LCOC with Uncertainty ($/GPU-hr)
Orbital vs Ground market price. Bands = scenario range.
SLA: 99.9% uptime (8.76 hrs/yr downtime)
Orbital (baseline)
Orbital range
Ground (market)
Inference Cost ($/M tokens, Llama 70B eq.)
Orbital vs Ground inference pricing. Same model as LCOC.
Orbital (baseline)
Orbital range
Ground (market)
Orbital Fleet (TW)
All shells: LEO, MEO, GEO, Cislunar
LEO
MEO
GEO
Cislunar
Carbon (gCO₂/TFLOP-hr)
Orbital = embodied; Ground = grid + embodied
Orbital
Ground
Orb 2026
--
Gnd 2026
--
Orb 2030
--
Gnd 2030
--
Orb 2040
--
Gnd 2040
--
Orb 2050
--
Gnd 2050
--
LCOC with Uncertainty ($/GPU-hr)
Orbital vs Ground market price. Bands = scenario range.
SLA: 99.9% uptime (8.76 hrs/yr downtime)
Orbital (baseline)
Orbital range
Ground (market)
Inference Cost ($/M tokens, Llama 70B eq.)
Orbital vs Ground inference pricing. Same model as LCOC.
Orbital (baseline)
Orbital range
Ground (market)
Orbital Fleet (TW)
All shells: LEO, MEO, GEO, Cislunar
LEO
MEO
GEO
Cislunar
Carbon (gCO₂/TFLOP-hr)
Orbital = embodied; Ground = grid + embodied
Orbital
Ground
Peak Premium
--
Premium 2035
--
Unmet 2035
--
Supply vs Demand (GW)
Gap drives scarcity premium. Orbital-eligible demand = total × eligible share.
Total Demand
Ground Supply
Total Supply
Scarcity Premium (×)
Price multiplier from unmet demand
BW Util 2035
--
Platforms 2035
--
Platforms 2050
--
Shell Utilization (%)
LEO: 300k, MEO: 50k, GEO: 1.8k, Cislunar: 700k capacity
LEO
MEO
GEO
Cislunar
Bandwidth Utilization (%)
Available / needed. Low = stranded capacity.
Bandwidth Capacity (Tbps)
Ground segment optical links
Stranded Capex Penalty (×)
LCOC multiplier from underutilization
Radiator Mass (kg/kW)
Droplet: 50× lighter thermal system
Bottleneck Analysis
Dominant constraint each period
Thermal
Power
Bandwidth
Orbital Slots
Demand
LEO Power
--
Cislunar
--
Mass 2035
--
Mass Breakdown (kg)
Droplet radiators 50× lighter
Solar/Reactor
Battery
Compute
Radiator
Structure
Launch Cost ($/kg)
Learning curve to $15/kg floor
Compute Efficiency (GFLOPS/W)
AI learning rate drives improvement
Orbital
Ground
Platform Power by Shell (MW)
Fission/fusion enable 10-1000× power jump
LEO
MEO
GEO
Cislunar
Specific Power (W/kg)
Key figure of merit for orbital
Power Budget (%)
Compute vs thermal vs housekeeping
Compute
Thermal
Housekeeping
Data Rate per Platform (Gbps)
Link budget based on compute output
Radiation: Reliability Overhead (%)
SEU-driven redundancy by shell (improves with rad-hard tech)
LEO
MEO
GEO
Cislunar
EROL (Energy Return on Launch)
Lifetime energy output vs launch energy
Aggressive
--
Baseline
--
Conservative
--
LCOC Scenarios ($/GPU-hr)
Deployment ramp timing under different assumptions
Orbital (baseline)
Orbital range
Ground (market)
Carbon Scenarios (gCO₂/TFLOP-hr)
Emissions trajectory under each scenario
Orbital (baseline)
Orbital range
Efficiency Scenarios (GFLOPS/W)
Compute efficiency improvement
Orbital (baseline)
Orbital range
Breakthroughs
Droplet Radiators2030
50× lighter thermal, MW-class platforms
Space Fission
10-100× orbital power only
Ground SMR Buildout
Ground supply +10-25 GW/yr
Fusion Power
GW stations, 100× ground
Thermal
Emissivity0.85
Op Temp350K
Power
Solar Eff32%
Base Power300kW
Compute %68%
Battery Dens280Wh/kg
Compute
Rad Penalty35%
AI Learn Rate20%/yr
Sat Lifetime8yrs
Economics
Launch $/kg$1500
Launch Learn30%
Launch Floor$15/kg
Mfg Mult2.5×
Maint Cost15%/yr
WACC Orbital12%
WACC Ground8%
Ground
PUE1.30
Energy Cost$0.065/kWh
Energy Escal3%/yr
Queue Delay36mo
Bandwidth
2026 BW50Tbps
BW Growth35%/yr
Batch Eligible35%
Market
2026 Demand65GW
Demand Growth45%/yr
2026 Supply480GW
Build Rate15GW/yr
Derived Values
Rad Power -- W/m²
Rad Mass -- kg/MW
Launch LEO -- $/kg
Launch GEO -- $/kg
Sat Power -- kW
Sat Mass -- kg
Demand '35 -- GW
Supply '35 -- GW
LCOC with Uncertainty ($/GPU-hr)
Orbital vs Ground market price. Bands = scenario range.
SLA: 99.9% uptime (8.76 hrs/yr downtime)
Orbital (baseline)
Orbital range
Ground (market)