top of page
Search

Off-World Data Centers: A Critical Look at the SpaceX-xAI Merger

A satellite and rocket over Earth. Diagram with text: Energy flow, Thermal management. Ocean and coastline visible below.

Introduction

The convergence of aerospace engineering and artificial intelligence, crystallized by the acquisition of xAI by SpaceX, represents a paradigm shift in the industrial organization of the 21st century. This report provides an exhaustive analysis of the proposal to migrate the "digital backbone" of human civilization—specifically the training and inference infrastructure for Artificial General Intelligence (AGI)—from terrestrial data centers to a constellation of orbital platforms. The initiative, valued at approximately $1.25 trillion, is predicated on the thesis that Earth’s energy grids and thermal sinks are insufficient to sustain the exponential growth of computational demand required by next-generation AI models.1 By deploying a constellation of up to one million satellites, the combined entity aims to harness the unattenuated solar flux of the orbital environment while bypassing the geopolitical and infrastructural bottlenecks of the terrestrial surface.3

This document dissects the project across four critical dimensions: the thermodynamic and physical feasibility of rejecting gigawatts of waste heat in a vacuum; the material science challenges of operating nanometer-scale semiconductors in high-radiation environments; the legal and antitrust implications of a vertically integrated monopoly controlling launch, connectivity, and compute; and the profound environmental ethics of privatizing the orbital commons. The analysis suggests that while the "energy wall" facing terrestrial AI is real, the solution proposed by SpaceX introduces second-order externalities—ranging from stratospheric ozone depletion to the monopolization of the space economy—that may outweigh the benefits of orbital computation.

Part I: The Terrestrial Saturation Point and the Energy Wall

1.1 The Exponential Demand of Silicon Intelligence

The genesis of the orbital data center concept lies in the physical limitations of the terrestrial power grid. The training of Large Language Models (LLMs) has transformed compute from a resource management problem into an energy infrastructure crisis. Current estimates suggest that by 2027, the global energy consumption of AI-dedicated data centers could reach between 85 and 134 terawatt-hours (TWh) annually, an amount comparable to the total electricity consumption of nations such as the Netherlands or Argentina.5 This demand is not merely a function of volume but of density; modern AI clusters, utilizing racks of NVIDIA H100 or Blackwell GPUs, require power densities exceeding 100 kilowatts per rack, far stripping the cooling and power delivery capabilities of legacy data centers designed for 5-10 kW racks.7

In major digital hubs like Northern Virginia ("Data Center Alley"), utility providers are imposing multi-year moratoriums on new grid connections due to transmission constraints. The physical infrastructure of the electrical grid—transformers, substations, and high-voltage transmission lines—cannot be scaled at the "Moore’s Law" pace of silicon innovation. Furthermore, the thermal management of these facilities consumes vast quantities of fresh water for evaporative cooling, creating conflict with municipal water supplies in drought-prone regions like the American Southwest.9

1.2 The "Vertical Innovation Engine" Created by the SpaceX-xAI Merger

The acquisition of xAI by SpaceX is a strategic maneuver to internalize these constraints. By merging the launch provider (SpaceX), the satellite network operator (Starlink), and the AI developer (xAI) into a single corporate entity, Elon Musk aims to create a "Vertical Innovation Engine" capable of bypassing terrestrial utilities entirely.10 This integration allows for the optimization of the entire technology stack: the silicon can be designed specifically for the radiation and thermal environment of the satellite (via Tesla/xAI expertise), the satellite bus can be optimized for the fairing of the launch vehicle, and the launch cadence can be dictated by the deployment needs of the compute cluster rather than market availability.12

The economic logic rests on the "Starship Singularity"—the theoretical reduction of launch costs to below $50 per kilogram. At historic launch prices ($10,000/kg), launching heavy, copper-rich data center hardware was economically nonsensical. However, if Starship achieves its target economics, the cost of placing a server in orbit becomes comparable to the cost of land acquisition and construction for a terrestrial Tier 4 data center, while offering access to "free" solar energy.1

Part II: The Thermodynamics of Orbital Intelligence

2.1 The Photovoltaic Advantage: Harvesting the Solar Constant

The primary technical argument for orbital computing is the superior quality of energy available in Low Earth Orbit (LEO). On Earth, solar energy is filtered by the atmosphere, reducing the peak irradiance to approximately 1,000 watts per square meter (W/m2) under ideal conditions (Air Mass 1.5). In orbit, above the atmosphere, solar arrays receive the Air Mass 0 (AMO) spectrum, which carries a solar constant of approximately 1,366 W/m2 —a 36% immediate increase in photon energy density.1

