Satellite Mass Budget per kW_IT

What is the total mass per kW_IT for an orbital compute satellite?

What is the total mass per kW_IT (kg/kW_IT) for an orbital compute satellite, combining solar arrays, thermal rejection, and compute hardware?

Answer

The total satellite mass per kW_IT ranges from 13.3 kg/kW_IT (optimistic, 2026) to 68.1 kg/kW_IT (conservative, 2040), with a central 2026 estimate of 33.3 kg/kW_IT. These values include solar arrays, thermal rejection systems, compute hardware, eclipse batteries, and a scenario-dependent structural overhead factor (optimistic 1.15x, central 1.25x, conservative 1.40x) for bus, wiring, and propulsion.

The mass budget is now time-varying because battery mass depends on cost-optimal battery sizing, which in turn depends on launch cost. At high launch costs (2026), the model may select no battery (accepting eclipse downtime); as launch costs fall, batteries become cost-effective and satellite mass increases. Solar array mass depends on orbital PUE (optimistic 1.035, central 1.05, conservative 1.10).

Inputs

Analysis

Component Breakdown

The satellite mass budget per kW_IT decomposes into four subsystems plus structural overhead:

Solar array mass is computed as the total power required (IT load times the orbital PUE) divided by the solar array specific power. The orbital PUE varies by scenario: 1.035 (optimistic), 1.05 (central), 1.10 (conservative). The central estimate yields 7 kg/kW_IT. The optimistic case at 300 W/kg yields 3.5 kg/kW_IT, while the conservative case at 100 W/kg yields 11 kg/kW_IT -- a ~3x range driven primarily by solar panel technology maturity.

Thermal system mass is computed as the total dissipated power (IT load × orbital PUE, since all electrical power — including conversion losses and housekeeping — ultimately becomes heat) divided by the radiative cooling specific power. The central estimate yields 13.6 kg/kW_IT. The range from 4.1 kg/kW_IT (optimistic, 250 W_rejected/kg) to 22 kg/kW_IT (conservative, 50 W_rejected/kg) reflects a 5x span driven by radiator panel areal density and operating temperature.

Compute hardware mass contributes the smallest fraction in all scenarios: 4.0-9.0 kg/kW_IT. The optimistic case now matches the HGX B200 baseboard floor (~4 kg/kW_IT), the irreducible mass of GPU packages, HBM stacks, VRMs, and PCB. Space-qualification overhead pushes the range modestly higher. Despite this, the compute subsystem remains a minor mass component relative to solar and thermal.

Eclipse battery mass provides power during eclipse periods in the dawn-dusk SSO. Battery sizing is determined by cost optimization: the model selects the battery duration that minimizes availability-adjusted TCO with a bias toward uptime. Battery mass is time-varying — at high launch costs, the mass penalty makes batteries uneconomical and the satellite accepts eclipse downtime; as launch costs fall, full ride-through batteries become cost-optimal. In the central case, battery mass ranges from 0 kg/kW_IT (2026, no battery) to 5.6 kg/kW_IT (2040, full ride-through).

Structural overhead covers the satellite bus, wiring harness, propulsion, ADCS, and communications hardware. This is applied multiplicatively to the sum of the four subsystems. The overhead factor varies by scenario: 1.15x (optimistic, reflecting tight integration and mass optimization), 1.25x (central), and 1.40x (conservative, reflecting heavier propulsion and redundancy requirements). See the structural overhead page for the full derivation.

Which Component Dominates?

The dominant mass contributor shifts across scenarios:

In the optimistic scenario, thermal is the largest subsystem at 4.1 kg/kW_IT (~36% of pre-overhead mass), with compute at 4.0 kg/kW_IT (~35%) and solar at 3.5 kg/kW_IT (~30%). Advanced solar arrays (300 W/kg) and lightweight radiators (250 W_rejected/kg with 2 kg/m^2 panels) bring all three subsystems to comparable mass.

In the central scenario, thermal is the heaviest single subsystem at 13.6 kg/kW_IT (~51% of pre-overhead mass), with solar at 7 kg/kW_IT (~26%). Compute contributes ~23%.

In the conservative scenario, thermal dominates decisively at 22 kg/kW_IT (~52% of pre-overhead mass), reflecting the steep penalty of conventional 5 kg/m^2 radiator panels at lower operating temperatures. Solar is second at 11 kg/kW_IT (~26%), and compute remains smallest at 9.0 kg/kW_IT (~21%).

Why Thermal Mass Is the Key Uncertainty

The thermal system exhibits the widest relative range of any subsystem (5x from optimistic to conservative), compared to ~3.7x for solar and ~2.6x for compute. This is because radiative cooling performance depends on the product of two highly uncertain variables: radiator panel areal density (2-5 kg/m^2) and system overhead factor (1.2-1.5x). The T^4 Stefan-Boltzmann scaling means that even modest differences in operating temperature (70C vs 85C) produce significant W/m^2 differences, which propagate through to mass. As noted in the radiative-cooling-density analysis, operating at 85C rather than 70C reduces required radiator area by ~40%.

Cross-Check Against Published Estimates

The central estimate of 33.3 kg/kW_IT aligns well with independent estimates. Industry proposals for 100 kW-class satellites span a 1–5.4 ton range (10–54 kg/kW_IT), bracketing our model's optimistic-to-conservative range. See the satellite GPU capacity scaling side page for the full analysis of satellite sizing and architecture tradeoffs.

• SpaceX AI Sat Mini at ~10 kg/kW sits below our optimistic case (13.3 kg/kW_IT). This is consistent: SpaceX assumes next-generation solar arrays (200+ W/kg vs our optimistic 300 W/kg), a custom D3 chip running at elevated temperatures to reduce radiator mass, and minimal batteries in a dawn-dusk SSO. Our optimistic already uses aggressive component parameters; SpaceX's figure suggests that a purpose-built, vertically integrated design could achieve slightly better than our optimistic mass budget spacex-ai-sat-mini-spacenews.1.
• Mach33's 100 kW scaling at ~54 kg/kW sits within our conservative range (~65.3 kg/kW_IT in 2026 including battery). Their breakdown shows solar arrays dominating at ~48% of mass, radiators at ~18%, and bus/structure at ~34% — consistent with our finding that solar and thermal are the primary mass drivers mach33-cooling.1.
• Elon Musk's observation that "the solar array is most of the weight on the satellite" is borne out in the optimistic and central scenarios where solar is the largest or second-largest subsystem, but is contradicted in the conservative case where thermal dominates.
• The 100 kW satellite is the industry consensus design point. SpaceX, Starcloud, K2 Space (Giga-Class at 110 kW), and independent analyses all converge on 100–130 kW_IT as the near-term target. This validates our model's parameterization per kW_IT — the economics are driven by mass-per-kW_IT regardless of whether the satellite houses 8 GPUs or 72 [spacex-ai-sat-mini-spacenews.1, k2-gravitas-orbital-today.1].

Time-Invariance Note

The base subsystem parameters (solar specific power, radiator specific power, compute hardware mass) are held constant across all years because they reflect hardware design choices rather than market prices. Total satellite mass varies by year only due to the time-varying cost-optimal battery sizing — as launch costs fall, heavier batteries become cost-effective. While subsystem technologies will improve over time, the improvement timelines are uncertain and are better captured in the scenario ranges than as a time series.