arxiv: 2604.27197 · v1 · submitted 2026-04-29 · ⚛️ physics.gen-ph · cs.SY· eess.SY

Recognition: unknown

Orbital Data Centers: Spacecraft Constraints and Economic Viability

Slava G. Turyshev

Pith reviewed 2026-05-07 09:17 UTC · model grok-4.3

classification ⚛️ physics.gen-ph cs.SYeess.SY

keywords orbital data centersspacecraft mass budgetphotovoltaic arearadiator sizingspace-to-ground communicationseconomic viabilitycompute power deliverylaunch cost constraints

0 comments

The pith

Orbital data centers require combined launch and build costs several times below current benchmarks to deliver competitive compute power.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper derives the full set of simultaneous engineering and economic constraints that an orbital data center must satisfy to outperform terrestrial infrastructure. Solar generation, eclipse storage, heat rejection, space-to-ground data rates, utilization, and mission lifetime all interact through the delivered IT power and the mass per kilowatt of that power. A reader cares because the resulting cost window shows that solar flux alone does not make space compute viable; the supporting hardware must reach price points far below today's launch and spacecraft costs.

Core claim

For a 1 MW IT-power anchor in high sunlight the design needs 5.64 thousand square meters of beginning-of-life photovoltaic area and 2.5 thousand square meters of radiator area. Photovoltaic, storage, and radiator mass alone reach 29.4 kg per kW; fixed spacecraft mass raises the total to 34-59 kg per kW. At roughly 40 kg per kW a terrestrial benchmark of 10-40 thousand dollars per kW allows only 250-1000 dollars per kg for the combined launch-plus-build cost before communications, operations, utilization, and lifetime terms are included. That allowance lies 3.4 to 13.5 times below the current public Falcon 9 dedicated low-Earth-orbit launch price alone.

What carries the argument

The cluster-level competitiveness conditions expressed through delivered IT power P_IT, mass per delivered kilowatt m_kW, communication intensity Gamma equals data sent to ground divided by IT energy, sustained communication ceiling Gamma_max, effective utilization U_eff, and lifetime penalty Pi_life.

If this is right

Space-native preprocessing and communications-integrated edge compute become credible early regimes.
Terrestrial-user general compute only closes for low Earth-coupled communication intensity, high effective utilization, long delivered lifetime, and very low combined launch-plus-build cost.
The beginning-of-life photovoltaic area and radiator area both scale linearly with the target IT power.
Fixed spacecraft mass raises the total mass budget above the pure power-system value of 29.4 kg per kW.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

The same mass-budget and communication-intensity logic applies to any power-intensive orbital platform, not only data centers.
If launch and build costs reach the required range, orbital centers could serve specific low-latency or high-bandwidth workloads that terrestrial sites handle poorly.
Direct measurement of sustained space-to-ground data rates per unit energy would test one of the parameters that most strongly affects the economic window.

Load-bearing premise

The conclusions rest on chosen values for communication intensity, effective utilization, and lifetime penalty that are not independently measured from real orbital operations.

What would settle it

A complete orbital compute spacecraft whose total launch-plus-build cost per kilogram falls inside or below the 250-1000 dollar window, while meeting the assumed sustained communication rate and utilization, would support the viability claim; measured costs remaining above that window would falsify it under the paper's parameter ranges.

Figures

Figures reproduced from arXiv: 2604.27197 by Slava G. Turyshev.