More critical than peak power is the capacity factor. A terrestrial solar farm is bound by the diurnal cycle, generating power for only 6-8 hours a day. To run a data center 24/7, this necessitates massive battery storage systems (BESS) or grid interconnects to fossil-fuel baseload, effectively tripling the infrastructure cost per watt of useful compute. By utilizing Sun-Synchronous Orbits (SSO), specifically "dawn-dusk" or terminator orbits, SpaceX plans to place satellites in a regime of perpetual sunlight. In these orbits, the satellite rides the boundary between day and night, keeping its solar arrays illuminated for nearly the entire year, achieving a capacity factor approaching 95-100%.14 This allows for a "flow-through" energy architecture where photons are converted directly to electrons and then to floating-point operations, eliminating the mass and chemical hazard of large battery banks.

2.2 The Stefan-Boltzmann Restriction: The Cooling Bottleneck

While energy generation favors space, energy rejection—cooling—is the project's Achilles' heel. Thermodynamics dictates that all computational work eventually degrades into waste heat. On Earth, this heat is removed via convection: fans blow air over heat sinks, or liquid loops transfer heat to cooling towers where it is exchanged with the atmosphere. In the vacuum of space, convection is physically impossible. Heat rejection relies exclusively on thermal radiation, a process governed by the Stefan-Boltzmann law.

The implications of the Stefan-Boltzmann law are brutal for high-performance computing. Space is not "cold" in a thermodynamic sense; it is a vacuum, which is a perfect thermal insulator.16 To reject the 100 kW of heat generated by a single rack of modern AI accelerators (e.g., an NVIDIA GB200 NVL72 rack), a satellite would require massive radiator surfaces. For comparison, the International Space Station (ISS) utilizes huge, deployable radiator wings to reject approximately 70-100 kW of heat—the entire thermal budget of the station is roughly equivalent to one modern AI server rack.17

2.3 Thermal Management Architectures

To achieve the claimed power density of "100 kW per ton" of satellite mass 19, the SpaceX-xAI engineering teams must effectively rewrite the rulebook of spacecraft thermal control. Current state-of-the-art deployable radiators have a specific mass of roughly 5-12 kg per square meter.21 Scaling this to megawatt-class clusters would result in radiators that are prohibitively heavy and large, creating immense atmospheric drag in LEO and increasing the risk of debris collisions.

Several theoretical approaches are likely being explored to overcome this limit:

  1. High-Temperature Operation: The Stefan-Boltzmann law is highly sensitive to temperature. If the silicon can operate at higher temperatures (e.g., 80C° vs 40C°), the radiator efficiency increases dramatically. This drives a need for wide-bandgap semiconductors (like Silicon Carbide or Gallium Nitride) in the power delivery systems, although the logic cores themselves (CPUs/GPUs) still generally require lower temperatures.23

  2. Active Refrigeration: Using heat pumps (vapor compression cycles) to boost the temperature of the fluid loop before it reaches the radiator. This consumes energy (increasing the total heat load) but allows the radiator to operate at a much higher temperature, reducing its required surface area. This trade-off is only viable because solar power in orbit is abundant and "free".24

  3. Liquid Droplet Radiators (LDR): An advanced concept where the working fluid is sprayed directly into space as a sheet of droplets, maximizing surface area for radiation before being recollected. While theoretically offering specific masses an order of magnitude lower than solid panels, LDR technology remains experimental and poses risks of fluid loss and contamination.25

Metric

Terrestrial Data Center

Orbital Data Center (Proposed)

Cooling Mechanism

Convection (Air/Water) + Evaporation

Radiation Only

Heat Rejection Capacity

Unlimited (Atmosphere)

Constrained by Surface Area (m2)

Water Consumption

High (Liters/kWh)

Zero

Energy Source

Grid (Mix of Fossil/Renewable)

100% Solar (AMO)

Capacity Factor

Variable (Grid/Weather dependent)

>95% (Terminator Orbit)

Thermal Inertia

High (Building mass)

Low (Vacuum flask effect)

Part III: The Material Science of the Orbital Environment

3.1 The Radiation Environment: Hardening the Silicon Brain

The transition to orbital computing exposes delicate semiconductor logic to the harsh ionizing radiation environment of space. LEO is permeated by high-energy protons and electrons trapped in the Van Allen radiation belts, as well as Galactic Cosmic Rays (GCRs) originating from deep space.

