Orbital AI Data Center: Solar and Radiator Sizing

Size solar arrays, eclipse batteries and radiators for an orbital AI data center. Compare temperature, orbit geometry, launch mass and launch-price assumptions.

#AI infrastructure #eclipse batteries #launch mass #orbital data centers #radiators #solar arrays #Stefan–Boltzmann law #thermal sizing

This orbital data center calculator turns a continuous GPU and IT electrical load into the solar array, eclipse battery, radiator area and launch mass needed to support it. Engineers and students can use it to compare early concepts for data centers in space and see which assumptions dominate the result.

Start with the total IT load and a bus power multiplier for pumps, avionics, communications and conversion losses. Choose a circular Earth orbit and the Sun’s beta angle above its plane. Zero beta gives the longest eclipse at that altitude; a sufficiently large beta eliminates eclipse in this simplified geometry. Beta changes with season and orbital orientation, so a favorable snapshot does not establish year-round sunlight.

The array supplies the operating load and recharges the battery during every sunlit interval. Enter solar cell efficiency and a retained output factor covering temperature, packing, pointing and end-of-life degradation. The battery result is rated energy capacity after discharge efficiency and usable depth of discharge. A zero-eclipse case returns zero eclipse storage; mission operations may still require reserve batteries.

For cooling, enter the mean radiator surface temperature, infrared emissivity and absorbed environmental heat flux. The default 76.85 °C is 350 K. The background defaults to deep space; add Earth infrared, reflected sunlight and direct solar absorption through the absorbed-flux input using your own attitude and view-factor analysis. Avoid counting the same radiation in both inputs.

The space radiator calculator reports total emitting surface area and panel footprint separately. Two useful faces halve the footprint for a given emitting area only when their average net rejection is equivalent. Radiator mass per m² refers to footprint. The often quoted 1,200 m² per MW at 350 K is approximately the ideal blackbody emitting area with negligible incoming radiation, before real emissivity, overhead power and sizing allowances.

The temperature curve shows the strong fourth-power cooling tradeoff. The beta-angle curve shows why eclipse storage cannot be inferred from altitude alone. The drawing distinguishes solar panels, heat-rejection panels and the orbit’s shadow arc; the spacecraft is schematic.

For orbital data center feasibility studies, replace the illustrative assembly masses and launch price with mission estimates. The price input is an assumption in USD per kg dated 27 October 2026, not a launch quotation. The result covers launch transport only, excluding hardware purchase, development, orbital assembly, maintenance and operations.

This is preliminary steady-state sizing. It does not qualify processor cooling, battery cycle life, radiation shielding, collision avoidance, deployment dynamics, radiator survivability or disposal. GAO’s April 2026 assessment identifies major unresolved engineering and economic barriers; favorable area arithmetic alone cannot establish a viable mission.

How it is calculated

The sizing equations are derived here from orbital geometry, conservation of electrical energy and the Stefan–Boltzmann law. NASA's Small Spacecraft Technology State of the Art (2026), chapters 3 and 7, provides power-system and thermal-control context. The reference cases test published numbers independently of this derivation.

For a circular orbit, r=RE+hr=R_E+h and to=2πr3/μEt_o=2\pi\sqrt{r^3/\mu_E}. In a cylindrical Earth shadow, the eclipse half angle is θe=cos⁡−1[1−(RE/r)2/cos⁡β]\theta_e=\cos^{-1}[\sqrt{1-(R_E/r)^2}/\cos\beta] while β<sin⁡−1(RE/r)\beta<\sin^{-1}(R_E/r), and zero otherwise. Then fe=θe/πf_e=\theta_e/\pi with angles in radians. The implemented degree form uses a half-turn of 180°. Periods and eclipse times are converted explicitly before computing energy. See ESA Navipedia, Satellite Eclipses; Earth radius and irradiance come from NASA's Earth Fact Sheet and μE\mu_E from JPL DE440.

