In a new 26-minute Forbes interview, Elon Musk laid out one of his most ambitious long-term visions yet: moving AI computing infrastructure off Earth entirely. He argued the concept is far more achievable than most people assume — and that SpaceX already has the foundation in place to make it happen. Here's what he said, and what it actually means.
Why would you put AI datacenters in space at all?
The core argument is energy and scale. Ground-based datacenters face real-world constraints: land acquisition, power grid capacity, cooling infrastructure, and regulatory hurdles. In space, solar energy is abundant and continuous — no weather, no day/night cycle interruption. The vacuum of space also solves cooling, one of the biggest operational costs for any large compute facility. For AI workloads that demand enormous, sustained power, the physics of space actually work in your favor once you can get hardware up there affordably.
What did Musk actually say in the interview?
His exact quote, shared by Sawyer Merritt, was direct: "Datacenters in space is much easier than people may think. SpaceX has 10k satellites in orbit right now. In the future with Starship, we'll be launching over 10k per year." He then connected it to the Kardashev scale — the theoretical framework for measuring a civilization's energy consumption and technological advancement. The implication is clear: space-based computing isn't a novelty project, it's infrastructure for a multi-planetary civilization.
What is the Kardashev scale and why does it matter here?
Proposed by Soviet astronomer Nikolai Kardashev in 1964, the scale classifies civilizations by their total energy use. A Type I civilization harnesses all energy available on its home planet. Type II captures the full output of its star. Type III controls energy at a galactic scale. Earth currently sits well below Type I. Musk's point is that climbing this ladder requires moving beyond Earth's energy budget — and space-based computing infrastructure is a logical step in that direction. It reframes the datacenter question from "is this practical?" to "is this necessary for long-term survival?"
How does SpaceX's existing Starlink constellation make this feasible?
SpaceX has already demonstrated it can deploy and operate satellites at a scale no other organization has matched. According to verified data, as of April 2026, SpaceX has launched at least 11,749 Starlink satellites, with approximately 10,202 currently in orbit and around 10,193 confirmed operational. The FCC has also granted SpaceX authorization to deploy up to 15,000 second-generation Gen2 Starlink satellites. That operational track record — supply chain, launch cadence, on-orbit management — is directly transferable to deploying compute hardware in orbit. The logistics are already proven.
What role does Starship play in making this economically viable?
Launch cost per kilogram is the gating factor for any space infrastructure. Starship Version 3, scheduled for its debut flight around the time of this interview, is designed for a fully reusable payload capacity of over 100 metric tons to Low Earth Orbit — roughly three times the capacity of the previous Starship version, according to SpaceX. Future expendable configurations could reach 200 tons per flight. SpaceX's stated ambition is launches "every hour carrying 200 tons per flight," which would deliver millions of tons to orbit annually. Musk has indicated he expects SpaceX to achieve over 10,000 Starship launches per year by 2031. At that cadence and payload capacity, the economics of putting compute hardware in orbit shift dramatically.
Is this actually happening, or is it still theoretical?
It's somewhere between the two. Musk and the SpaceX team have stated they believe they have a viable approach for building orbiting datacenters, but no operational space-based AI compute facility exists yet. The enabling infrastructure — Starlink's constellation management, Starship's reusability — is being built in parallel. The honest read is that the foundational work is real and advancing, but the datacenter application remains a declared goal rather than an announced product. The timeline Musk points to — 10,000+ Starship launches per year by the early 2030s — is the trigger point where this stops being theoretical.
Whether or not the timeline holds, the underlying logic is harder to dismiss than it might initially seem. SpaceX has already put more hardware in orbit than any entity in history. The jump from satellite internet infrastructure to orbital compute infrastructure is a large one — but it's built on a foundation that genuinely exists.
Sarah focuses on Tesla Energy, SpaceX missions, and the broader Musk AI portfolio. Former data analyst in clean energy. Based in San Francisco.
Sources verified at publish time. Spotted an inaccuracy? Email editorial@basenor.com.