Modern AI chips, such as the NVIDIA H100 or Tesla’s D1, are fabricated on extremely advanced process nodes (3nm or 4nm FinFET). These nanoscale transistors are uniquely vulnerable to radiation effects:

  • Total Ionizing Dose (TID): As radiation passes through the silicon dioxide insulating layers, it generates electron-hole pairs. Over time, trapped charges accumulate, shifting the threshold voltage required to switch the transistor. This leads to increased leakage current, timing errors, and eventually total device failure.27

  • Single Event Upsets (SEU): A single high-energy particle striking a memory cell or logic gate can deposit enough charge to flip a bit from 0 to 1. In a neural network, a bit flip in a weight parameter might be benign, but a bit flip in a control register could crash the system.29

  • Single Event Latch-up (SEL): The most catastrophic failure mode, where a particle strike triggers a parasitic thyristor structure within the chip, creating a short circuit that allows massive current to flow until the device burns out.30

3.2 The "Rad-Tolerant" Paradigm

Historically, space missions relied on "Radiation Hardened" (Rad-Hard) chips, which use larger transistors (e.g., 65nm, 130nm) and specialized substrates (Silicon-on-Insulator) to physically prevent these errors. However, Rad-Hard chips are generations behind consumer technology in terms of speed and efficiency. To make orbital AI viable, SpaceX cannot use legacy space chips; they must use COTS (Commercial Off-The-Shelf) hardware and protect it.

The strategy likely involves a multi-layered defense:

  • Shielding: Encasing server racks in aluminum or hydrogen-rich polymers (like polyethylene) to attenuate particle flux. However, shielding has diminishing returns; thick metal shields can interact with high-energy GCRs to produce showers of secondary particles (neutrons) that are even more damaging than the primary radiation.31

  • Software Resilience: Neural networks exhibit a degree of inherent fault tolerance. Unlike a financial database where a single bit flip is unacceptable, a deep learning model is probabilistic. A minor error in a neuron's activation value may not significantly affect the final inference. This allows xAI to accept higher hardware error rates than traditional computing applications.28

  • System-Level Redundancy: Utilizing Triple Modular Redundancy (TMR), where the same calculation is performed on three separate cores simultaneously. A voting logic compares the results; if one core has suffered a bit flip, the other two overrule it. This sacrifices compute density for reliability.28

3.3 Re-entry Pollution: The Alumina Threat

The environmental impact of the hardware lifecycle presents a severe, under-regulated risk. A constellation of one million satellites implies a continuous conveyor belt of launches and de-orbits. Satellites in LEO have finite lifespans (5-7 years) due to orbital decay and component degradation. To maintain the constellation, thousands of satellites must be de-orbited annually, burning up in the Earth's atmosphere.

When satellites constructed of aluminum alloys burn up during re-entry, they do not simply vanish. They vaporize into aluminum oxide (alumina, Al2O) nanoparticles in the mesosphere and stratosphere.33 These particles have profound atmospheric effects:

  • Ozone Depletion: Alumina particles provide surface area for chlorine activation reactions, catalyzing the destruction of ozone molecules. Recent studies estimate that the re-entry of megaconstellations could deposit up to 360 metric tons of aluminum oxide into the upper atmosphere annually—an increase of 650% over natural levels—potentially stalling or reversing the recovery of the ozone layer.34

  • Radiative Forcing: The accumulation of metallic ash in the stratosphere alters the Earth's albedo (reflectivity). These particles can scatter incoming sunlight or trap outgoing infrared heat, acting as an unintentional geoengineering agent with unknown climatic consequences.36

This "atmospheric dumping" creates a stark ethical contradiction: a project ostensibly designed to save the biosphere from the carbon emissions of terrestrial data centers may end up damaging the biosphere through stratospheric pollution.

Part IV: The Economics of Ascent

4.1 The Cost-to-Orbit Equation

The entire economic viability of the orbital data center thesis rests on the launch economics of the SpaceX Starship. Traditional aerospace costs ($2,000 - $10,000 per kg) make space hardware the most expensive real estate in existence. At those prices, launching a 500 kg server rack costing $200,000 would cost $5 million in launch fees alone, destroying any ROI.