**Figure 1.** Figure 1: FIG. 1. Reference distributed ODC architecture used throughout the paper. Compute nodes are linked by optical inter-satellite view at source ↗

**Figure 2.** Figure 2: FIG. 2. Schematic overview of the competitiveness model. Delivered IT power sets total platform power through view at source ↗

**Figure 3.** Figure 3: FIG. 3. First-order technology levers in the subsystem floor. The photovoltaic term decreases nearly as view at source ↗

**Figure 4.** Figure 4: FIG. 4. Constellation granularity trade for fixed total cluster power. Reducing node power lowers the fraction of capacity lost in view at source ↗

**Figure 5.** Figure 5: FIG. 5. Benchmark break-even boundary. The three markers at view at source ↗

read the original abstract

Orbital data centers are being evaluated as solar-powered compute constellations and relay-integrated processing platforms. Their feasibility is not set by orbital solar flux alone, but by simultaneous closure of photovoltaic generation, eclipse recharge, radiative heat rejection, sustained space-to-ground communications, utilization, replacement cadence, and delivered compute-years over finite mission life. This paper derives necessary cluster-level competitiveness conditions using delivered information-technology (IT) electrical power $P_{\rm IT}$, deployed mass per delivered IT power $m_{\rm kW}$ in kg/kW, communication intensity $\Gamma=D_{\rm sg}/E_{\rm IT}$, sustained communication ceiling $\Gamma_{\max}$, effective utilization $U_{\rm eff}$, and lifetime penalty $\Pi_{\rm life}$. For a representative $P_{\rm IT}$=1 MW high-sunlight anchor, the base case gives beginning-of-life photovoltaic area $A^{\rm BOL}_{\rm PV}=5.64 \times 10^3 {\rm m}^2$, radiator area $A_{\rm rad}=2.50 \times 10^3 {\rm m^2}$, and 29.4 kg/kW for photovoltaic, storage, and radiator mass; fixed spacecraft mass raises the total to 34-59 kg/kW. At m_kW ~ 40 kg/kW, a terrestrial infrastructure benchmark of 10-40 k\$/kW allows only 250-1000 \$/kg for the combined launch and spacecraft-build cost before space-to-ground communications, operations, utilization, and lifetime terms are included. That allowance is 3.4-13.5 times below the current public Falcon 9 dedicated low-Earth-orbit launch-price benchmark alone, before spacecraft build is included. Space-native preprocessing and communications-integrated edge compute are credible early regimes; terrestrial-user general compute closes only for low Earth-coupled communication intensity, high effective utilization, long delivered lifetime, and very low combined launch-plus-build cost.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Desk Editor's Note private letter to a colleague

The paper derives explicit engineering numbers for a 1 MW orbital data center but its economic gap claim depends on chosen ranges for communication intensity, utilization, and lifetime rather than independently derived bounds.

read the letter

The key point here is that orbital data centers would require launch and spacecraft costs several times lower than today's Falcon 9 prices to make economic sense against ground-based compute, but only if the assumptions on data transfer rates, utilization, and mission life hold up. The paper does a solid job pulling together the main constraints. It starts from a 1 MW IT power target and works out the solar array size at 5640 square meters, radiator at 2500 square meters, and subsystem mass around 29 kg per kW, pushing total to 34-59 kg/kW once fixed mass is added. These come from standard equations for power generation, eclipse handling, and heat rejection. The competitiveness model then folds in the communication intensity Gamma, effective utilization, and lifetime factor to set the allowed cost per kg. That integrated framing is new enough for this application. It turns the usual separate discussions of power budgets and launch economics into one set of thresholds. The main limitation is that the final cost gap of 3.4 to 13.5 times below current launch prices depends directly on the chosen ranges for those operational parameters. The abstract does not show independent derivations or citations for realistic sustained Gamma or U_eff values. If space-to-ground links cannot sustain the assumed data volumes or if utilization drops, the allowed cost window moves and the shortfall against Falcon 9 changes by the same factor. The engineering numbers look more robust because they rest on physical constants. This work is aimed at teams planning space infrastructure or evaluating new compute locations. A reader interested in quantitative feasibility studies for orbital systems will find the explicit targets helpful for setting requirements. I would send it to peer review. The calculations are grounded and the claims are specific enough to be checked or refined. The parameter choices need more support, but that is fixable.

