By Yang Tianzi
Space-based AI: a radical answer to exploding compute demand
According to documents submitted by SpaceX to the U.S. Federal Communications Commission (FCC), the company is seeking approval to build what it calls an “orbital data center” network: up to one million satellites designed specifically to provide the massive computing power required for artificial intelligence.
If realized, the project would fundamentally redefine the concept of data centers, shifting them from Earth-bound warehouses filled with high-performance servers into outer space. SpaceX argues that this approach represents the most cost-effective and energy-efficient response to the explosive growth in AI computing demand.
From large language models to increasingly complex machine-learning systems, artificial intelligence now requires unprecedented levels of computation. Conventional data centers occupy vast physical space, consume enormous amounts of electricity, and generate extreme heat that must be managed through energy- and water-intensive cooling systems. SpaceX contends that terrestrial expansion is reaching its limits—and that space offers a virtually boundless alternative.
The proposed system would rely on solar-powered satellites capable, in theory, of delivering enough computing capacity to serve billions of users worldwide. Compared with SpaceX’s existing Starlink network—currently numbering close to 10,000 satellites—the leap to one million represents a qualitative transformation in scale.
For the best of our weekly content！
You are now signed up for our newsletter
Check your email to complete sign up
These satellites would operate in low Earth orbit, at altitudes ranging roughly from 500 to 2,000 kilometers, overlapping with the orbital range already occupied by Starlink. SpaceX maintains that despite the headline number, the vastness of space means individual satellites would remain widely dispersed, often too far apart to be visible from one another.
A step toward a Kardashev Type II civilization
In its FCC filing, SpaceX frames the project in explicitly civilizational terms, presenting it as a first step toward what it calls a “Kardashev Type II civilization.”
The Kardashev scale, proposed in the 1960s by Soviet astronomer Nikolai Kardashev, classifies hypothetical civilizations by their ability to harness energy. A Type I civilization can utilize all the energy available on its home planet; a Type II civilization can capture the total energy output of its host star. Humanity has not yet reached even Type I status.
SpaceX nevertheless positions the orbital AI network as a meaningful move toward higher levels of technological civilization—an articulation consistent with Elon Musk’s long-standing rhetoric about humanity’s future and its expansion beyond Earth.
The company further argues that orbital data centers would be more environmentally sustainable than their terrestrial counterparts. Cooling is one of the largest challenges for conventional data centers, consuming vast amounts of electricity and water. Globally, data centers already account for a significant share of total electricity use, a figure expected to rise sharply with the growth of AI.
Space offers natural advantages: heat can be dissipated through radiation in a vacuum, and solar energy provides a clean, nearly limitless power source. In theory, this model could reduce dependence on Earth’s finite resources and lower overall carbon emissions.
Environmental claims under scrutiny
These environmental arguments, however, face substantial skepticism.
Launching hardware into orbit on such a scale would itself require immense energy expenditure, generating significant carbon emissions during rocket launches. Satellite manufacturing, maintenance, and end-of-life disposal all carry complex and poorly quantified environmental costs.
Experts note that even with SpaceX’s reusable rocket technology—widely credited with reducing launch costs—the deployment of one million satellites would remain an extraordinarily expensive and resource-intensive undertaking. The claim that orbital data centers are inherently “green” remains, at best, unproven.
The technical challenges are formidable. Satellites must integrate robust protection, cooling, and power systems in an environment characterized by radiation exposure and the constant risk of micrometeoroid impacts. Hardware must be durable yet lightweight to minimize launch costs. Solar panels must deliver reliable, high-efficiency power within tight spatial constraints.
Data transmission presents another major hurdle. High-speed, stable communication links are required both between satellites and between satellites and ground stations—at a scale never before attempted.
Most concerning is the growing threat of space debris. As the number of low-Earth-orbit objects increases, so does the probability of collisions. A single impact can generate thousands of high-velocity fragments capable of destroying other satellites, potentially triggering a cascading chain reaction known as the Kessler syndrome—where debris collisions multiply until entire orbital regions become unusable.
SpaceX’s existing Starlink constellation has already intensified concerns about orbital congestion. An additional one million satellites would dramatically exacerbate these risks.
Astronomers sound the alarm
The global astronomy community has expressed deep alarm over the proposal. As early as 2024, astronomers warned that Starlink’s radio emissions were effectively blinding radio telescopes, severely impairing scientific research.
Satellites reflect sunlight, creating bright streaks across optical telescope images, and their communication systems emit radio signals that interfere with radio astronomy. With satellite numbers rising into the hundreds of thousands—or millions—such interference would increase exponentially.
For certain types of astronomical observation, the damage could be irreversible, rendering entire classes of research extraordinarily difficult or altogether impossible. Critics argue that this poses a serious threat to humanity’s ability to explore the universe and understand its place within it.
Elon Musk has dismissed these criticisms. Writing on X, the social media platform he owns, Musk has insisted that space is vast and that even one million satellites would not meaningfully crowd orbital space. He argues that the distances between satellites would remain enormous.
Critics counter that the issue is not the abstract size of space, but the density of objects in specific orbital bands and the cumulative effects over time. Congestion, they argue, is already visible—and accelerating.
An unsettled future for orbital infrastructure
SpaceX’s proposal remains at the application stage, with no public timeline for implementation. From technical feasibility and economic cost to environmental impact and international regulation, the plan faces a formidable array of unresolved challenges.
The project embodies a radically new vision for humanity’s use of space—one that pushes technological ambition to its limits. At the same time, it underscores the urgent need to balance innovation against the protection of the orbital environment and the preservation of scientific inquiry.