Will Space Data Centers Destroy Stargazing? Columbia Researchers Warn of a Satellite Crisis

The race to build increasingly powerful artificial intelligence systems is creating demand for computing infrastructure on an extraordinary scale. Data centers require enormous quantities of electricity, advanced cooling systems, networking equipment, and suitable locations. As companies encounter constraints on land, energy availability, and infrastructure development, some are exploring an unconventional solution: moving AI data centers into space.
The concept of orbital data centers involves placing computing hardware in satellites equipped with solar power systems, thermal management infrastructure, and communications links. In theory, these platforms could use continuous access to solar energy and operate above many of the geographical and infrastructure constraints associated with terrestrial facilities.
However, a November 2026 working paper prepared for the Innovation for Cool Earth Forum (ICEF) raises serious questions about the environmental, astronomical, economic, and geopolitical consequences of deploying such systems at scale.
Led by David Sandalow of Columbia University's Center on Global Energy Policy, the report, titled Space-Based Data Centers Roadmap: The Case for Caution, warns that proposed satellite constellations could dramatically change the appearance of the night sky. If announced plans were fully implemented, satellites could become more numerous than visible stars across many locations and periods of the night.
The researchers also identify concerns involving orbital congestion, space debris, rocket emissions, atmospheric effects, and the enormous launch requirements associated with operating computing infrastructure beyond Earth.
The debate illustrates a fundamental challenge facing the AI industry: solving the energy and infrastructure requirements of artificial intelligence without creating environmental and technological costs that outweigh the benefits.
Why Technology Companies Want to Build AI Data Centers in Space
Modern AI systems depend on large computing clusters containing specialized processors, high-speed networking equipment, and sophisticated power infrastructure. Training advanced models and operating them at commercial scale can require substantial computing capacity, creating pressure to expand data center construction.
On Earth, developers must overcome several constraints:
Electricity availability: Large AI facilities require reliable power supplies, and grid connections can take considerable time to secure.
Cooling requirements: High-performance processors generate significant heat, requiring extensive thermal management.
Land and construction: Suitable locations must accommodate buildings, substations, networking infrastructure, and supporting facilities.
Environmental pressures: Data centers can place additional demands on local electricity and water resources.
Infrastructure development: Planning, permitting, equipment procurement, and grid upgrades can delay deployment.
Orbital computing offers a different infrastructure model. Satellites can use solar arrays to generate electricity and operate without occupying conventional industrial land. Their computing equipment could potentially serve AI workloads through communication links connecting orbital platforms with users and terrestrial networks.
The concept is attractive because AI infrastructure demand is expanding while the physical resources required to support it remain constrained.
However, space does not eliminate infrastructure challenges. It changes their location and nature. Instead of relying on terrestrial electricity grids and conventional cooling systems, orbital data centers must manage launch costs, radiation exposure, thermal rejection, communications latency, maintenance limitations, and end-of-life disposal.
The central question is therefore not simply whether computers can operate in orbit, but whether an orbital computing network can deliver reliable services at a competitive economic and environmental cost.
The Scale of Proposed Space Data Center Constellations
The Columbia-led report highlights the extraordinary scale of proposals currently under consideration.
Approximately 19,000 satellites are in orbit around Earth today, according to the supplied report. Full implementation of the proposed space-based data center plans could increase the number of satellites in low Earth orbit by more than 50 times.
Several companies have submitted ambitious constellation proposals:
Company or project | Proposed satellite constellation |
SpaceX | Up to 1,000,000 satellites |
Starcloud | 88,000 satellites |
Blue Origin, Project Sunrise | Up to 51,600 satellites |
Orbital | Up to 100,000 satellites |
These figures represent proposed or requested constellation sizes, not confirmed deployments.
SpaceX submitted an application to the Federal Communications Commission in January 2026 seeking permission to operate up to one million data-center satellites. Other companies subsequently filed proposals involving hundreds of thousands of additional satellites.
The distinction between regulatory applications and actual deployment is important. Caleb Henry, research director at Quilty Space, cautioned that companies may request authorization for more satellites than they ultimately intend to operate. Filing for a large constellation can provide flexibility for future expansion without establishing that the full network will be constructed.
Nevertheless, the scale of these proposals demonstrates how seriously parts of the technology and space industries are considering orbital computing.
The investment trend also suggests growing commercial interest. According to market research firm Quilty Space, approximately $197 million was associated with orbital data center investments in 2025. That figure had increased to $1.11 billion during 2026 at the time of the supplied reporting.
This financial momentum is developing alongside substantial uncertainty about technical feasibility, commercial returns, and long-term environmental consequences.
Why Space Data Centers Could Make Satellites More Visible Than Stars
One of the report's most striking concerns involves light pollution.
