SpaceX Unveils Updated Vision for Orbital AI Data Centres at Global Space Congress

SpaceX has offered the public a fresh glimpse of its ambitious concept for placing artificial intelligence data centres in orbit, presenting a new visualisation of the project at the 2026 International Astronautical Congress in Turkiye. The presentation, reported by Al Jazeera's NewsFeed segment, centres on a proposal to deploy as many as one million satellites in space to power AI computing infrastructure — a scale that would dwarf anything currently operating in Earth orbit.

The reveal took place at one of the aerospace industry's most closely watched annual gatherings, where space agencies, private companies and research institutions converge to showcase their latest programmes. The International Astronautical Congress has historically served as a stage for major announcements, from deep-space exploration plans to next-generation launch systems. SpaceX's decision to use the event to share an updated visualisation of its orbital data centre concept signals that the company wants the idea to be taken seriously by the wider space and technology communities.

At the heart of the proposal is a simple but radical premise: rather than building more energy-hungry data centres on the ground, SpaceX envisions processing artificial intelligence workloads in space using a vast constellation of satellites. The company has previously argued that orbital facilities could benefit from near-continuous solar power and the natural cooling environment of space, potentially easing some of the constraints that terrestrial data centres face.

The figure of one million satellites is the most striking element of the plan. For context, the total number of active satellites currently in orbit around Earth is estimated to be in the low tens of thousands, with SpaceX's own Starlink broadband network accounting for a substantial share of that total. A constellation of one million spacecraft would represent an increase of several orders of magnitude over the entire historical record of satellite launches, raising immediate questions about manufacturing capacity, launch cadence and orbital traffic management.

Data centres on the ground are coming under growing strain as demand for artificial intelligence accelerates. Training and running large AI models requires enormous amounts of electricity, and cooling those facilities consumes vast quantities of water. Technology companies have increasingly explored alternatives, including locating servers in cold climates, submerging them in specialised liquids, and purchasing dedicated nuclear or renewable power supplies. The orbital concept represents an attempt to move the entire problem off the planet's surface.

SpaceX is not alone in exploring the idea. Several startups and research groups have floated proposals for space-based computing, arguing that solar energy collected in orbit is uninterrupted by weather or night cycles, and that heat can be radiated away into the vacuum of space. However, engineers have long cautioned that radiating heat in a vacuum is far more difficult than cooling equipment in an atmosphere, and that servicing or replacing hardware in orbit remains expensive and logistically complex.

The regulatory dimension of the proposal is equally daunting. The International Telecommunication Union coordinates radio spectrum and orbital slots for satellite operators, and national regulators license individual spacecraft. A constellation of one million satellites would test those frameworks in ways they were never designed to handle. Astronomers have also raised repeated concerns about the impact of large satellite constellations on ground-based observations, warning that bright spacecraft can streak across telescope images and interfere with scientific research.

SpaceX's Starlink network has already drawn scrutiny over its effect on the night sky, prompting the company to experiment with darkening coatings and other mitigation measures. A vastly larger constellation dedicated to computing rather than communications would likely intensify those debates, particularly if it involved satellites in low Earth orbit, where they are most visible from the ground.

The timing of the presentation is notable. SpaceX continues to expand its launch capabilities, and its Starship vehicle is designed to carry very large payloads at lower cost per kilogram than earlier rockets. Supporters of orbital infrastructure argue that such heavy-lift capacity is a prerequisite for any project on this scale, since assembling and maintaining a million-satellite system would require an unprecedented number of launches.

Critics, however, point out that no public timeline, cost estimate or technical roadmap accompanied the new visualisation. The presentation appears to be conceptual in nature, illustrating the company's long-term ambitions rather than announcing a scheduled programme. That distinction matters for policymakers, investors and rival firms trying to assess how soon — if ever — orbital AI computing could become commercially viable.

The involvement of artificial intelligence as the intended workload also reflects the broader economic moment. AI has become the dominant driver of demand for computing power worldwide, reshaping investment priorities across the technology and energy sectors. If a credible path to space-based AI infrastructure emerged, it could upend assumptions about where data is processed and who controls the underlying hardware.

For the space industry, the proposal reinforces a shift in which orbital assets are viewed not only as tools for communication, observation and navigation, but as industrial platforms in their own right. That framing has attracted both enthusiasm and scepticism, with some analysts warning about the risk of concentrating critical infrastructure in orbit and others celebrating the potential for new markets.

Al Jazeera's report did not include comments from SpaceX executives or independent experts on the technical feasibility of the plan, and the company has not published detailed specifications. The NewsFeed segment focused on the unveiling of the visualisation itself, describing it as a new look at SpaceX's proposal to place a million satellites in space for AI data centres.

As the concept stands, it remains firmly in the realm of ambition rather than execution. Whether orbital data centres become a genuine component of the AI economy or remain a striking illustration of what private spaceflight companies like to imagine will depend on engineering breakthroughs, regulatory accommodation, launch economics and public tolerance for an increasingly crowded sky. For now, SpaceX has given the world another image of the future it wants to build — and a new round of questions about how, and whether, it can be built at all.