Starship aims to fundamentally break this equation by achieving full reusability and high launch cadence, targeting launch costs below $50/kg.1 If achieved, the launch cost becomes a minor line item compared to the cost of the silicon itself.

  • Terrestrial Cost Structure: High OPEX (electricity bills, cooling water, real estate taxes, maintenance staff).

  • Orbital Cost Structure: High CAPEX (Launch, satellite manufacturing) but near-zero OPEX (free solar energy, passive cooling, no rent, no staff).

If the lifespan of the satellite is sufficiently long, the "free" energy of the sun eventually pays back the initial launch cost. Analyses by startups like Starcloud (formerly Lumen Orbit) suggest that over a 10-year period, the Total Cost of Ownership (TCO) for an orbital data center could be significantly lower than a terrestrial counterpart, provided launch costs hit their aggressive targets.37

4.2 Latency and the Speed of Light

A critical economic differentiator is network latency. Light travels approximately 47% slower in glass fiber optic cables (refractive index ~1.5) than it does in the vacuum of space.

  • Fiber Speed: ~200,000 km/s.

  • Vacuum Speed: ~300,000 km/s.

For long-haul data transmission (e.g., London to Singapore), routing data via laser inter-satellite links (OISL) in LEO is physically faster than routing it through undersea cables.39 This "latency arbitrage" makes orbital compute highly attractive for specific high-value workloads like high-frequency trading (HFT) or real-time inference for autonomous systems. However, for the bulk of AI training workloads, bandwidth is more critical than latency. The massive datasets required to train models like GPT-5 (petabytes of text and video) would be difficult to upload to orbit via RF uplinks. This suggests a bifurcated model: Training may remain terrestrial (or require physical shipment of data storage modules to orbit), while Inference (serving the model to users) moves to space to leverage the global low-latency distribution network of Starlink.1

Part V: Legal Frameworks: Antitrust and Competition

5.1 The Vertical Monolith and Market Foreclosure

The merger of SpaceX and xAI creates a vertically integrated entity with potential monopoly power across multiple sectors: launch, satellite internet, and AI compute. From an antitrust perspective, this raises the specter of input foreclosure.41

  • Launch Dominance: SpaceX currently launches over 80% of all global payload mass to orbit. It effectively controls the "railroad to space."

  • Foreclosure Risk: Antitrust regulators (FTC, DOJ, EU Commission) will scrutinize whether SpaceX has the incentive and ability to deny launch services or charge supracompetitive prices to rival satellite operators (like Amazon's Project Kuiper or OneWeb) to prevent them from deploying competing orbital data centers.43

  • Self-Preferencing: By offering xAI preferential access to Starship launches and Starlink backhaul, the combined entity could make it impossible for terrestrial AI companies (like OpenAI or Anthropic) or other space startups to compete on price or performance.43

5.2 The Essential Facilities Doctrine

Legal scholars may argue that the Starship launch system and the finite orbital shells of LEO constitute "essential facilities".44 The Essential Facilities Doctrine in antitrust law mandates that a monopolist controlling a bottleneck facility must provide access to competitors on reasonable terms if duplication of the facility is not feasible. Given the immense capital and technical barriers to building a reusable heavy-lift rocket, competitors could argue that they cannot compete in the orbital compute market without fair access to Starship. However, U.S. courts have historically been reluctant to apply this doctrine broadly, creating a complex legal battleground.46

5.3 Cross-Market Leverage and Tying

The merger allows for "tying" arrangements, where a customer buying Starlink connectivity might be compelled or incentivized to use xAI's compute services. This is analogous to the United States v. Microsoft case, where Microsoft used its dominance in operating systems to monopolize the browser market.47 Regulators will likely assess whether SpaceX is using its launch monopoly to unfairly leverage its way into the AI market.12

Part VI: Governance of the Void – Space Law and Sovereignty

6.1 The Jurisdictional Vacuum

The deployment of data centers in space challenges the Westphalian concept of sovereignty. Under Article VIII of the Outer Space Treaty (OST), the state of registry retains "jurisdiction and control" over a space object.50 This means a SpaceX data center registered in the United States is legally U.S. territory, even when orbiting over China or the European Union.

