Space Datacenter TCO Model
A first-principles framework for orbital compute: the full cost stack of a space datacenter, from launch vehicle physics and thermal rejection limits to GPU-level cost of ownership, stress-tested against the terrestrial constraints that would push compute off Earth in the first place.
- Horizon
- 2026-2050
- Output
- TCO, LCOC
If and when compute leaves Earth.
The SemiAnalysis Space Datacenter TCO Model evaluates the economics, engineering constraints, and supply-demand dynamics of orbital compute, and answers one question: if and when AI datacenters in space become cost-competitive with terrestrial ones. The space cost stack is built line by line from first principles, and every item is adjustable for cost and technology scaling.
The model spans 2026 to 2050 on an annual basis with dynamic, user-controlled scenarios, and computes total cost of ownership for both space and terrestrial deployments, output per GPU-hour, per PFLOP-hour, and per billion tokens.
Who uses it, and what it decides.
The buyers of this model, and the calls they make with it.
- Investors & strategists
- Whether orbital compute is real and when it matters: crossover and parity-ratio outputs under fully adjustable scenarios, with sensitivity to launch costs and terrestrial constraints.
- Engineering & infrastructure teams
- The space cost stack from first principles: subsystem-level capital costs, launch vehicle economics, thermal rejection limits, and the terrestrial power and silicon constraints stress-tested against them.
From launch physics to cost per token.
Every layer of the stack is built from first principles, and every assumption stays in your hands.
The cost stack
Total cost of ownership for space and terrestrial deployments, bridged to a levelized cost of compute:
Three cost buckets
IT capital (servers, networking, storage, software, burn-in), datacenter capital including launch, and operating cost
TCO to LCOC
Bridged via radiation availability and GPU redundancy and SLA provisioning, in dollars per GPU-hour, PFLOP-hour, and billion tokens
Your assumptions
Adjustable WACC, useful life, utilization, and PUE across every scenario
The space system, line by line
Space capital costs built from first principles, every item adjustable for cost and technology scaling:
Power & thermal
Solar arrays from cell technology to end-of-life sizing, radiators from Stefan-Boltzmann limits to the droplet roadmap, cold plates, pumps, and the transport loop
Bus & propulsion
Battery and PMAD, ADCS with star trackers, reaction wheels and magnetorquers, propulsion hardware, feed system, propellant and tank
Protection & the rest
Radiation and MMOD shielding, communications and the ground segment, structure and integration, and assembly, integration and test
Launch economics
Forecasting from Falcon 9 to Starship, priced the way the rockets actually fly:
Cost per kilogram
Cost per launch, payload to low Earth orbit, and the cost per kilogram that falls out
Reuse maturity
Early, maturing, and mature reuse scenarios for each vehicle
Cost trajectories
Slow, base, and Elon launch-cost paths, with their sensitivity to the levelized cost of compute
The terrestrial constraint
The reason space gets considered at all: terrestrial supply and the silicon constraint modeled as five layers of incremental power supply, with user-adjustable foundry and HBM wafer capacity additions and Terafab scenarios. Unconstrained against silicon-constrained AI power demand, terrestrial capacity, the implied space datacenter install base by scenario, and the cost crossover, parity ratios, and sensitivity tables that fall out.
Priced part by part.
Every part of the machine is a group of line items in the model. Point at one, in the drawing or in the list, and see what it covers.
Bill of materials
A schematic spacecraft: a compute hull at the center, solar wings to the left and right, radiator panels above and below, a bus module with propellant tanks and thrusters, a shielding contour around the hull, and a dashed launch fairing envelope around everything. Selecting an item in the bill of materials highlights that part of the drawing.
The drawing is illustrative. In the model each subsystem is a set of adjustable line items, priced from first principles, and the launch line reprices every kilogram of it.
How the model is built.
First principles up, constraints down, and the crossover where they meet.
-
Build the space stack
Every subsystem is costed line by line from first principles, from solar arrays and radiators to shielding and integration, with launch priced per vehicle and reuse scenario.
-
Model the constraints
Terrestrial supply and the silicon constraint are modeled as five layers of incremental power supply, with orbital mechanics setting availability in space.
-
Find the crossover
Both deployments are levelized to cost of compute and compared year by year, with crossover, parity ratios, and full sensitivity tables as the output.
Common questions.
Anything not covered here, ask the team directly through the form below.
What does the Space Datacenter TCO Model include?
The full cost stack of a space datacenter built from first principles, total cost of ownership for space and terrestrial deployments bridged to a levelized cost of compute, launch economics for Falcon 9 and Starship, terrestrial supply and silicon constraints as five layers of incremental power, demand and installed base by scenario, cost crossover and parity outputs with sensitivity tables, orbital mechanics, and 3-D interactive CAD models of major components. Annual from 2026 to 2050.
What are the two headline scenarios?
A Base Case with a bullish chip and Earth datacenter capacity ramp, and an “Elon Musk” Case with constrained terrestrial supply and aggressive launch and Terafab assumptions. Both are fully adjustable, and every scenario is driven by user-controlled assumptions.
What outputs does it produce?
Total cost of ownership in dollars per GPU-hour, per PFLOP-hour, and per billion tokens, bridged to a levelized cost of compute via radiation availability and GPU redundancy and SLA provisioning, plus cost crossover and parity-ratio outputs and full launch-cost and terrestrial-cost sensitivity tables.
What is in the space cost stack?
Solar arrays, radiators and the thermal transport loop, battery and power management, bus hardware from ADCS to reaction wheels, propulsion and propellant, radiation and MMOD shielding, communications and the ground segment, and structure, integration, and test. Every line is adjustable for cost and technology scaling.
Is it part of the SemiAnalysis newsletter subscription?
No. Industry models are separate institutional offerings and are not included with the annual newsletter membership.
Models that pair with this one.
The terrestrial side of the comparison lives in its own models. These carry the ground truth.
AI Cloud TCO Model
The economicsGPU-level cost of ownership on the ground, the baseline this model measures orbit against.
View model
Datacenter Industry Model
The groundThe terrestrial half of the crossover: datacenter capacity, power, and the buildout on Earth.
View model
Energy Model
The powerThe terrestrial power constraint in depth: the grid that orbital compute would be escaping.
View modelGet the Space Datacenter TCO Model.
Start with the sales team. They come back with scoping, licensing, and pricing for your mandate.
- Scoped to your use case
- Every assumption adjustable, scenario by scenario
- Custom research engagements available
