"It might not be rational. But it might be physically possible."
An interactive first-principles cost analysis comparing orbital solar power satellites to terrestrial natural gas for datacenter capacity. Built by Andrew McCalip.
This is an interactive economic model that asks a simple question: Can you make space-based commodity compute cost-competitive with the cheapest terrestrial alternative?
Instead of hand-waving about "the sun being huge" or "space being big," this tool lets you manipulate the actua 8000 l variables and see what assumptions need to be true for orbital compute to make economic sense.
- Interactive Sliders: Adjust launch costs, satellite hardware costs, specific power, gas prices, PUE, and dozens more parameters
- Real-Time Calculations: See cost breakdowns, LCOE, and engineering outputs update instantly
- Thermal Analysis: Full bifacial panel thermal model with equilibrium temperature calculations
- Propellant & Logistics: LOX/CH4 production energy, Texas grid impact, tanker truck estimates
- Mobile Responsive: Works on desktop and mobile with adaptive layouts
- First-Principles Physics: Solar flux, Stefan-Boltzmann radiation, cell degradation, view factors
- Transparent Math: Formulas are implemented directly in
static/js/math.js
- Starlink V2 Mini heritage bus as reference point
- Configurable specific power (3-75 W/kg)
- Launch costs from theoretical floor ($20/kg) to Falcon 9 ($2,940/kg)
- Cell degradation, GPU failure rates, NRE costs
- Thermal equilibrium calculations for bifacial panels
- Starship/Super Heavy propellant: 6,750 metric tons per launch (78.3% LOX)
- Energy requirements: ~0.4 MWh/ton LOX, ~0.8 MWh/ton CH4
- Texas LOX capacity impact: % of total (~14,500 t/day) and 10% surplus
- Warning indicators when demand exceeds surplus capacity
- Tanker truck estimates per launch
- H-Class combined cycle gas turbines
- 5-bucket capex model: Power Gen, Electrical, Mechanical, Civil/Shell, Network
- Fuel costs tied to natural gas prices
- PUE and capacity factor adjustments
Quick verification scripts live in temp/ and mirror the JavaScript model.
├── Flask 2.2.3 # Web framework
├── Gunicorn 20.1.0 # Production WSGI server
├── Jinja2 # Templating
├── KaTeX # Math rendering
├── Vanilla JS # No framework bloat
└── CSS Custom Props # Theming & responsive design
# Clone the repo
git clone https://github.com/andrewmccalip/thoughts.git
cd thoughts
# Create virtual environment
python -m venv venv
# Activate (Windows)
venv\Scripts\activate
# Activate (Unix/Mac)
source venv/bin/activate
# Install dependencies
pip install -r requirements.txt
# Run the app
python app.pyVisit http://localhost:5000 in your browser.
# Build the image
docker build -t space-datacenters .
# Run the container
docker run -p 8080:8080 space-datacentersVisit http://localhost:8080 in your browser.
thoughts/
├── app.py # Flask application
├── requirements.txt # Python dependencies
├── Dockerfile # Container configuration
│
├── audits/ # Archived review artifacts
├── static/
│ ├── css/
│ │ └── style.css # All styling (2400+ lines)
│ ├── js/
│ │ ├── math.js # Cost model calculations
│ │ └── main.js # UI controller & interactions
│ └── references.json # Citation data
│
├── templates/
│ ├── index.html # Landing page
│ ├── space-datacenters.html # Main analysis page
│ └── (none)
│
├── reports/ # Research notes
└── temp/ # Development scripts
| Variable | Default | Description |
|---|---|---|
PORT |
8080 |
Server port (Docker) |
Total Cost = Hardware + Launch + Ops + GPU Replacement + NRE
Hardware = $/W × Initial Capacity (degradation-adjusted)
Launch = $/kg × Total Mass
Mass = Initial Capacity / Specific Power (W/kg)
T_eq = [ΣQ_in / (σ·A·ε_tot) + T_space⁴]^¼
Where:
- Q_in = Q_solar + Q_earthIR + Q_albedo + Q_heatloop
- σ = Stefan-Boltzmann constant
- ε_tot = ε_pv + ε_rad (bifacial)
LCOE = Total Cost / Energy Output ($/MWh)
Energy = Capacity × Hours × Capacity Factor
From the analysis (with default assumptions):
- It's not obviously stupid — The physics doesn't immediately kill it
- The economics are savage — Currently 3-4x more expensive than terrestrial
- Vertical integration is everything — Margin stacking kills external procurement
- One organization is positioned — SpaceX's integration advantage is the whole ballgame
- The math is explicit — Formulas are in
static/js/math.js
Found an error in the model? Disagree with an assumption? Good.
- Fork the repo
- Run the numbers yourself
- Open a PR with your corrections
- Include sources
The goal is to drag the conversation back to first principles—assumptions you can point at and outputs you can sanity-check.
MIT License — Do whatever you want with it.
Andrew McCalip
- Twitter: @andrewmccalip
- GitHub: andrewmccalip