This creates a potential "Data Haven" scenario. Data stored on a US-registered satellite might be beyond the reach of EU data protection authorities, complicating the enforcement of laws like the General Data Protection Regulation (GDPR).51 While GDPR asserts extraterritorial reach (applying to the data of EU subjects regardless of processing location), enforcing a "right to be forgotten" or a data audit on a server flying at 7.6 km/s in a vacuum presents novel enforcement challenges.53

6.2 Data Sovereignty vs. Orbital Reality

Nations are increasingly asserting "digital sovereignty"—the requirement that citizen data be stored and processed within national borders. Orbital data centers inherently violate this principle. A satellite in a polar orbit passes over every country on Earth. Does processing German banking data on a satellite overflying Russia constitute a data transfer? This ambiguity could lead to a fragmentation of the orbital internet, with nations demanding "sovereign clouds" in space—a concept already being explored by the EU's ASCEND project to ensure European strategic autonomy.53

Part VII: Environmental Ethics and Externalities

7.1 The Kessler Syndrome: Enclosure of the Commons

The proposal to launch one million satellites 3 represents a statistical assault on the orbital environment. The current active satellite population is roughly 10,000. Increasing this by two orders of magnitude drastically increases the probability of collision.

  • The Chain Reaction: The Kessler Syndrome describes a cascade where debris from one collision destroys other satellites, creating a debris belt that renders LEO unusable for centuries.56

  • Active vs. Passive Debris: While SpaceX satellites have automated collision avoidance, this system relies on the satellite being functional. A "dead" data center satellite—a multi-ton unguided mass—becomes a lethal projectile. With a million satellites, even a 0.1% failure rate leaves 1,000 massive derelict objects in orbit, acting as triggers for a cascade event.57

This raises an ethical question of intergenerational equity: Does one generation's desire for cheap AI compute justify the risk of foreclosing access to space for future generations?

7.2 Light Pollution: Erasing the Night

A constellation of this magnitude poses an existential threat to ground-based astronomy. Satellites reflect sunlight, appearing as bright streaks in telescope exposures. Even with "DarkSat" mitigation coatings, the sheer density of objects could saturate the detectors of wide-field survey telescopes like the Vera Rubin Observatory.58 This degradation of the night sky represents a loss of global cultural heritage and scientific potential, effectively privatizing the view of the cosmos.59

Conclusion: The Sky as the New Silicon Valley

The SpaceX-xAI orbital data center initiative is a technological wager of staggering proportions. It bets that the constraints of physics on Earth (energy scarcity, thermal inefficiency) are more binding than the constraints of physics in space (radiation, vacuum, launch mass).

Technically, the project hinges on unproven breakthroughs: the ability of Starship to make mass to orbit negligible, and the ability of thermal engineers to reject gigawatts of heat without convection. Economically, it relies on the "Energy Wall" becoming so impermeable that the extreme capital expenditure of space infrastructure becomes rational.

Ethically and legally, the project pushes humanity into uncharted territory. It challenges the foundational principle of the Outer Space Treaty—that space is the "province of all mankind"—by threatening to enclose the orbital commons with a dense mesh of private infrastructure. It introduces environmental risks, from ozone depletion to orbital debris, that are global in scope but decided by a single corporate board. As the first Starships load their server racks, the world faces a definitive choice: will the future of intelligence be grounded in the biosphere, or will it ascend to the cold, abundant, and lawless expanse of the orbital frontier?

Statistical Appendices

Table 1: Comparative Metrics of Terrestrial vs. Orbital Data Centers

Parameter

Terrestrial Data Center

Orbital Data Center (Proposed)

Implication

Solar Irradiance

~1,000 W/m2 (Peak AM1.5)

~1,366 W/m2 (Constant AMO)

36% higher base energy density in space.

Capacity Factor

~20-25% (Day/Night cycle)

>95% (Terminator Orbit)

Eliminates need for grid/battery storage.

Cooling Method

Convection (Air/Water)

Radiation (Stefan-Boltzmann)

Space requires massive surface area for heat rejection.

Water Usage

High (Evaporative cooling)

Zero

Eliminates impact on local water stress.

Radiation Risk

Negligible (Atmosphere shielded)

High (Van Allen Belts/GCR)

Requires expensive shielding or resilient hardware.

Latency

Variable (Fiber routing)

Low (Speed of light in vacuum)

Faster long-haul transmission via laser links.

Maintenance

Easy (Human accessible)

Impossible (No servicing)

Requires extreme hardware reliability.

