SpaceX’s Orbital AI Gambit: A $1.75 Trillion Bet on Computing Beyond the Atmosphere
(SeaPRwire) –
By: Lucas Caldwell
SpaceX is pivoting from a launch provider to a celestial data center operator. The company is betting that the physical constraints of Earth-based power grids will force the next generation of AI compute into orbit. This is not just about satellites; it is a fundamental shift in where we house the digital brain of the future. By leveraging the existing Starlink V3 architecture, Musk is attempting to bypass the terrestrial energy crisis that currently throttles massive AI clusters. The ambition is to reach 1 gigawatt of space-based computing power by the end of 2027, a target that sounds like science fiction until you look at the company’s track record of rapid iteration.
The technical blueprint centers on the AI1 satellite. Each unit is designed to deliver 150 kilowatts of computing power, roughly equivalent to a single Nvidia GB300 rack. These satellites will rely on solar energy for power and use the vacuum of space for cooling, eliminating the massive HVAC overhead required by ground-based facilities. SpaceX engineer Ian Dahl notes that these units are actually simpler than current Starlink hardware because they lack large broadband antennas. The company has already petitioned the FCC for approval to deploy up to 1 million of these AI-capable satellites into low-Earth orbit.
Production is already moving from the drawing board to the factory floor. The company is establishing the Terafab facility in Bastrop, Texas, to manufacture its own AI chips in partnership with Tesla and Intel. This vertical integration is the core of the strategy. By controlling the chip design, the satellite manufacturing, and the launch vehicle—the Starship rocket—SpaceX aims to slash the cost of orbital deployment. The company is currently eyeing a $1.75 trillion valuation in an upcoming IPO, positioning this infrastructure play as the primary driver for its future revenue growth.
The industry is divided on whether this is a viable path or a massive overreach. Critics like Jeff Bezos and teams at Blue Origin point to the prohibitive costs of both chip manufacturing and heavy-lift launches. Independent models suggest the current economics of space-based compute do not yet pencil out against the efficiency of terrestrial data centers. Musk himself has advised taking the 2027 gigawatt target with a grain of salt, acknowledging the technical hurdles ahead. The company’s own IPO filing is more conservative, projecting the first launches in 2028 with revenue streams appearing only toward the end of the decade.
This move represents a high-stakes game of industrial chess. If SpaceX succeeds, it effectively creates a sovereign compute layer that exists outside the jurisdiction of terrestrial power grids and local regulatory bottlenecks. It forces competitors to either invest in their own space-based infrastructure or remain tethered to the increasingly expensive and power-constrained Earth surface. The move also secures a massive captive market for Starship, ensuring that the rocket has a permanent, high-value payload to carry into orbit. It is a classic move to create a new market where the company holds all the infrastructure cards.
The success of this entire orbital AI strategy hinges entirely on whether the cost of launching a gigawatt of compute can drop below the cost of building a gigawatt of terrestrial data center capacity.
Author bio: Lucas Caldwell, a tech opinion leader with millions of followers on X/Twitter, specializes in analyzing the intersection of aerospace, semiconductor supply chains, and emerging infrastructure markets.