🎧 Listen to this article
By Tech Space Desk
In a detailed briefing, SpaceX founder Elon Musk and Starlink engineer Ian Dahl laid out a sweeping vision to move artificial intelligence infrastructure into orbit, build what they term a “TeraFab” semiconductor megafactory, and eventually manufacture and launch satellites from the Moon using electromagnetic mass drivers. The discussion framed these engineering goals not merely as commercial expansions, but as a civilizational imperative to advance humanity up the Kardashev scale.
The Kardashev Framing
Musk anchored the presentation in the Kardashev scale, the theoretical measure of a civilization’s energy harnessing capability. Type I denotes a society using all available energy on its home planet; Type II harnesses the bulk of its star’s output; Type III commands galactic-level power. By Musk’s assessment, contemporary civilization is effectively “non-registered” on the scale.
“The sun is about 99.86% of all mass in the solar system,” Musk noted. “We currently use much less than a trillionth of the power output of the sun.” To become what he described as a “respectable civilization” worthy of eventual contact with extraterrestrial observers, humanity must begin capturing a meaningful fraction of stellar energy. The only path to do so, he argued, is to move power generation and compute off-world.
Three Limiting Factors
The speakers identified three bottlenecks to orbital-scale infrastructure: mass-to-orbit capability, orbital power and thermal management, and semiconductor supply.
1. Mass to Orbit: Starship
Central to the plan is Starship, which SpaceX describes as the first rocket design capable of full and rapid reusability. Currently, SpaceX delivers approximately 85% to 90% of all Earth mass to orbit via Falcon 9 and Falcon Heavy, a figure Musk said totals roughly 2,500 tons annually. Starship is intended to change the unit economics entirely.
“Starship is the first rocket where [rapid reusability] is possible,” Musk said. The vehicle is already the most powerful flying object ever built, with the upcoming Block 4 version expected to produce nearly three times the thrust of the Saturn V. The company is targeting a turnaround time that eventually allows flights more than once per hour, and aims to reach an annualized rate of “millions of tons per year to orbit” within approximately three years.
2. Orbital Power and the AI-1 Satellite
Dahl and Musk showcased a draft design for the company’s first dedicated orbital AI satellite, internally referenced as AI-1. Rather than launching terrestrial data-center buildings, the satellite is essentially a rack-scale compute node wrapped in solar arrays and radiators.
The initial reference design targets 150 kilowatts of peak power and roughly 120 kilowatts of sustained average compute—roughly equivalent to an Nvidia GV300 rack. The satellite would use solar arrays sized around 250 watts per square meter and double-sided radiators capable of rejecting 1,400 watts per square meter. Musk emphasized that much of this technology derives directly from the Starlink V3 program, arguing that an AI satellite is “actually much simpler than a Starlink satellite” because it lacks complex phased-array and parabolic antennas.
Operating at an altitude of roughly 600 to 800 kilometers, the latency to orbit would be approximately three milliseconds, with data relayed via terabit-class laser links to the Starlink constellation and then to Earth. The speakers noted that even constellations numbering in the hundreds of thousands would not crowd low Earth orbit, given the volume of space available.
Production of the AI satellites and their solar arrays is planned for Bastrop, Texas, where SpaceX already manufactures Starlink user terminals. The company expects solar and AISAT production lines to be operational at “reasonable volume” by the end of next year.
3. The Chip Gap and the TeraFab
Even with launchers and satellites, Musk argued the global chip industry is on a trajectory to plateau around 100 gigawatts of annual AI compute—an order of magnitude short of what orbital infrastructure would eventually demand. To bridge the gap, SpaceX is planning what it calls the TeraFab.
The proposed facility would span roughly 100 million square feet, or about ten times the footprint of Tesla’s Gigafactory Texas. Its target: produce enough logic dies to output a terawatt of AI compute annually, translating to approximately one billion “full radical equivalent” chips operating at one kilowatt each, plus associated memory.
Reference designs for the orbital compute nodes include current and next-generation Nvidia processors (GB300 and Rubin) as well as TPUs, though the long-term goal is to scale custom and commercially available silicon indiscriminately.
An Aspirational Timeline
Musk repeatedly cautioned that the timeline was a “best guess,” not a promise. The company’s stated internal targets include reaching an annualized rate of one gigawatt of space-based AI compute by the end of next year, scaling to 10 gigawatts within roughly two and a half years, 100 gigawatts in three and a half years, and eventually a terawatt per year—an output roughly double the current total electricity consumption of the United States.
The Lunar Mass Driver
To advance another three orders of magnitude beyond a terawatt, Musk said the only viable path is manufacturing on the Moon. Because the Moon lacks an atmosphere and has one-sixth of Earth’s gravity, satellites could be launched into deep space using a linear electric motor—an electromagnetic mass driver—without chemical rockets. Under this long-term vision, the Moon would produce photovoltaics and radiators locally, potentially importing only the most advanced logic dies from Earth.
Conclusion: What the Next Five Years May Hold
If even a fraction of this roadmap materializes on the timeline described, the next five years would represent a phase change in the space economy. Starship’s transition from experimental flights to rapid reusability would collapse launch costs and shift the bottleneck from “how to lift” to “what to build.” Orbital AI compute would cease to be a speculative concept and become a measurable utility, potentially reconfiguring terrestrial energy markets as data-center demand increasingly migrates to space-based solar and radiative cooling.
However, the hurdles are commensurate with the ambition. Achieving a gigawatt of orbital compute within roughly a year would require flawless execution in Starship operations, satellite manufacturing at unprecedented speed, and the resolution of regulatory and orbital-debris questions that accompany megaconstellations. The TeraFab, meanwhile, implies capital expenditures and supply-chain demands that dwarf any previous private manufacturing project.
Yet the strategic signal is unambiguous. SpaceX is explicitly positioning itself not merely as a launch provider or a satellite internet company, but as the primary architect of orbital heavy industry. Within five years, the world will likely know whether that transition is achievable—or whether the Kardashev climb remains a longer march than even its chief proponent anticipates.
Disclaimer: This blog post was automatically generated using AI technology based on news summaries. The information provided is for general informational purposes only and should not be considered as professional advice or an official statement. Facts and events mentioned have not been independently verified. Readers should conduct their own research before making any decisions based on this content. We do not guarantee the accuracy, completeness, or reliability of the information presented.