Analysts and experts point to major safety and technical hurdles with SpaceX CEO Elon Musk’s vision for a network of orbiting data centers powered by one million satellites.<\/b>
\nElon Musk claims to be serious about creating a network of orbiting data centers using one million satellites, but some industry analysts and space experts have their doubts.
„Feasible in what timeline?“ asked Lluc Palerm, a satellite research director at consulting firm Analysys Mason. „This seems more a long-term goal,\u201d he said, likening it to a mission to Mars.
Musk has been hyping up the one-million-satellites plan, betting that the AI gold rush belongs in orbit. „In the long term, space-based AI is obviously the only way to scale“, the SpaceX CEO said in a Monday post that announced SpaceX’s acquisition of xAI and cited the benefits of harnessing the sun’s energy through a network of satellites. SpaceX is preparing an IPO that\u2019s expected to raise as much as $50 billion in an effort to fund orbital data centers.
SpaceX argues that the next-generation satellite network will beat ground-based data centers on cost and energy efficiency. A key challenge facing orbiting data centers is the lack of air to cool a GPU in the vacuum of space. However, with one million satellites, SpaceX could make the data centers relatively small, and thus easier to cool, according to industry analyst Carlos Placido.
„Large, power-hungry processors require disproportionately large radiators. Rather than deploying fewer, massive satellites, SpaceX appears to be exploring a highly distributed architecture: many smaller AI nodes, each with modest compute power, interconnected by laser link,\u201d he wrote.
But some question whether we’d switch from one environmental hazard to another, given the number of rocket launches that would be required to send up one million satellites. Not to mention the clutter in Earth\u2019s orbit. The number is staggering considering its monumental jump (or about a 68x increase) from the 14,500+ satellites already in orbit, according to data from astronomer and satellite tracker Jonathan McDowell.
„I think it is going to be extremely difficult to operate such a huge number of satellites safely“, he said, noting that if you include plans from China, 1.7 million satellites have been proposed for the future. \u201cThis is a factor of 100 increase over the already large number extant today.“
The also means the increased risk of potential collisions between satellites and space debris. \u201cSpaceX will say they can do that station-keeping successfully, but it doesn’t take many failures to have you end up in a bad situation,\u201d McDowell added. In a worst-case scenario, a mishap could trigger the \u201cKessler syndrome,\u201d where Earth\u2019s low orbit has become so cluttered with satellites and debris that any pile-up could create a chain reaction of colliding material.
Hugh Lewis, a space debris expert and professor of astronautics at the University of Birmingham, added: \u201cI think these plans are somewhat optimistic, premature, and possibly na\u00efve from safety and sustainability perspectives.\u201d Although SpaceX has designed its existing Starlink satellites with thrusters so they can move out of harm’s way, a constellation of one million satellites increases the chance of satellite malfunctions and failures, which would result in more space debris, he noted.
Still, Lewis said: \u201cI suspect that this is indeed an attempt to build the profile before any public offering, but I also expect SpaceX to move forwards with the plan.\u201d This might include revising the approach when the company\u2019s current application with the Federal Communications Commission is light on certain technical details. Lewis expects the company\u2019s orbital data center proposal to face intense scrutiny along with vocal opposition.
Currently, SpaceX retires aging Starlink satellites by letting them „de-orbit“ and burn up in the Earth’s atmosphere, which has raised questions about the atmospheric impact. But to decommission the data center satellites, the company is proposing moving at least some of them into higher orbits, or on a path to orbit the Sun.
Musk has pointed to the vastness of space while the company’s FCC proposal teases a „future hardware recycling and material harvesting“ for the retired satellites. However, Lewis says: „Scattering space junk into the solar system doesn\u2019t seem to me to be environmentally responsible.“Can Starship Get It Done?
A major unknown is the timeline for deployment. The project\u2019s success hinges on Starship, SpaceX\u2019s next-generation heavy-lift rocket, which has also been designed to send humans to Mars. The vehicle is still going through flight testing, but SpaceX is hoping it\u2019ll start deploying V3 Starlink satellites later this year.
But satellite industry analyst Tim Farrar questions if the company will have rocket capacity into 2027 and 2028, given that Starship is also needed to send astronauts to the Moon for NASA\u2019s Artemis program.
\u201cOf course, there might be some test satellites up there sooner,\u201d according to Farrar, who laid out a scenario where SpaceX can launch Starship rockets on a weekly cadence, paving the way for it to send up 1,000 data center satellites into orbit by the end of 2028. \u201cThat\u2019s certainly not enough for space-based data centers to account for more than a tiny fraction of total data center capacity, let alone to make them cheaper than terrestrial options,\u201d he added.
Although Lluc Palerm at Analysys Mason estimates it\u2019ll take years to develop the constellation, he could see SpaceX leveraging its existing business with Starlink to create test satellites for the data centers. \u201cIt is possible that they launch some sort of early version very soon. They have the capacity to innovate very rapidly, [and] they have their own internal development and manufacturing capabilities. It would be very easy for them to have a test satellite launched in parallel with one of their Starlink launches,\u201d he said. \u201cBut scaling the constellation is a very different story.\u201d
Palerm noted the data center constellation is dependent on Starlink to route data and communicate with users below. \u201cSo they really need to keep scaling Starlink, otherwise this will become a bottleneck,\u201d he added. Getting Beyond Strategic Storytelling
Meanwhile, satellite technology consultant Christian Freiherr von der Ropp says that SpaceX\u2019s plan belongs \u201cmore to the realm of speculative vision than near- or mid-term engineering reality.\u201d He pointed to other hurdles such as shielding the data centers from radiation and the economics when it\u2019s easier to maintain and upgrade ground-based data centers.
\u201cLaunch, replacement, and station-keeping costs for vast numbers of high-power, short-lifetime compute platforms (in space) would likely exceed the cost of building and operating ground-based data centers supplied by cheap renewable energy,\u201d he said.
Still, von der Ropp suspects Musk\u2019s ambitions are also about marketing. \u201cProposals of million-satellite ‚data centers in space‘ appear far more aligned with long-term narrative building than with executable infrastructure planning. It is reasonable to interpret such visions as strategic storytelling that highlights technological ambition and future AI compute potential,\u201d he said.
Indeed, Musk has long talked about turning science-fiction concepts into reality. His Monday post about the project even mentions colonizing the Moon to help manufacture the data center satellites. He and his company also views the one million satellites as a stepping stone toward humanity becoming „a Kardashev type-2 level civilization“, where our technology has advanced to the point we can fully harness the entire energy output of the sun.
„My estimate is that within two to three years, the lowest cost way to generate AI compute will be in space“, Musk wrote in a Monday post announcing the SpaceX merger with xAI. „This cost-efficiency alone will enable innovative companies to forge ahead in training their AI models and processing data at unprecedented speeds and scales, accelerating breakthroughs in our understanding of physics and invention of technologies to benefit humanity.“<\/p>\n