Referee Report

1 major / 2 minor

Summary. The manuscript derives spacecraft-level engineering constraints for orbital data centers using standard orbital power and thermal balance equations. For a representative 1 MW IT power anchor in high-sunlight orbit it reports explicit values A_BOL_PV = 5.64 × 10^3 m², A_rad = 2.50 × 10^3 m² and 29.4 kg/kW subsystem mass (total 34–59 kg/kW once fixed spacecraft mass is added). It then introduces competitiveness conditions that fold in communication intensity Γ = D_sg/E_IT, effective utilization U_eff, lifetime penalty Π_life and a sustained-communication ceiling Γ_max to obtain an allowed combined launch-plus-build cost of only 250–1000 $/kg at ~40 kg/kW. This figure is stated to lie 3.4–13.5× below the public Falcon 9 dedicated LEO launch-price benchmark, implying that general terrestrial-user compute is not viable while space-native preprocessing or edge-compute regimes remain credible.

Significance. If the engineering stack and the framing of the cost gap hold, the work supplies concrete, equation-derived benchmarks (PV area, radiator area, mass per delivered kW) that can serve as reference points for subsequent studies. The explicit numerical outputs grounded in orbital flux, eclipse fraction and radiative rejection equations constitute a reproducible starting point, even though the final economic thresholds remain conditional on operational parameters.

major comments (1)

[competitiveness conditions] Competitiveness conditions (abstract and associated derivation): the headline claim that the allowed combined cost is 3.4–13.5× below the Falcon 9 benchmark is obtained only after the delivered compute-years are scaled by the chosen ranges for Γ, U_eff and Π_life. These three quantities are introduced as free parameters without external bounds, literature citations, or a sensitivity table showing how the allowed $/kg changes when each is varied by a factor of two. Because the gap scales linearly with these factors, the quantitative viability conclusion is load-bearing on their selection rather than on the engineering stack alone.

minor comments (2)

[abstract] Notation: Γ, U_eff and Π_life appear in the abstract and competitiveness conditions without immediate units or a compact definition table; adding a short nomenclature box would improve readability.
[engineering derivations] The manuscript states explicit numerical results (A_BOL_PV, A_rad, 29.4 kg/kW) but does not display the intermediate algebraic steps or the precise efficiency and eclipse-fraction values used; a short appendix or inline derivation box would allow independent verification.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for the careful and constructive review. The major comment concerns the dependence of the competitiveness conclusions on the operational parameters. We address this point below and will revise the manuscript to improve transparency.

read point-by-point responses

Referee: Competitiveness conditions (abstract and associated derivation): the headline claim that the allowed combined cost is 3.4–13.5× below the Falcon 9 benchmark is obtained only after the delivered compute-years are scaled by the chosen ranges for Γ, U_eff and Π_life. These three quantities are introduced as free parameters without external bounds, literature citations, or a sensitivity table showing how the allowed $/kg changes when each is varied by a factor of two. Because the gap scales linearly with these factors, the quantitative viability conclusion is load-bearing on their selection rather than on the engineering stack alone.

Authors: We agree that the factor of 3.4–13.5× is obtained by scaling with the chosen ranges of Γ, U_eff, and Π_life, and that these are operational parameters rather than fixed engineering quantities. The manuscript frames them explicitly as variables that delineate use-case regimes (low-communication edge compute versus general terrestrial workloads), with the quoted factor corresponding to the subset of values that would render space-native preprocessing viable. The core engineering results (PV area, radiator area, and 34–59 kg/kW total mass) are independent of these parameters and derived solely from orbital flux, eclipse, and thermal-balance equations. To address the absence of a sensitivity table and external grounding, the revised manuscript will include a new table showing the allowed launch-plus-build cost for ± factor-of-two excursions in each parameter around the base values used in the abstract. We will also expand the discussion section to supply example bounds drawn from terrestrial edge-computing literature (e.g., utilization rates for content-delivery and IoT preprocessing workloads) together with appropriate citations. These additions will make the load-bearing assumptions explicit without altering the underlying engineering stack. revision: yes