Table 2: Environmental Impact Analysis


Risk Factor

Mechanism

Potential Consequence

Source Data

Ozone Depletion

Alumina (Al2O) from re-entry

Catalytic destruction of ozone

33

Radiative Forcing

Black Carbon from launch

Stratospheric heating / Albedo change

61

Orbital Debris

Collision Cascade

Kessler Syndrome (LEO unusable)

56

Light Pollution

Solar reflection

Degradation of astronomy

58



Works cited

  1. Elon Musk’s Vision: Why He Wants To Build AI Data Centres In Space, accessed February 2, 2026, https://www.outlookindia.com/international/elon-musks-vision-why-he-wants-to-build-ai-data-centres-in-space

  2. SpaceX formalizes plan to build 1 million satellite Orbital Data Center System — FCC filing sketches out plans, but over-packed orbits could be limiting factor, accessed February 2, 2026, https://www.tomshardware.com/tech-industry/spacex-formalizes-plan-to-build-1-million-satellite-orbital-data-center-system-fcc-filing-sketches-out-plans-but-over-packed-orbits-could-be-limiting-factor

  3. Elon Musk lays out a new vision of AI satellites as SpaceX acquires xAI, accessed February 2, 2026, https://www.universetoday.com/articles/elon-musk-lays-out-a-new-vision-of-ai-satellites-as-spacex-acquires-xai

  4. AI to drive 165% increase in data center power demand by 2030 | Goldman Sachs, accessed February 2, 2026, https://www.goldmansachs.com/insights/articles/ai-to-drive-165-increase-in-data-center-power-demand-by-2030

  5. AI data centers are pushing power grids to the brink, accessed February 2, 2026, https://www.devdiscourse.com/article/technology/3786888-ai-data-centers-are-pushing-power-grids-to-the-brink

  6. Gartner Predicts Power Shortages Will Restrict 40% of AI Data Centers By 2027, accessed February 2, 2026, https://www.gartner.com/en/newsroom/press-releases/2024-11-12-gartner-predicts-power-shortages-will-restrict-40-percent-of-ai-data-centers-by-20270

  7. Can US infrastructure keep up with the AI economy? - Deloitte, accessed February 2, 2026, https://www.deloitte.com/us/en/insights/industry/power-and-utilities/data-center-infrastructure-artificial-intelligence.html

  8. New study highlights the viability of space data centres - Innovation News Network, accessed February 2, 2026, https://www.innovationnewsnetwork.com/new-study-highlights-the-viability-of-space-data-centres/48810/

  9. SpaceX bails out xAI in mega-deal — here’s what it means for Tesla, accessed February 2, 2026, https://electrek.co/2026/02/02/spacex-bails-out-xai-in-mega-deal-heres-what-it-means-for-tesla/

  10. Musk merges SpaceX with xAI ‘to accelerate humanity’s future’, accessed February 2, 2026, https://www.washingtonexaminer.com/policy/technology/4443577/elon-musk-spacex-xai-merger/

  11. SpaceX-xAI merger: Billionaire's blunt take signals major tech shift - Winvesta, accessed February 2, 2026, https://www.winvesta.in/blog/us-market-news/spacex-xai-merger-billionaires-blunt-take-signals-major-tech-shift

  12. SpaceX Acquires xAI: What the Deal Signals for AI, Starlink, and “Compute in Space” (2026), accessed February 2, 2026, https://www.abzglobal.net/web-development-blog/spacex-acquires-xai-what-the-deal-signals-for-ai-starlink-and-compute-in-space-2026

  13. How data centres in space sustainably enable the AI age - The World Economic Forum, accessed February 2, 2026, https://www.weforum.org/stories/2026/01/data-centres-space-ai-revolution/

  14. Can You Really Put a Data Center in Space? (yeah probably) - Medium, accessed February 2, 2026, https://medium.com/@Elongated_musk/orbital-data-centers-on-starlink-feasibility-outlook-a97d1a4e3a2f

  15. If Datacenters Go to Space, Heat Becomes the Hardest Problem - New England Die Cutting, accessed February 2, 2026, https://www.nedc.com/if-datacenters-go-to-space-heat-becomes-the-hardest-problem/

  16. Heat transfer in space-based data centers?, accessed February 2, 2026, https://www.reddit.com/r/AskPhysics/comments/1pj2dpp/heat_transfer_in_spacebased_data_centers/

  17. Lightweight, High-Temperature Radiator for Space Propulsion - NASA Technical Reports Server, accessed February 2, 2026, https://ntrs.nasa.gov/api/citations/20130001608/downloads/20130001608.pdf

  18. accessed February 2, 2026, https://newspaceeconomy.ca/2026/01/31/spacex-files-plans-for-massive-orbital-data-center-constellation/#:~:text=The%20Kessler%20Syndrome%20is%20a,necessitating%20advanced%20collision%20avoidance%20systems.