The continuous bus load is Pb=fauxPITP_b=f_{aux}P_{IT}. Eclipse delivery is Ee=PbteE_e=P_b t_e. Stored energy removed is Es=Ee/ηdE_s=E_e/\eta_d and rated pack capacity is Epack=Es/DE_{pack}=E_s/D. Sunlight charging needs Pch=Es/(ηcts)P_{ch}=E_s/(\eta_c t_s), so the array delivers PSA=Pb+PchP_{SA}=P_b+P_{ch}. Array area is ASA=PSA/(SηPVfarray)A_{SA}=P_{SA}/(S\eta_{PV}f_{array}), using consistent W and W/m².

Peak internal waste heat is the greater of Pb+(1−ηc)PchP_b+(1-\eta_c)P_{ch} in sunlight and Pb/ηdP_b/\eta_d during eclipse. With no eclipse the latter mode is absent. All IT, auxiliary and battery losses become heat. Array cell inefficiency heats the solar panels, which reject it separately.

For a uniform gray radiator, qnet=ϵσ(Tr4−Ts4)−qabsq_{net}=\epsilon\sigma(T_r^4-T_s^4)-q_{abs}. Temperatures are converted from °C to K. With thermal factor FTF_T, total emitting area is Ae=FTQpeak/qnetA_e=F_TQ_{peak}/q_{net}. Panel footprint is Ar=Ae/nfacesA_r=A_e/n_{faces}. Nonpositive net rejection is blocked by a condition, never presented as a feasible size. As a derived ideal limit, σ(350 K)4≈851\sigma(350\,K)^4\approx851 W/m², giving about 1,175 m² of blackbody emitting area per MW. The familiar 1,200 m²/MW shorthand excludes emissivity, absorbed radiation, bus overhead and allowances.

Assembly masses are footprint or array area times their respective kg/m² values; battery mass is rated Wh divided by pack Wh/kg. Other spacecraft mass is entered separately. Their sum times the mass factor gives estimated launch mass. Launch-only cost is mass in kg times the assumed USD/kg rate. Monetary amounts stay in USD under the unit switch.

Assumptions

  • Continuous IT load in both sunlight and eclipse; no eclipse load shedding. Constant Sun irradiance and beta angle represent one design condition, not an annual trajectory.
  • Uniform radiator temperature and emissivity. Both faces, if enabled, have the entered average environmental flux and net heat rejection. No mutual shading, Earth occultation of radiator view, fin efficiency or finite thermal conductance is calculated.
  • The 3 K background is a deep-space reference. The separate 100 W/m² absorbed flux is an illustrative hot-environment allowance; replace it with a mission attitude and view-factor analysis. Do not include the same Earth IR contribution in both the background temperature and absorbed flux.
  • Retained array output includes packing, pointing, operating temperature and end-of-life loss. Onboard load conversion heat belongs in the bus multiplier. Solar panel waste heat is not assigned to the server radiator.
  • The 30% cell efficiency is consistent with NASA's 2026 power chapter. The 0.80 array retention, 0.95 charge and discharge efficiencies, 0.30 usable battery fraction, 180 Wh/kg pack value, 5 kg/m² radiator, 3 kg/m² array, 10,000 kg other mass and sizing factors are editable scenario assumptions, not a qualified design or supplier specifications.
  • NASA/Boeing ISS ATCS Overview, attached PDF page 3, gives a PVR mass of 740.7 kg and deployed dimensions 3.12 × 13.6 m. That is approximately 17.5 kg/m² of footprint, substantially above the optimistic default radiator assembly density. Its different temperatures, geometry and environment prevent using its rated heat load as a direct 350 K radiator benchmark.
  • Launch price is an illustrative input in USD/kg dated 27 October 2026. No provider tariff, vehicle capacity, discount or future price forecast is assumed.