Circularity Check

0 steps flagged

No circularity; cost-gap claim follows from independent physical constants and external market benchmarks

full rationale

The derivation begins with physical inputs (solar flux, eclipse duration, thermal rejection efficiency) to compute BOL photovoltaic area, radiator area, and subsystem masses for a 1 MW P_IT anchor, producing m_kW = 29.4 kg/kW (PV/storage/radiator) or 34-59 kg/kW total. The allowed combined launch-plus-build cost is obtained by dividing the independent terrestrial benchmark (10-40 k$/kW) by this m_kW, then comparing the resulting 250-1000 $/kg range to the separate Falcon 9 public price benchmark. The parameters Γ, U_eff, and Π_life enter only the broader competitiveness conditions and are explicitly excluded from the quoted allowance and shortfall factor. No equations reduce to their own inputs by construction, no parameters are fitted then renamed as predictions, and no self-citations or imported uniqueness theorems appear in the chain.

Axiom & Free-Parameter Ledger

3 free parameters · 3 axioms · 0 invented entities

The central claims rest on standard orbital engineering assumptions for solar generation and thermal rejection plus external economic benchmarks; no new physical entities are postulated.

free parameters (3)

P_IT
Representative 1 MW high-sunlight anchor power level selected for the base-case calculation of areas and masses.
terrestrial cost benchmark 10-40 k$/kW
External market figure used to back-calculate the allowable launch-plus-build cost per kg.
Falcon 9 launch price benchmark
Current public dedicated LEO price used to quantify the gap to viability.

axioms (3)

domain assumption Orbital solar flux and beginning-of-life photovoltaic efficiency yield A_BOL_PV = 5.64e3 m2 for 1 MW IT power
Invoked to obtain the photovoltaic area in the base case.
domain assumption Radiative heat rejection model produces A_rad = 2.50e3 m2
Standard thermal balance used for radiator sizing.
domain assumption Subsystem mass of 29.4 kg/kW for PV, storage, and radiator plus fixed spacecraft mass totals 34-59 kg/kW
Mass budget closure assumed without detailed breakdown in the abstract.

pith-pipeline@v0.9.0 · 5659 in / 1881 out tokens · 76418 ms · 2026-05-07T09:17:20.481452+00:00 · methodology

discussion (0)

Reference graph

Works this paper leans on

39 extracted references · 4 canonical work pages · 1 internal anchor

[1]

Shehabi, S

A. Shehabi, S. J. Smith, A. Hubbard, A. Newkirk, N. Lei, M. A. B. Siddik, B. Holecek, J. Koomey, E. Masanet, and D. Sartor,2024 United States Data Center Energy Usage Report, Tech. Rep. LBNL-2001637 (Lawrence Berkeley National Laboratory, 2024)

2024
[2]

North American Electric Reliability Corporation,2024 Long-Term Reliability Assessment, Tech. Rep. (NERC, 2024)

2024
[3]

International Energy Agency,Energy and AI, Tech. Rep. (International Energy Agency, 2025)

2025
[4]

Towards a future space-based, highly scalable ai infrastructure system design,

B. Ag¨ uera y Arcas, T. Beals, M. Biggs, J. V. Bloom, T. Fischbacher, K. Gromov, U. K¨ oster, R. Pravahan, and J. Manyika, Towards a Future Space-Based, Highly Scalable AI Infrastructure System Design (2025), arXiv:2511.19468 [cs.DC]

work page arXiv 2025
[5]

Bargatin, D

I. Bargatin, D. Jin, Z. Alansari, and J. R. Raney, Tether-Based Architecture for Solar-Powered Orbital AI Data Centers (2025), arXiv:2512.09044 [astro-ph.IM]

work page arXiv 2025
[6]

J. F. Gutierrezet al.,Data Centres in Space: Orbital Backbone of the Second Digital Era?, Tech. Rep. ESPI Report 98 (European Space Policy Institute, 2025)

2025
[7]

Technical University of Munich, Space-Based Data Centres: From Vision to Viable Infrastructure for Europe (2025), news release, November 19, 2025