  19. The basic math is that launching a million tons per year of satellites generat... - Hacker News, accessed February 2, 2026, https://news.ycombinator.com/item?id=46862510

  20. 6 Radiators, accessed February 2, 2026, http://matthewwturner.com/uah/IPT2008_summer/baselines/LOW%20Files/Thermal/Spacecraft%20Thermal%20Control%20Handbook/06.pdf

  21. Advanced Lightweight Heat Rejection Radiators for Space Nuclear Power Systems, accessed February 2, 2026, https://isnps.unm.edu/reports/ISNPS_Tech_Report_97.pdf

  22. Realities of Space-Based Compute - Per Aspera, accessed February 2, 2026, https://www.peraspera.us/realities-of-space-based-compute/

  23. Why Putting AI Data Centers in Space Doesn’t Make Much Sense, accessed February 2, 2026, https://www.reddit.com/r/space/comments/1pix4md/why_putting_ai_data_centers_in_space_doesnt_make/

  24. Liquid Droplet Radiator Development Status - NASA Technical Reports Server, accessed February 2, 2026, https://ntrs.nasa.gov/api/citations/19870010920/downloads/19870010920.pdf

  25. Liquid droplet radiator for space applications - Thermopedia, accessed February 2, 2026, https://www.thermopedia.com/de/content/212/

  26. Review of Passive Shielding Materials for High-Energy Charged Particles in Earth's Orbit, accessed February 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC12156362/

  27. A Case for Application-Aware Space Radiation Tolerance in Orbital Computing - arXiv, accessed February 2, 2026, https://arxiv.org/html/2407.11853v1

  28. Edge AI Trustworthiness: Revisiting Bit-Flip Impacts and Criticality Conditions - MDPI, accessed February 2, 2026, https://www.mdpi.com/2079-9292/14/16/3186

  29. Space Radiation Effects on Electronic Components in Low-Earth Orbit - NASA Lessons Learned, accessed February 2, 2026, https://llis.nasa.gov/lesson/824

  30. Galactic and Solar Cosmic Ray Shielding in Deep Space - NASA Technical Reports Server, accessed February 2, 2026, https://ntrs.nasa.gov/api/citations/19980006777/downloads/19980006777.pdf

  31. Toward the Use of Electronic Commercial Off-the-Shelf Devices in Space: Assessment of the True Radiation Environment in Low Earth Orbit (LEO) - MDPI, accessed February 2, 2026, https://www.mdpi.com/2079-9292/12/19/4058

  32. accessed February 2, 2026, https://downloads.regulations.gov/FAA-2024-1395-0037/attachment_1.pdf

  33. Satellite megaconstellations threaten ozone layer recovery, study confirms - Space, accessed February 2, 2026, https://www.space.com/megaconstellations-threat-to-ozone-layer-recovery

  34. Envisioning a sustainable future for space launches: a review of current research and policy, accessed February 2, 2026, https://pmc.ncbi.nlm.nih.gov/articles/PMC11459831/

  35. ENVIRONMENTAL IMPACT OF LARGE LEO CONSTELLATIONS - Viasat, accessed February 2, 2026, https://www.viasat.com/content/dam/us-site/corporate/documents/Environmental_Impact_of_Large_LEO_Constellations_-_13th_IAASS.pdf

  36. Why we should train AI in space - White Paper - GitHub Pages, accessed February 2, 2026, https://starcloudinc.github.io/wp.pdf

  37. This Startup Wants to Tackle AI Energy Demands With Data Centers in Space | PCMag, accessed February 2, 2026, https://www.pcmag.com/news/this-startup-wants-to-tackle-ai-energy-demands-with-data-centers-in-space

  38. Starlink vs. Fiber Internet: A Complete Comparison of Speed, Latency, and Cost, accessed February 2, 2026, https://www.tataplayfiber.com/blog/starlink-vs-fiber-internet-complete-comparison-speed-latency-and-cost

  39. Starlink latency vs fiber!? - Reddit, accessed February 2, 2026, https://www.reddit.com/r/Starlink/comments/x09n6a/starlink_latency_vs_fiber/

