Elon Musk posted a single sentence on Wednesday that reframes how we should think about the future of computing: "The amount of compute in space will obviously round up to 100% of all compute." It reads like a throwaway prediction, but behind it sits a multi-year engineering program already moving through regulatory approvals, hardware development, and investor roadshows. Here's what's actually going on.
What exactly is SpaceX building here?
SpaceX is developing a constellation of orbital data centers — satellites purpose-built to run AI workloads rather than relay communications. The program centers on a satellite design codenamed AI1: 20 meters tall, 70-meter wingspan, capable of 150 kilowatts peak power with roughly 120 kilowatts available for compute. The technology is adapted directly from Starlink V3 hardware, including its solar arrays and thermal radiators. According to public filings and investor materials, the FCC's Space Bureau formally accepted SpaceX's application for this system on February 4, 2026 — a constellation of up to one million satellites classified as a non-geostationary orbit (NGSO) system for large-scale computing.
Why move data centers to orbit at all?
SpaceX's argument is fundamentally about physics and resource constraints on Earth. Terrestrial data centers are running into hard limits on electricity supply and heat dissipation — cooling a hyperscale AI cluster requires enormous amounts of water and land. In orbit, solar power is abundant and continuous, and the vacuum of space is an effective thermal sink. The AI1 design assumes 250 watts per square meter for solar collection and 1,400 watts per square meter for radiators — numbers that would be impractical to achieve at ground level. Operating at 600–800 km altitude also means roughly 3-millisecond latency, which is competitive with terrestrial intercontinental links for many AI inference tasks.
When does this actually start?
According to SpaceX's investor presentation at a JPMorgan-hosted event in June 2026 ahead of the company's IPO, the timeline is as follows: a dedicated orbital factory for manufacturing AI satellites is targeted before the end of 2026, with the first orbital AI data center deployments beginning in Q4 2027. SpaceX projects reaching "significant scale" by 2028. Current ground-based compute capacity already hit 1.4 gigawatts in Q2 2026, with a target to exceed 2 GW by year-end.
How large could this get?
The stated long-term targets are staggering. SpaceX aims for 10 GW, then 100 GW, and ultimately 1 terawatt per year of solar-powered AI compute in orbit. Musk's logic for the 1 TW figure rests on Starship's projected launch cadence: he estimates Starship could lift approximately 10 million tons of payload to orbit annually, which at AI1-class power density would enable terawatt-scale compute production. Nvidia is reportedly optimizing its Vera chips specifically for use in these orbital systems, which signals that the hardware ecosystem is already being built around the concept.
How does this connect to xAI?
SpaceX acquired xAI — Musk's AI company — as part of a broader vertical integration strategy. The goal is a single platform that owns the launch vehicles, the satellites, the compute infrastructure, and the AI models running on top of it. Orbital data centers become the physical backbone of that stack. Rather than renting GPU clusters from cloud providers, xAI would operate its own compute in orbit, launched on its own rockets, powered by its own solar arrays.
Is "100% of all compute" a serious claim or a rhetorical flourish?
Almost certainly both. The physics argument for space-based compute is real — unlimited solar power, no cooling water, no land constraints. But the practical path from today's 1.4 GW of ground-based capacity to a majority of global compute living in orbit involves solving manufacturing at a scale that doesn't yet exist, maintaining reliability in a radiation-heavy environment, and proving that orbital latency is acceptable for the workloads that matter most. Musk has a long track record of stating ambitious end-states as obvious conclusions while the engineering catches up over years. Whether the timeline holds, the direction of travel appears to be set — and the regulatory, financial, and hardware groundwork is already being laid.
Sources & reporting notes
The links below identify the material source records used for this report.
- @elonmusk on X (2026-09-24T18:22:56.000Z) — Direct source
Source links are preserved as published or accessed. See our editorial standards and corrections policy.
The BASENOR Editorial Desk covers Tesla, SpaceX, and related technology, curating reporting from primary sources — official accounts, regulatory filings, and software release data. Every article passes source-record and fact-checking review before publication. About the newsroom.
This report was curated by the BASENOR Editorial Desk from the sources listed above. Read our editorial standards or email editorial@basenor.com to report an error.