2025
[8]

National Aeronautics and Space Administration, Communications Services Project (2026)

2026
[9]

SpaceX,Starlink 2025 Progress Report, Tech. Rep. (SpaceX, 2026)

2025
[10]

Federal Communications Commission, Space Bureau, Space Bureau Accepts for Filing SpaceX’s Application for Orbital Data Centers (2026), Public Notice, ICFS File No

U.S. Federal Communications Commission, Space Bureau, Space Bureau Accepts for Filing SpaceX’s Application for Orbital Data Centers (2026), Public Notice, ICFS File No. SAT-LOA-20260108-00016

2026
[11]

NVIDIA, NVIDIA Launches Space Computing, Rocketing AI Into Orbit (2026), Newsroom release, March 16, 2026

2026
[12]

Chabra, C

O. Chabra, C. Li, K. Hsieh, S. Segarra, B. Arzani, P. Olsen, and R. Chandra, OrbitalBrain: A Distributed Framework for Training ML Models in Space, in1st New Ideas in Networked Systems (NINeS 2026), Open Access Series in Informatics (OASIcs), Vol. 139 (Schloss Dagstuhl – Leibniz-Zentrum fuer Informatik, Dagstuhl, Germany, 2026) pp. 5:1–5:32

2026
[13]

Thummala and G

A. Thummala and G. Falco, When to Compute in Space (2025), arXiv:2512.17054 [cs.DC]

work page arXiv 2025
[14]

A. A. Periola and M. O. Kolawole, Space-Based Data Centres: A Paradigm for Data Processing and Scientific Investigations, Wireless Personal Communications107, 95 (2019)

2019
[15]

Kopp and J

G. Kopp and J. L. Lean, A New, Lower Value of Total Solar Irradiance: Evidence and Climate Significance, Geophys. Res. Letters38, L01706 (2011)

2011
[16]

National Aeronautics and Space Administration,Small Spacecraft Technology State of the Art: Power Chapter, Tech. Rep. (NASA, 2024). 29

2024
[17]

A. J. Juhasz,Design Considerations for Lightweight Space Radiators Based on Fabrication and Test Experience with a Carbon-Carbon Composite Prototype Heat Pipe, Tech. Rep. NASA/TP-1998-207427/REV1 (NASA Lewis Research Center, 2002)

1998
[18]

D. G. Gilmore, ed.,Spacecraft Thermal Control Handbook. Volume I: Fundamental Technologies, 2nd ed. (American Institute of Aeronautics and Astronautics and The Aerospace Press, Reston, VA, 2002)

2002
[19]

C. M. Schieler, K. M. Riesing, B. C. Bilyeu, J. S. Chang, A. S. Garg, N. J. Gilbert, A. J. Horvath, R. S. Reeve, B. S. Robinson, J. P. Wang, S. Piazzolla, W. T. Roberts, J. M. Kovalik, and B. Keer, On-orbit demonstration of 200-Gbps laser communication downlink from the TBIRD CubeSat, inFree-Space Laser Communications XXXV, Vol. 12413, edited by H. Hemmat...

2023
[20]

Wang, TeraByte InfraRed Delivery (TBIRD): Delivering 200 Gbps from Space to Ground (2025), presentation, dated April 1, 2025

J. Wang, TeraByte InfraRed Delivery (TBIRD): Delivering 200 Gbps from Space to Ground (2025), presentation, dated April 1, 2025

2025
[21]

D. A. Vallado,Fundamentals of Astrodynamics and Applications, 4th ed. (Microcosm Press, Hawthorne, CA, 2013)

2013
[22]

Vavrin, A

A. Vavrin, A. Manis, J. Seago,et al.,NASA Orbital Debris Engineering Model (ORDEM) 3.1 – Software User Guide, Tech. Rep. NASA/TP-2019-220448 (NASA Johnson Space Center, Orbital Debris Program Office, 2019)

2019
[23]