  40. Chapter II. The Final Economic Frontier: Satellite Competition in Low Earth Orbit, accessed February 2, 2026, https://www.newamerica.org/oti/wireless-future-project/reports/leo-satellites/chapter-ii-the-final-economic-frontier-satellite-competition-in-low-earth-orbit/

  41. Antitrust Regulators Release New Vertical Merger Guidelines - Congress.gov, accessed February 2, 2026, https://www.congress.gov/crs-product/LSB10521

  42. The Antitrust Case Against SpaceX - The Sling, accessed February 2, 2026, https://www.thesling.org/the-antitrust-case-against-spacex/

  43. The Essential Facilities Doctrine: What Does It Mean To Be Essential? - Marquette Law Scholarly Commons, accessed February 2, 2026, https://scholarship.law.marquette.edu/cgi/viewcontent.cgi?article=1517&context=mulr

  44. The Consequences of Search Bias: How Application of the Essential Facilities Doctrine Remedies Google's Unrestricted Monopoly - The George Washington Law Review, accessed February 2, 2026, https://www.gwlr.org/wp-content/uploads/2015/05/83-Geo-Wash-L-Rev-721.pdf

  45. Essential Facilities - Criterion Economics, Inc., accessed February 2, 2026, https://www.criterioneconomics.com/docs/lipsky-sidak-essential-facilities.pdf

  46. United States v. Microsoft Corp. - Wikipedia, accessed February 2, 2026, https://en.wikipedia.org/wiki/United_States_v._Microsoft_Corp.

  47. Antitrust Division | U.S. V. Microsoft: Court's Findings Of Fact - Department of Justice, accessed February 2, 2026, https://www.justice.gov/atr/us-v-microsoft-courts-findings-fact

  48. Amazon's Antitrust Paradox - Yale Law Journal, accessed February 2, 2026, https://yalelawjournal.org/note/amazons-antitrust-paradox

  49. Post-Territorial Sovereignty: Reassessing State Jurisdiction over Space Objects under International Space Law | European Journal of Law and Political Science, accessed February 2, 2026, https://eu-opensci.org/index.php/politics/article/view/8189

  50. The Application of Data Protection Laws in (Outer) Space - Rothwell Figg, accessed February 2, 2026, https://www.rothwellfigg.com/assets/htmldocuments/ICLG__Data_Protection_2019_RothwellFigg_Outer_Space.pdf

  51. The Extra-Territorial Reach of EU Data Protection Law | Insights | Sidley Austin LLP, accessed February 2, 2026, https://www.sidley.com/en/insights/publications/2019/07/the-extra-territorial-reach-of-eu-data-protection-law

  52. Legal Considerations for Data Processing From Outer Space Through Remote Sensing | Insights | Mayer Brown, accessed February 2, 2026, https://www.mayerbrown.com/en/insights/publications/2025/07/legal-considerations-for-data-processing-from-outer-space-through-remote-sensing

  53. The “Third Way” to Space Power: Europe's Digital Sovereignty Advantage, accessed February 2, 2026, https://law.stanford.edu/2026/01/23/the-third-way-to-space-power-europes-digital-sovereignty-advantage/

  54. Data Centres in Space | ASCEND, accessed February 2, 2026, https://ascend-horizon.eu/data-centres-in-space/

  55. Kessler syndrome - Wikipedia, accessed February 2, 2026, https://en.wikipedia.org/wiki/Kessler_syndrome

  56. Space Debris and the Kessler Problem: A Reality Check | by Marc Bara - Medium, accessed February 2, 2026, https://medium.com/@marc.bara.iniesta/space-debris-and-the-kessler-problem-a-reality-check-99484cfa5e86

  57. Blinded by the light: How bad are satellite megaconstellations for astronomy? - Reddit, accessed February 2, 2026, https://www.reddit.com/r/space/comments/1ct58aw/blinded_by_the_light_how_bad_are_satellite/

  58. Light pollution from satellite mega-constellations threaten space-based observations, accessed February 2, 2026, https://physicsworld.com/a/light-pollution-from-satellite-mega-constellations-threaten-space-based-observations/

  59. As space gets more commercial, how can it be governed ethically? | Aeon Essays, accessed February 2, 2026, https://aeon.co/essays/as-space-gets-more-commercial-how-can-it-be-governed-ethically

  60. Projected increase in space travel may damage ozone layer - NOAA Research, accessed February 2, 2026, https://research.noaa.gov/projected-increase-in-space-travel-may-damage-ozone-layer/

bottom of page