Where it applies

Use for preliminary comparisons of circular Earth orbits from 200 to 2,000 km. The cylindrical shadow ignores penumbra, finite Sun size, atmospheric refraction, Earth oblateness, eccentricity and seasonal orbit evolution. The thermal calculation is steady-state peak sizing with no thermal storage credit. The background model is an equivalent radiative environment, not a local air temperature.

Processor junction temperature must exceed the radiator temperature by the entire cold-plate, coolant and panel temperature drop. Raising the surface temperature to reduce area does not establish that the computing hardware can tolerate it. Battery mass is energy-limited only; verify power capability, charge rate, cycle life, thermal control, end-of-life capacity, eclipse contingency and operational reserves separately.

Launch mass is only as complete as the entered assembly masses and other spacecraft allowance. There is no calculation of shielding, redundancy, propellant, structural modes, deployment, micrometeoroid damage, repair, data throughput, station keeping or disposal. Launch cost excludes manufacture, integration, development, orbital assembly and operations.

GAO-26-109012 (28 April 2026) describes unproven large-scale cooling and power systems. The FCC notice of 4 February 2026 accepted SpaceX's application for filing and comment; that notice is not an operating authorization or engineering validation. No regulatory-compliance determination is made.

References

  • [NASA Small Spacecraft Technology State of the Art (2026), chapter 7: Thermal Control](https://www.nasa.gov/smallsat-institute/sst-soa/thermal-control/). Sections 7.1 and 7.2 discuss radiative balance, environmental heating and radiator design.
  • [NASA Small Spacecraft Technology State of the Art (2026), chapter 3: Power](https://www.nasa.gov/wp-content/uploads/2026/05/3-soa-power-2026-final.pdf). Table 3-1, attached PDF page 6 (printed p.31), supplies the independent 403 W/m² cell benchmark.
  • [NASA/Boeing ISS Active Thermal Control System Overview](https://www.nasa.gov/wp-content/uploads/2021/02/473486main_iss_atcs_overview.pdf). Attached PDF pp.3 and 5: PVR dimensions/mass and 70 kW total EATCS capacity; context, not a calibration of this model.
  • [OpenStax College Physics 2e, §14.7 Radiation, Example 14.9](https://openstax.org/books/college-physics-2e/pages/14-7-radiation). Published net radiative heat-transfer example, including its rounded Kelvin temperatures.
  • [CEOS CARBONITE-2 Mission Summary](https://database.eohandbook.com/database/missionsummary.aspx?missionID=855). Independent orbit data: 500 km altitude and 94.6 min period.
  • [NASA Earth Fact Sheet](https://nssdc.gsfc.nasa.gov/planetary/factsheet/earthfact.html). Mean radius 6371 km and solar irradiance 1361 W/m².
  • [JPL Astrodynamic Parameters](https://ssd.jpl.nasa.gov/astro_par.html). DE440 Earth GM: 398600.435507 km³/s².
  • [ESA Navipedia: Satellite Eclipses](https://gssc.esa.int/navipedia/index.php/Satellite_Eclipses). Cylindrical Earth-shadow condition; eclipse fractions here are derived from that geometry.
  • [GAO-26-109012: Science & Tech Spotlight—Data Centers in Space, 28 April 2026](https://www.gao.gov/products/gao-26-109012). Mission feasibility context.
  • [FCC DA 26-113, 4 February 2026](https://docs.fcc.gov/public/attachments/DA-26-113A1.pdf). Public notice accepting the SpaceX orbital data center application for filing.
  • [IEEE Spectrum: How Stupid Would It Be to Put Data Centers in Space?](https://spectrum.ieee.org/orbital-data-centers). Background discussion; its concept arithmetic is not a validation case.
  • [AZUR SPACE General Information, 8 April 2025, p.7](https://www.azurspace.com/media/uploads/file_links/file/bdb_00010890-01-00_generalinformationazur.pdf). Bare-cell efficiencies on both 1353 and 1367 W/m² AM0 rating bases; supports the independent NASA table benchmark conditions.