National Aeronautics and Space Administration,Process for Limiting Orbital Debris, Tech. Rep. (NASA, 2019)

2019
[24]

Hemmati, ed.,Deep Space Optical Communications(Wiley-Interscience, 2006)

H. Hemmati, ed.,Deep Space Optical Communications(Wiley-Interscience, 2006)

2006
[25]

D. J. Israel, B. L. Edwards, R. L. Butler, J. D. Moores, S. Piazzolla, N. du Toit, and L. Braatz, Early results from NASA’s laser communications relay demonstration (LCRD) experiment program, inOptics Damage and Materials Processing by EUV/X-ray Radiation (XDam8), Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, Vol. 12413, edi...

2023
[26]

M. A. Felix, W. S. Slater, D. C. Landauer, R. E. Pinson, and B. B. W. Rutherford, Total Ionizing Dose Radiation Testing of NVIDIA Jetson Orin NX System on Module, in2024 IEEE Space Computing Conference (SCC)(2024) pp. 116–121

2024
[27]

Rodriguez-Ferrandez, E

I. Rodriguez-Ferrandez, E. Rufart Blasco, L. Kosmidis, M. Tali, and D. Steenari, Radiation Effects on NVIDIA Jetson SoCs: Insights from Heavy Ion Irradiation, in2025 IEEE International Symposium on Defect and Fault Tolerance in VLSI and Nanotechnology Systems (DFT)(2025) pp. 1–4

2025
[28]

G. P. Ginet, T. P. O’Brien, S. L. Huston, W. R. Johnston, T. B. Guild, R. Friedel, C. D. Lindstrom, C. J. Roth, P. Whelan, R. A. Quinn, D. Madden, S. K. Morley, and Y.-J. Su, AE9, AP9, and SPM: New Models for Specifying the Trapped Energetic Particle and Space Plasma Environment, Space Science Reviews179, 579 (2013)

2013
[29]

T. B. Guild, S. C. Davis, A. J. Boyd,et al.,Best Practices for Generating Space Environment Specifications with Modern Tools, Tech. Rep. TOR-2022-00016 (The Aerospace Corporation, 2021)

2022
[30]

L. R. Rodeck, F. Reina, R. R”uddenklau,et al., Cube1G: A Full-Hemispherical, High-Speed Optical Communication Terminal for CubeSats, inProceedings of the 76th International Astronautical Congress (IAC 2025)(Sydney, Australia,

2025
[31]

paper IAC-25,B2,IP,29,x94524
[32]

S. Buchner, Radiation Hardness Assurance (RHA) for Space Systems (2008), presentation at the 4th International School on the Effects of Radiation on Embedded Systems for Space Applications (SERESSA), West Palm Beach, FL

2008
[33]

Turner and Townsend, Data Centre Construction Cost Index 2025–2026 (2025), Report, Nov 5, 2025

2025
[34]

Cushman and Wakefield, Data Center Development Cost Guide 2025 (2025), Research report

2025
[35]

Rahkonen, A

T. Rahkonen, A. Lawrence, and R. Ascierto,Best-in-Class Data Center Provisioning, Tech. Rep. (Uptime Institute Intelligence, 2020) UII-36 v1.0, published June 8, 2020

2020
[36]

International Organization for Standardization, Space Systems – Space Debris Mitigation Requirements (2019), ISO 24113:2019

2019
[37]

S. G. Turyshev, Orbital Debris in Earth Orbit: Operations, Stability, Control, and Market Formation (2026), arXiv:2603.23552 [astro-ph.IM]

work page internal anchor Pith review Pith/arXiv arXiv 2026
[38]

R. R. Romanofsky,Optical Array Versus Monolithic Telescope Ground Station Cost Assessment, Tech. Rep. NASA/TM- 2019-220216 (NASA Glenn Research Center, 2019)

2019
[39]

SpaceX,Capabilities and Services, Tech. Rep. (SpaceX, 2026) Falcon 9 service capabilities and standard payment plan

2026