Astronomical observations depend on dark skies, particularly when researchers study faint stars, distant galaxies, and other objects that produce relatively little visible light. Increasing numbers of bright artificial objects crossing the sky can interfere with these observations and alter the natural appearance of the night.
The problem could become substantially more serious if orbital data center satellites are significantly larger and brighter than conventional communications satellites.
The report identifies SpaceX's proposed AI1 satellite as having a wingspan of approximately 70 meters, more than twice the size of a Starlink satellite. Some experts cited by the authors expect future data-center satellites to reach approximately 100 meters from tip to tip because of their extensive solar arrays and radiators.
These large structures introduce several potential sources of visible brightness.
Solar Arrays and Reflective Surfaces
Large solar arrays are necessary to generate electricity for orbital computing systems. Their surfaces can reflect sunlight toward observers on Earth, making satellites visible even when the ground below is dark.
The brightness of a satellite depends on several factors, including its size, orientation, surface materials, altitude, and position relative to the Sun and observer.
A much larger satellite can present a substantially greater reflective surface than a conventional spacecraft, increasing the potential for visible interference.
Radiators and Thermal Management
AI processors produce heat during operation. Spacecraft must transfer that heat away from their electronics to prevent overheating and maintain reliable performance.
On Earth, data centers commonly use air-based or liquid-based cooling systems to transport heat away from processors. Spacecraft operate in a different thermal environment, where radiators play a central role in releasing heat through infrared radiation.
The report identifies a difficult engineering trade-off. Radiator coatings designed to reflect sunlight can also appear bright in visible wavelengths. Darkening these surfaces may reduce visible brightness, but it can increase solar absorption and make thermal management less effective.
This creates a conflict between astronomical protection and computing efficiency.
The heat generated by operating hardware must still be released. Infrared radiation is not an optional by-product that can simply be eliminated through darker coatings.
Consequently, large orbital computing platforms could become significant sources of both visible reflection and infrared emission.
The American Astronomical Society raised concerns about the potential impact in a filing opposing SpaceX's January 2026 application, describing the proposal as an unprecedented threat to the visual integrity of the night sky.
The Astronomical Consequences of Artificial Satellite Constellations
The night sky is a shared scientific, cultural, and environmental resource. Its value extends beyond professional astronomy.
Astronomers use ground-based observatories to study objects across the electromagnetic spectrum, including visible light. Satellite trails can interfere with observations, complicate image processing, and reduce the amount of useful data collected during observing sessions.
Large orbital data center constellations could intensify these challenges.
For optical astronomy, bright satellites may cross the field of view during long-exposure observations. Software can sometimes identify and remove satellite trails, but this does not fully restore the information obscured by the moving object.
Large numbers of satellites could increase the frequency of such interference, making observation scheduling and data processing more difficult.
Infrared astronomy faces a different set of concerns. Orbital data centers must release waste heat, and large constellations could introduce additional sources of infrared radiation into the near-Earth environment.
The combined impact would depend on satellite design, operating altitude, orbital distribution, thermal characteristics, and observation conditions.
The report's warning about satellites potentially becoming more visible than stars is therefore not simply an aesthetic concern. It raises questions about access to dark skies, the scientific value of astronomical observations, and the long-term management of Earth's orbital environment.
The Hidden Economics of Launching Orbital AI Infrastructure
Space-based data centers are frequently discussed in terms of their access to solar energy. Yet transporting computing hardware into orbit remains one of the most demanding parts of the business model.
A terrestrial data center can receive equipment through conventional logistics networks. An orbital facility must be launched aboard a rocket, deployed, integrated into a functioning network, and eventually replaced or de-orbited.
The Columbia-led report questions weight estimates associated with proposed orbital computing satellites.
SpaceX has suggested in regulatory filings that a 100-kilowatt data-center satellite could weigh approximately one tonne. The report's authors estimate that the power system alone could weigh around 2.2 tonnes.
This difference matters because satellite mass directly influences launch requirements.
The researchers estimate that serving just 10% of projected 2030 AI data center demand from orbit would require at least 700 Starship-sized launches annually. That is approximately two launches every day, continuing for as long as the orbital fleet remains operational.
<text size="sm" color="secondary">The estimate illustrates the scale of the transportation system required to support orbital computing. It is a projection from the report, not a confirmed launch schedule or an existing operational requirement.</text>
The launch challenge extends beyond the initial deployment. Satellites have finite operational lifetimes, hardware can become obsolete, and failures may require replacement missions. Maintaining a large constellation could therefore create continuing demand for launch capacity.