Inputs

InputSymbolUnitDefault
GPU and IT electrical power
Continuous IT load, including CPUs, memory and networking assigned to the computing payload.
P_gpukW1000
Total bus power / IT power
Include pumps, avionics, communications and downstream load-converter losses. Battery charge/discharge losses are calculated separately; do not count them here. This is a sizing assumption, not a measured orbital PUE.
f_aux1.2
Circular orbit altitudeh_orbkm550
Sun angle above the orbit plane (β)
0° gives the longest eclipse at this altitude. Use a mission's worst seasonal beta, not its orbital inclination. 90° is continuously sunlit in this model.
beta_sdeg0
Sunlight irradiance at the array
NASA Earth Fact Sheet: 1361 W/m² near 1 AU. Consider the annual minimum for mission sizing.
S_sunW/m^21361
Solar cell efficiency
Fraction, e.g. 0.30 = 30%. NASA's 2026 power chapter describes nominal 30% multijunction cells.
eta_pv0.3
Retained array output factor
Combined cell packing, hot-cell performance, end-of-life degradation and cosine pointing factor. Relative to cell efficiency; include each loss once. Include onboard converter heat in the bus power factor.
f_array0.8
Battery charge efficiency
Bus electrical energy to stored energy. Charge and discharge efficiencies multiply to give round-trip efficiency.
eta_ch0.95
Battery discharge efficiency
Stored energy to delivered bus energy; includes battery-side conversion losses.
eta_dis0.95
Usable depth of discharge
Mission end-of-life usable fraction per orbit. Default 0.30 limits depth per cycle; it is a scenario, not proof of battery life. Use mission testing and include degradation in pack specific energy.
d_use0.3
Battery pack specific energy
Rated end-of-life pack energy per mass, including enclosure and battery electronics; not cell-only specific energy.
e_packWh/kg180
Mean radiator surface temperature
76.85 °C = 350 K. Use the mean emitting surface temperature, below the coolant and processor temperatures. Higher values require compatible hardware.
T_rad°C76.85
Effective background radiation temperature
Default 3 K represents deep space. Earth IR and sunlight enter the separate absorbed heat flux. Use a warmer equivalent background only if that radiation is removed from the absorbed-flux input.
T_sink°C-270.15
Infrared emissivity
Use end-of-life hemispherical emissivity of the emitting faces.
eps_rad0.9
Additional absorbed heat per emitting area
Already-absorbed, worst-case solar, albedo and Earth IR flux, averaged over all emitting faces. Include view factors and absorptivity here; the calculator does not infer them from the orbit.
q_absW/m^2100
Effective emitting faces per panel
Two assumes both faces have the same average net rejection, unobstructed. Choose one if the reverse face is blocked or unusable.
n_face2
Radiator area sizing factor
Applied to peak internal heat. Allowance for uncertainty and temperature nonuniformity; not a model of redundant loops or failure tolerance.
f_thermal1.15
Radiator assembly mass per panel area
Per one-sided panel footprint, including coolant, tubes, deployment and supports allocated to the radiator. 5 kg/m² is an optimistic scenario. NASA's ISS PVR is about 17.5 kg/m² of deployed footprint.
rho_radkg/m^25
Solar array assembly mass per array area
Include substrate, wiring, deployment and allocated support mass. 3 kg/m² is a scenario assumption.
rho_sakg/m^23
Computing hardware and remaining spacecraft mass
Include payload, bus, shielding, communications, propulsion, tanks and hardware not included in the array, radiator or pack estimates.
m_otherkg10000
Launch mass allowance factor
Multiplies the sum of specified component masses. Explicit mission hardware should go in its mass input; this factor covers remaining uncertainty.
f_mass1.2
Assumed launch price (USD per kg)
User scenario in USD/kg, not a provider quote. Default is an illustrative assumption dated 27 October 2026. Currency units are written in labels; prices do not change with the metric/US switch.
c_launch2000

Results

With the default inputs above, the calc gives:

ResultValue
Solar array area6,088 m²
Radiator panel footprint1,091 m²
Total radiator emitting surface2,182 m²
Minimum rated eclipse battery2,494.7 kWh
Array power during sunlight1,988.4 kW
Eclipse duration35.55 min
Estimated launch mass57,092 kg
Launch-only cost (million USD)114.18
Net radiator heat rejection665.8 W/m²
Radiator design heat load1,452.6 kW
Solar assembly mass18,263 kg
Radiator assembly mass5,454 kg
Eclipse battery pack mass13,860 kg
Orbital period95.5 min

Checked against sources

Each case below was calculated by this calc and agrees with the value from its source (within the tolerance where one is shown).

OpenStax 14.9: net radiation from a 33 °C surface

Source: OpenStax College Physics 2e §14.7, Example 14.9: 1.50 m², emissivity 0.97, skin 33 °C and room 22 °C gives 99 W net outward radiation. Thus published net flux is 99/1.50 = 66 W/m². The book rounds to 306 K and 295 K; the calculator uses 306.15 K and 295.15 K. Absolute tolerances of 0.5 W/m² in flux and 0.01 m² in area cover this rounding; no coefficient is fitted. https://openstax.org/books/college-physics-2e/pages/14-7-radiation

Inputs: GPU and IT electrical power = 0.099 kW; Total bus power / IT power = 1; Sun angle above the orbit plane (β) = 90 deg; Mean radiator surface temperature = 33 °C; Effective background radiation temperature = 22 °C; Infrared emissivity = 0.97; Additional absorbed heat per emitting area = 0 W/m²; Effective emitting faces per panel = 1; Radiator area sizing factor = 1; the others as above.

ResultSourceThis calc
Net radiator heat rejection66 ± 0.5 W/m²65.791 W/m²
Total radiator emitting surface1.5 ± 0.01 m²1.505 m²

CEOS CARBONITE-2: 500 km orbit, 94.6 min period

Source: CEOS Mission, Instruments and Measurements Database, CARBONITE-2 mission: orbit altitude 500 km and period 94.6 minutes. Compare a circular spherical-Earth approximation; 0.2 min absolute tolerance allows the published rounded orbit and Earth-radius convention. https://database.eohandbook.com/database/missionsummary.aspx?missionID=855

Inputs: Circular orbit altitude = 500 km; the others as above.

ResultSourceThis calc
Orbital period94.6 ± 0.2 min94.469 min

NASA Table 3-1: AZUR 3G30-Adv cell output

Source: NASA 2026 Small Spacecraft Technology State of the Art, Power §3.2.1, Table 3-1: AZUR Space 3G30-Adv has BOL efficiency 29.5% and Pmp 403 W/m². At the conventional 1367 W/m² AM0 cell-test irradiance these rounded values imply approximately one m² for 403 W. This verifies the bare-cell relation; it does not validate mission derating or an orbital array design. https://www.nasa.gov/wp-content/uploads/2026/05/3-soa-power-2026-final.pdf Absolute tolerances: 1 W/m² and 0.003 m². The 1367 W/m² rating basis and 29.5% bare-cell efficiency for the 100 mm, two-cells-per-wafer 3G30-Advanced configuration are corroborated by AZUR SPACE General Information, issue 8 April 2025, p.7: https://www.azurspace.com/media/uploads/file_links/file/bdb_00010890-01-00_generalinformationazur.pdf .

Inputs: GPU and IT electrical power = 0.403 kW; Total bus power / IT power = 1; Sun angle above the orbit plane (β) = 90 deg; Sunlight irradiance at the array = 1,367 W/m²; Solar cell efficiency = 0.295; Retained array output factor = 1; the others as above.

ResultSourceThis calc
Delivered solar power per array area403 ± 1 W/m²403.3 W/m²
Solar array area1 ± 0.003 m²0.9993 m²

Related Resources

Responses

Your email address will not be published. Required fields are marked *

L ading...