Elon Musk’s ambitious initiative to deploy one million artificial intelligence data center satellites by 2028 might lead to a financial disaster, potentially putting SpaceX on a precarious path, warn leading experts in North American space studies.
This venture appears to follow in the footsteps of SpaceX’s existing Starlink project, which has already launched 10,000 satellites to provide broadband services. According to Robert Zubrin, a renowned rocket designer, each of these satellites carries a construction and launch price tag of approximately $2 million.
Utilizing SpaceX’s current assembly and launch technology to send one million AI satellites into low Earth orbit could potentially demand a staggering $2 trillion investment. This figure coincidentally matches the anticipated market valuation of SpaceX, projected during its forthcoming initial public offering (IPO).
If Musk proceeds with this colossal vision of creating a vast network of space-based AI data centers, it could abruptly end his series of business successes, Zubrin indicated in an interview.
Dr. Zubrin, who played a pivotal role in designing an early version of NASA’s Space Launch System Moon rocket, which recently facilitated a mission around the moon for four astronauts, pointed out that deploying satellites for hosting advanced AI models like OpenAI’s ChatGPT and Google’s Gemini would be significantly more costly than traditional land-based data centers.
In the lead-up to SpaceX’s IPO, the company’s leader published a mission statement online, highlighting the plan to “launch a constellation of a million satellites functioning as orbital data centers.” This initiative aims to support “AI-driven applications for billions today and secure humanity’s future as a multi-planetary species.”
Aerospace engineer Robert Zubrin designed an early prototype of the Space Launch System super-rocket, which just sent four Allied astronauts on a trek around the Moon. (Photo by Jim WATSON / AFP via Getty Images)
AFP via Getty Images
During the countdown to the SpaceX IPO and listing, its commander-in-chief posted a mission statement on the outfit’s website stating: “Launching a constellation of a million satellites that operate as orbital data centers” would support “AI-driven applications for billions of people today and ensuring humanity’s multi-planetary future.”
Promising to launch its next-generation Starship super-capsule once every hour to rocket these satellites hundreds of kilometers above the Earth, Musk predicted: “My estimate is that within 2 to 3 years, the lowest cost way to generate AI compute will be in space.”
While SpaceX recently completed the 12th flight test of its Starship, with the upper stage making a picture-perfect splashdown in the Indian Ocean, its first-stage rocket made an unscheduled kamikaze dive into the Gulf of Mexico after its 33 engines failed to reignite for a boost-back burn.
SpaceX has led a remarkable rocket revolution with its reusable Falcon 9 booster, but its colossal Starship is still undergoing a series of flight tests. (Photo by Joe Raedle/Getty Images)
Getty Images
Musk ultimately aims to recover both stages, via titanic robotic arms attached to the launch tower, and slash launch prices in the process.
NASA’s leaders initially commissioned SpaceX, back in 2021, to shuttle their astronauts from a Moon-orbiting capsule down to the lunar South Pole. But a cascade of delays in demoing the Starship in orbit caused them to reopen the lunar landing competition, and SpaceX is now facing off against Blue Origin in a race to complete a lunar touchdown spacecraft by 2028.
SpaceX’s chief designer prophesied in his Web-posted missive that: “The capabilities we unlock by making space-based data centers a reality will fund and enable self-growing bases on the Moon, an entire civilization on Mars and ultimately expansion to the Universe.”
Dr. Zubrin told me SpaceX has led a revolution in the design of reusable spacecraft with its Falcon 9 rocket.
SpaceX is now the global superpower in terms of total rocket launches, with an average of three lift-offs per week in 2025, more than NASA or any other government space agency.
But to skyrocket from that flight cadence to one launch every hour, or 8700 flights every year, with the still unproven Starship by 2028 is unimaginable.
Likewise, Zubrin says, “Launching a million satellite orbital data center constellation is fantasy.”
Dr. Zubrin is not only at the global forefront in aeronautical engineering, but also founded the Mars Society, a planet-spanning think tank and design studio to promote terraforming and colonizing the Red Planet.
