update with your site and scale.
What SUN 1.0 runs on.
Three layers, each on the hardware that fits its job. The safety layer is fixed by law and physics. The optimizer, the part that decides every hour where the sun goes, is a real choice: it is light enough that a Mac could run it.
Optimizer computer options >>real specs
Replay the plant through real weather, hour by hour.
Five measured years at this site (2019–2023): sun from PVGIS on a single-axis tracker, air and rain from NASA POWER. The plant is sized on all 43,824 hours and replayed here exactly: battery, hydrogen and water levels as they would have been.
Run conditions
What the run shows
The graphene layer eats the GPU power swings.
Synchronized training jobs swing power by tens of percent in under a second. The supercapacitor bank absorbs the swing so the fuel cells and batteries see a flat line.
From one 5 kW unit to a 1 GW AI factory.
Pick a scale. Equipment, land, water, materials, vendors, capital and schedule recalculate for the site, design and case in the header.
Bill of system
Capital by subsystem
Materials list >>vendors, quantities, cost
Schedule >>months to first power
Race to energize
Outcomes
Team & supply chain
Permits it removes >>and keeps
Assumptions >>downside · base · upside, each sourced
What it costs, what it earns, and when it beats the grid.
A 25-year project-finance model on the plant, then the owner's view: what powering compute years early is worth against waiting for a grid hookup.
How SUN 1.0 got cheaper >>price it must charge, lever by lever
Levers
Sources & uses
25-year cash flow
Sensitivity >>required tariff, downside ↔ upside
Three cases side by side
Beating the grid >>100% sun, no gas, no grid: price SUN 1.0 must charge vs the local grid, on real 2019–2023 weather
Against the alternatives >>per kW of firm 24/7 load
DYFRENT economics >>proposed license terms
Annual detail
The constraint is the plug, not the money.
The four largest hyperscalers guided to roughly $720–745B of 2026 capex. Texas froze new large-load grid connections and then state permits. Ohio's utility bills data centers for 85% of contracted power for 12 years. A new gas turbine ordered today arrives around 2031. A building that never asks the grid for anything skips every one of those lines.
Texas
Ohio
The SUN 1.0 answer
- No interconnection request. Nothing enters the ERCOT or PJM queue, the SB 6 audit or an AEP minimum bill.
- No combustion. Electrolyzers and fuel cells burn nothing, so there is no turbine air permit.
- No municipal water. Rain off the solar field, water from air, and fuel-cell water returned to the loop.
- No curtailment exposure. A site off the grid cannot be remotely disconnected in a grid emergency.
- Flat power for spiky AI. The graphene bank absorbs training swings on site.