Elon Musk is orchestrating a strategic move valued at $1.25 trillion. Last week, the tech magnate revealed an ambitious plan to merge SpaceX with his unprofitable AI startup, xAI. This goes beyond a mere financial maneuver; at its core lies a seemingly science-fiction business wager—that the future of AI lies not on Earth, but in orbit. Musk is convinced that within the next three years, vast satellite constellations powered by solar energy and cooled by the vacuum of space will become the cheapest source of computational power for generative AI. The critical question remains: Is this another compelling narrative crafted for a potential IPO, or an inevitable path the AI industry must take?
A "Stellar Computing" Race Among Titans Musk is not the only tech titan looking to the stars. While former xAI employees have criticized the idea as "completely unproven," Amazon founder Jeff Bezos is already a participant. He co-manages the mysterious AI startup "Project Prometheus" and predicted last October the future construction of "gigawatt-scale space data centers." Simultaneously, Google is actively involved, planning to launch "Project Suncatcher" in early 2027 in partnership with Planet. This project aims to send two prototype satellites equipped with TPU chips into low Earth orbit for "learning missions." Baiju Bhatt, founder of Aetherflux, succinctly noted that the AI boom has created a "once-in-a-century inflection point in energy usage." While terrestrial data centers grapple with power supply and regulatory approvals, space offers theoretically "unlimited energy." The logic appears sound: space provides free, constant solar power and an environment near absolute zero. Sending power-hungry GPUs into orbit and transmitting data back via Starlink seems a logical progression.
Deutsche Bank's Ledger: The 2030 Cost Tipping Point Capital markets are often the first to sense financial opportunity. Deutsche Bank's latest report provides economic underpinning for this concept. The current reality, however, is stark: deploying a 1-gigawatt data center in space costs seven times more than on Earth (approximately $114 billion vs. $16 billion). But Deutsche Bank analysts predict this cost gap will narrow at a remarkable rate, with a potential inflection point around 2032. Their model suggests that space data center costs could drop to levels "roughly comparable" to terrestrial ones if two key conditions are met: a dramatic reduction in launch costs—relying on reusable rocket technology like SpaceX's Starship to slash the cost per kilogram from around $1,600 to just $67—and extreme hardware optimization, reducing individual satellite costs below $2 million and developing chips specifically customized for the space environment. Achieving this would mean overcoming the energy and cooling bottlenecks that plague Earth's computing industry.
Engineers' Reality Check: The "Thermodynamic Prison" and Radiation Nightmare In contrast to the optimistic projections of financial analysts, aerospace engineers and physicists on the front lines present a markedly different view, highlighting critical physical challenges inherent in Musk's grand vision. Firstly, the notion of superior cooling is misleading. A common misconception is that space is cold and ideal for heat dissipation. This is incorrect. Space is a vacuum, lacking air for convective cooling, making fans useless. "Space is a thermodynamic prison," one industry commentator aptly described. The immense heat generated by millions of processors can only dissipate through thermal radiation. This means a 5,000-megawatt facility might require approximately 16 square kilometers of radiator panels—a fragile structure the size of a small town that needs maintenance. It's akin to protecting a vast sheet of aluminum foil in a battlefield filled with shrapnel. Secondly, space is filled with radiation and debris, necessitating expensive shielding. Cosmic and solar radiation constantly bombard electronics, causing data corruption or damage. Solutions involve either encasing servers in heavy lead "armor" (drastically increasing launch weight) or using costly, lower-performance "rad-hard" chips. Furthermore, low Earth orbit has become a "junkyard." Space debris traveling at 15,000 miles per hour, even the size of a marble, could cause catastrophic damage to large radiator panels.
Who Holds the Key? Geopolitical Concerns Beyond technology and cost lies a more serious question: who will control this infrastructure? If SpaceX's Starship program succeeds and monopolizes cheap space transport, and if U.S. tech giants dominate orbital energy and computing resources... Jermaine Gutierrez, a researcher at the European Space Policy Institute (ESPI), warns, "If you let all space-based infrastructure be dominated by Americans, that in itself is a risk." Under the U.S. CLOUD Act, American companies could be compelled to切断 services anywhere globally. For Europe, this risks creating another dependency, following its lag in cloud infrastructure behind Amazon AWS and Google, in this new "space computing" race. Himanshu Tyagi, co-founder of Sentient, stated bluntly that the real risk isn't superintelligent AI run amok from science fiction, but oligopoly. "When a small group controls multiple choke points—launch, communications, AI, robotics, and consumer platforms—you face an oligopoly that is difficult to regulate and compete against."
The Starting Gun Has Fired Musk's predictions are notoriously aggressive. As industry veteran Steve Collar noted, "Elon is always years ahead of reality, but that doesn't mean he's wrong." Whether space data centers represent the next trillion-dollar frontier or a mirage constrained by physics remains to be seen. The current economic case heavily relies on the extreme assumption of launch costs falling to $67 per kilogram. For investors, this is an area requiring extreme caution. Investing in more efficient nuclear or geothermal data centers on Earth today might represent "smarter" capital than building expensive radiators in space. Regardless, the race to migrate computing power to orbit has officially begun.