He has known Elon Musk for a quarter century, since Musk made a contribution to the Society and was briefly named to its board of directors, before he founded SpaceX with the goal of speeding the first human explorers to the Martian dunes.
Since then, Zubrin has been a de facto mentor to Musk, trumpeting his spaceflight triumphs and his goal of creating the first hyper-tech cosmopolis on Mars, but also warning against strategic missteps that could jeopardize that quest, like the new quixotic, out-of-the-blue scheme to rocket generative AI agents to circle the Earth.
The American space seer says he wrote an appeal to Musk, published in a magazine article, to caution him against the cosmic boondoggle of constructing a mega-cluster of space-based data centers that could not compete economically with terrestrial counterparts.
To an outside observer, especially neophytes to the spaceflight sector, Musk’s masterplan for the super-constellation might seem like a logical progression after he assembled the rings of Starlink satellites that now transmit internet connections to 10 million people across the face of the Earth.
SpaceX’s master architect, Zubrin writes in the appeal, “takes things step by step, with each step providing the ample financial basis that enables the next.”
“He helped create PayPal, which gave him the funds to launch SpaceX.”
“The SpaceX Falcon 9 became a money maker, which enabled him to launch Starlink.”
“Starlink has not only made him even richer, it has made him far more powerful than any other businessman on Earth.”
SpaceX’s Starlink satellites have not only helped skyrocket Elon Musk across the pantheon of space tycoons, but also have “made him far more powerful than any other businessman on Earth.” (Photo by: Alan Dyer/VW Pics/Universal Images Group via Getty Images)
VW Pics/Universal Images Group via Getty Images
But the new scheme to build sensationally costly and colossal bands of AI satellites, Zubrin adds, could lead to SpaceX’s financial crash and burn.
Musk proposes powering these super-armadas of satellites with solar panels.
But Zubrin, echoing a host of other eminent space scholars, says that now, and for the foreseeable future, “space solar power is vastly more expensive than solar power on Earth, and even that is not fully competitive with fossil fuels, hydro, or nuclear.”
The solar panels on each SpaceX Starlink satellite generate 20 kilowatts of electric power. The $2,000,000 price tag on each of these spacecraft means one kilowatt of power costs $100,000, Zubrin says.
In contrast, he adds, solar arrays positioned on rooftops scattered across the planet produce electricity that costs about $3000 per kilowatt, while power from gas-fired generators costs roughly $1000 per kilowatt.
Zubrin, who holds a doctorate degree in nuclear engineering, says commercial nuclear power plants produce power at $5000 to $10,000 per kilowatt.
Any of these terrestrial power options, he points out, would be far cheaper than the solar wings that Elon Musk has proposed to fly his AI spacecraft.
While the Starlink satellites have to be sent into orbit to beam internet connections to users across the continents, Zubrin says, there is no comparable need to send AI data centers into space, especially one that would justify the sky-high price tag on each spacecraft.
Brian Hurley, founder of the influential New Space Economy think tank and digital magazine, told me in an interview that top researchers worldwide on the potential launch of orbital data centers customized to host AI models have reached a consensus that these stations will not become economically feasible until the mid-2030s.
A world-leading expert who chronicles the rapid-fire expansion of the modern space sector, and its rippling effects across the spheres of national security and international affairs, Canada-based Hurley points to a just-published study titled “Towards a future space-based, highly scalable AI infrastructure system design,” co-authored by nine vanguard Google scholars on AI and spaceflight, that predicts these AI space stations might only become competitive with terrestrial data centers around the middle of the next decade.
“High launch costs have historically stymied efforts to harvest solar power in space on large scales,” the Google scholars say.
Researchers at Google, creator of the Gemini artificial intelligence model, are now working on a “moonshot” exploration of launching AI labs into orbit in the future (Photo by Jonathan Raa/NurPhoto via Getty Images)
NurPhoto via Getty Images
Yet “there is a feasible path for launch costs to drop sufficiently to no longer be prohibitive.”