The economics must account for the entire lifecycle:
Cost category | Commercial implication |
Launch and deployment | Determines the initial cost of placing computing equipment in orbit |
Solar power systems | Adds mass and structural complexity |
Thermal management | Requires radiators and specialized spacecraft engineering |
Communications | Connects orbital computing platforms to terrestrial customers |
Maintenance and replacement | Creates recurring capital requirements |
De-orbiting | Adds end-of-life operational and financial obligations |
Even if orbital computing eventually offers advantages in energy availability, those benefits must be weighed against transportation, maintenance, and replacement expenses.
The report also notes that high costs may not prevent initial deployments. Companies could accept financial losses to establish an early position in a potentially important market. This introduces a distinction between commercial experimentation and long-term economic sustainability.
Space Debris and the Risk of Orbital Congestion
Low Earth orbit is a finite and increasingly valuable operating environment.
Communications satellites, Earth-observation systems, scientific spacecraft, crewed missions, and other orbital assets already depend on access to particular orbital regions.
Introducing enormous numbers of additional satellites could make orbital management significantly more difficult.
Every spacecraft creates some level of collision risk. Even small fragments can damage satellites because orbital objects travel at extremely high relative velocities.
The Axios report highlights that satellites in low Earth orbit can travel at approximately 17,000 miles per hour. At such speeds, collisions can produce destructive impacts and generate debris that may remain in orbit for extended periods.
A large collision could create additional fragments, increasing the probability of further collisions. This cascading effect is commonly associated with the Kessler syndrome, a theoretical scenario in which growing orbital debris makes certain regions increasingly hazardous for spacecraft operations.
The report warns that debris fields could make parts of low Earth orbit unusable for years.
For orbital AI infrastructure, this risk has direct commercial consequences. A data center constellation depends on reliable access to orbital space, functioning communications links, and the ability to replace damaged or outdated equipment.
A congested orbital environment could increase insurance costs, complicate deployment schedules, and threaten the availability of computing services.
The issue also extends beyond the companies operating data centers. A collision or debris event could affect unrelated satellites and public infrastructure, creating risks that are shared across the space industry.
Environmental Impacts: Rocket Emissions, Ozone, and Atmospheric Warming
Orbital data centers are often presented as a potential alternative to energy-intensive terrestrial infrastructure. Their solar-powered systems could reduce dependence on electricity generated through conventional power grids.
However, the environmental assessment must include the entire deployment and operating lifecycle.
The report identifies several potential environmental consequences.
Rocket emissions: Large-scale deployment would require substantial launch activity. Rocket exhaust can release greenhouse gases and other pollutants, including black carbon, depending on the propulsion system and operating conditions.
Ozone depletion: Rocket exhaust and the atmospheric effects of re-entering satellites could contribute to chemical processes that affect ozone.
Atmospheric warming: Repeated launches and satellite re-entries could introduce additional materials and emissions into atmospheric layers, with potential implications for climate systems.
Satellite disposal: Large constellations would eventually generate substantial de-orbiting activity, requiring careful management of the materials released during atmospheric re-entry.
The fact that orbital data centers use solar energy does not automatically make them environmentally neutral. Their environmental footprint includes manufacturing, launch vehicles, communications infrastructure, replacement cycles, and disposal.
A meaningful comparison with terrestrial data centers must therefore examine lifecycle emissions rather than comparing only the electricity source used during operation.
Military and Geopolitical Implications of Orbital Computing
Large orbital data center networks could also become strategically important infrastructure.
AI computing supports intelligence analysis, communications, logistics, simulation, and other capabilities relevant to national security. Placing such resources in orbit could provide new forms of computing access, but it could also create additional vulnerabilities.
The Axios report notes that if the US military adopts orbital data center technology, these systems could become potential targets during future space conflicts.
A constellation designed to provide commercial computing services could therefore acquire strategic significance beyond its original business purpose.
Potential concerns include:
Disruption of communications links.
Interference with satellite operations.
Physical threats to orbital infrastructure.
Increased competition over orbital positions and access.
Risks to civilian services if commercial and military uses overlap.
The growing importance of space infrastructure makes international coordination increasingly relevant. Orbital computing would not operate independently of existing satellite networks, national space policies, and international responsibilities.
This is particularly important because the consequences of orbital congestion or debris generation can extend beyond the country or company responsible for launching a satellite.
Why Existing Space Governance May Not Be Enough
The Columbia-led report argues that current governance arrangements are fragmented and insufficient for constellations of the proposed scale.
Satellite licensing, orbital coordination, environmental oversight, and space traffic management involve different institutions and regulatory processes.
These arrangements developed around a space industry with much smaller numbers of spacecraft than those contemplated by some current data center proposals.
The researchers recommend a more cautious and structured approach.