If SpaceX succeeds in making the Starship fully reusable, the Google researchers predict, and the ship’s average launch cost to low Earth orbit ultimately plummets to $200/kilogram, “a space-based system could achieve performance roughly comparable to a terrestrial datacenter.”
They also forecast that if SpaceX reaches a cadence of “∼180 Starship launches/year, launch prices could fall to <$200/kg by ∼2035.”
These leaders of Google’s new Project Suncatcher describe their fascinating research as “a moonshot” exploring the future of space-based AI outposts.
“If AI is a foundational general-purpose technology,” they say, “we should anticipate that demand for AI compute — and energy — will continue to grow.”
“This work explores a scalable compute system for machine learning in space, using fleets of satellites equipped with solar arrays, inter-satellite links using free-space optics, and Google tensor processing unit [AI] accelerator chips.”
The next step in this frontier research, they add, will involve a small-scale demonstration mission in an alliance with the vanguard imagery satellite operator Planet Labs. These allies aim to launch twin prototype satellites early next year to test their models and the deployment of super-speed inter-satellite links for an orbiting machine learning lab.
Leading-edge researchers at Google and at the cutting-edge imagery satellite operator Planet Labs are teaming up to launch twin prototype satellites early next year for an orbiting machine learning lab. (Photo By Michael Macor/The San Francisco Chronicle via Getty Images)
San Francisco Chronicle via Getty Images
Space scholar Brian Hurley says the tech demo being sent into orbit by frontier AI researchers at Google and the planet’s leading observation satellite operators at Planet Labs shows “orbital AI data centers may become technically demonstrable within a few years.”
But “economically meaningful deployment at scale is more likely a decade-plus proposition.”
“The cost of orbital infrastructure would need to be on par with, or better than, terrestrial infrastructure for the relevant workloads,” he says. “That does not mean matching only the cost of electricity or only the cost of launching hardware.”
“It means matching the full delivered cost of computation, including capital expenditures, operating costs, refresh cycles, reliability, financing, insurance, data transport, maintenance, customer acquisition, and risk.”
“Terrestrial data centers are expensive,” he adds, “but they benefit from mature supply chains, ground access, utility-scale power contracts, equipment replacement options, service crews, competitive fiber networks, established financing models, and decades of operational learning.”
Orbital infrastructure starts out with much steeper costs.
To become competitive with existing ground-based data centers scattered across the globe, he says, celestial stations would depend on nose-diving launch prices and “extremely high flight rates.”
At the same time, “satellite manufacturing cost would have to drop through mass production, design standardization, and high-yield assembly.”
All of these changes are likely to require at least a decade to materialize.
Dr. Zubrin says that jumping the gun by lofting a colossal constellation of orbital AI stations now, with still high launch costs, would swiftly drain the coffers of the station operators.
Yet he adds the idea of combining vanguard breakthroughs in artificial intelligence with those in leading-edge rocketry is now in vogue across the tech sector.
Some crystal-gazing investors, he told me, see AI as sparking a new tech gold rush, similar to the internet-powered investment rush of a generation ago.
For the foreseeable future, he predicts, the contending Titans in the kingdom of AI will promise untold fortunes and attract cascades of funding.
The internet boom, he told me, marked the super-speed rise of Google and Yahoo and an army of contenders vying to craft the perfect portal to search the limitless treasure houses of info that appeared across the Web.
During that era, he says, there was a mass sense that “there was gold in those hills.”
“It was question who was going to get it.”
Today, he adds, history is repeating itself, but across the next-generation realm of AI.
“There clearly is gold in the hills of artificial intelligence,” with a new race to become the ultimate prospector.
Musk can tap this exploding gold fever across the spheres of AI and rocketry, Zubrin predicts, “in order to amplify his IPO.”
“I think the thing that’s driving him right now is the desire to make the SpaceX xAI IPO a big success – I think that’s what he’s trying to do.”
“He’s calculating that people are looking at this and saying well I don’t know if this is really going to work but you know no one’s ever lost money betting on Elon Musk.”