Their proposals include:
Phased deployment: Companies should expand orbital constellations gradually rather than launching enormous fleets before their impacts are understood.
Annual license renewals: Regulatory approval should be reconsidered based on performance in areas such as brightness, de-orbiting, and emissions.
Protection of visible night skies: Regulators should refuse licenses for data-center constellations visible from Earth's surface.
Environmental review: The report recommends ending the FCC's categorical exclusion of satellite launches from National Environmental Policy Act review.
International coordination: The authors call for an international space traffic management organization capable of addressing the growing complexity of orbital activity.
The report also highlights SpaceX's own statement to the FCC that initial operations under its application would involve significantly fewer satellites and re-entries than the maximum authorized number.
The authors argue that companies should follow this principle strictly, using limited initial deployments to evaluate real-world performance before expanding.
Can Orbital AI Data Centers Become Commercially Viable?
The technology has potential advantages, but commercial viability depends on several interconnected factors.
Solar energy availability is attractive, but it must be evaluated alongside the mass and cost of power systems. Space offers a different thermal environment, but rejecting heat from high-performance processors requires specialized engineering. Global computing demand may be substantial, but serving customers from orbit introduces communications and latency considerations.
The business case also depends on hardware durability.
AI accelerators evolve rapidly. A terrestrial data center can replace processors, upgrade networking equipment, and modify its cooling infrastructure without launching an entirely new facility. Orbital computing systems face more complicated replacement and servicing requirements.
This creates a possible mismatch between the operational life of spacecraft and the commercial life of AI hardware.
The most promising applications may initially involve specialized workloads where orbital access, power availability, or strategic requirements justify the additional expense. Broad replacement of terrestrial data centers would require much stronger evidence of cost competitiveness and operational reliability.
The industry should therefore distinguish between the technical possibility of operating computers in space and the economic practicality of deploying them at enormous scale.
The AI Infrastructure Race Must Not Come at the Cost of Earth's Night Sky
Space-based AI data centers represent an ambitious attempt to overcome some of the physical constraints associated with expanding artificial intelligence infrastructure. Solar power, orbital deployment, and large-scale computing platforms could create new opportunities for specialized services.
Yet the Columbia-led ICEF report demonstrates that the consequences could extend far beyond the companies building these systems.
Proposed constellations involving hundreds of thousands or even one million satellites raise serious questions about astronomical visibility, orbital congestion, collision risks, atmospheric pollution, and international governance. The estimated requirement of at least 700 Starship-sized launches annually to serve 10% of projected 2030 AI data center demand illustrates the transportation challenge.
The next phase of orbital computing should be guided by measurable performance, transparent environmental assessment, and coordinated regulation. Phased deployment can help determine whether the benefits justify the costs before irreversible changes occur across the orbital environment.
For Dr. Shahid Masood and the expert team at 1950.ai, this debate highlights a critical dimension of the AI revolution: technological progress must be evaluated not only through computing performance and commercial potential, but also through its long-term consequences for shared global resources.
Key Takeaways
Orbital AI data centers could provide an alternative infrastructure model using solar-powered computing platforms.
A Columbia-led ICEF working paper warns that proposed constellations could make satellites more visible than stars across many locations and times.
Approximately 19,000 satellites currently orbit Earth, while full implementation of proposed data center plans could increase the low Earth orbit population more than 50-fold.
SpaceX has applied for permission to operate up to one million data-center satellites.
Starcloud, Blue Origin, and Orbital have also proposed large constellations.
Serving 10% of projected 2030 AI data center demand from orbit could require at least 700 Starship-sized launches annually, according to the report.
Large solar arrays and thermal radiators could increase satellite visibility and introduce additional infrared emissions.
Orbital congestion could increase collision risks and create debris fields that threaten other spacecraft.
Rocket launches and satellite re-entries could introduce additional environmental impacts.
Researchers recommend gradual deployment, stronger environmental oversight, annual licensing reviews, and international space traffic management.
Future Outlook
The development of orbital AI infrastructure will test the relationship between technological ambition and responsible space governance. Companies may continue exploring the commercial potential of space-based computing, but the scale of deployment should depend on demonstrated technical performance and credible assessments of environmental consequences.
The challenge is not simply to determine whether AI data centers can operate in orbit. It is to establish whether they can do so without undermining astronomy, increasing orbital hazards, or imposing environmental costs on the wider world.
As artificial intelligence drives demand for computing infrastructure, protecting the space environment and preserving access to the night sky will become increasingly important parts of the global technology conversation.
Further Reading / External References
Space data centers could leave more satellites than stars in the night sky, a Columbia-led report warns
Space data centers could end stargazing as we know it





Comments