2 Sources
2 Sources
[1]
Even the Sky May Not Be the Limit for A.I. Data Centers
Eli Tan reported from San Francisco and Ryan Mac from Los Angeles. If the architects of the artificial intelligence boom are right, it is only a matter of time before data centers -- the giant computing facilities that power A.I. -- will float in orbit and be visible in the night sky like planets. The science-fiction-like dream is being driven by A.I. and space industry leaders who are growing increasingly worried that data centers will eventually require more energy and land than are available on Earth. So one solution -- perhaps the only solution, they say -- is to start building them in space. Google announced in November that it was working on Project Suncatcher, a space data center project that would begin test launches in 2027. Elon Musk said at a recent conference that space data centers would be the cheapest way to train A.I. "not more than five years from now." Others pledging support for the idea include Jeff Bezos, the founder of Amazon and Blue Origin; Sam Altman, the chief executive of OpenAI; and Jensen Huang, the chief executive of Nvidia. "It is not a debate -- it is going to happen," said Philip Johnston, the chief executive of Starcloud, a space data center start-up. "The question is when." The notion has gained traction as the A.I. race hits a fever pitch, fueling fears of a potential bubble. Meta, OpenAI, Microsoft, Amazon and other big tech companies are investing hundreds of billions in data centers worldwide, with OpenAI alone committing $1.4 trillion to such projects. Saudi Arabia and other nations are also pouring money into these efforts, while smaller companies pile up debt and take on financial risks to join the frenzy. Yet earthbound data centers are increasingly running into limits. In many places, the projects lack enough available power for the computing needs. Local opposition has also flared over whether data centers are driving up utility bills and exacerbating water shortages. That has led to more creative -- some might say wishful -- thinking with space data centers. Technologists and scientists have researched the idea and concluded that some version of these projects may be possible in the next few decades. But skeptics said the proposals flew in the face of physics and would be astronomically expensive. Tech luminaries like Mr. Musk have also recently made comments about space data centers that are a magnitude larger than what current research suggests is possible, said Pierre Lionnet, a space economist and director at Eurospace, a trade association. "It's completely nonsensical," he said. The National Aeronautics and Space Administration introduced the idea of space data centers in the 1960s. In the 1980s, the concept of "data repositories" in space popped up in science fiction stories. In the last decade, the notion of space data centers that could power modern A.I. also emerged. The main benefit to building a data center in space is abundant energy, with nearly 24/7 access to the sun and no clouds to obstruct the project's solar panels, Mr. Johnston said. There are also fewer environmental regulations than on Earth, not to mention fewer neighbors to oppose the imposition or complain about electric bills. But the feasibility hinges on whether it will become cheaper to launch materials into space and whether technical issues like radiation and cooling can be solved in the meantime. Experts are split on how soon those conditions can be met. "As a business case, it's plausible," said Phil Metzger, a physics professor at the University of Central Florida and a former physicist at NASA. "It's been an evolving discussion." Data centers in space would look different from the football-stadium-size facilities on Earth. Most mock-ups from companies like Starcloud look like large satellites with a cluster of servers housing A.I. chips at the center of miles of solar panels to power them. The data centers would need to be rebuilt every five years, which is when the computer chips are typically replaced, Mr. Johnston of Starcloud said. They would be visible at dawn and dusk from Earth, he said, appearing in the sky as about a quarter the width of the moon. But it is too expensive to create space data centers today. A kilogram of material costs around $8,000 to launch into space, Mr. Lionnet said. The cheapest rate -- around $2,000 per kilogram -- is offered by the rocket maker SpaceX, he added. Individual server racks in a data center can weigh over 1,000 kilograms. If the space-launch costs fall to around $200 per kilogram, the economics will start to make sense, Dr. Metzger said. He predicted that would take around a decade. In a research paper about Suncatcher published in November, Google predicted the costs could decline to that level "in the mid-2030s," comparing the timeline to its driverless robot taxis, which took 15 years to develop. Others said they were not sure the costs would drop in such a short time. "It is like saying if we can get the cost of a McDonald's cheeseburger down to 10 cents, we will buy so many of them," Mr. Lionnet said. Modern computer chips and semiconductors are also not built to withstand the radiation in space, which would hurt their ability to compute reliably, said Benjamin Lee, an electrical and systems engineering professor at the University of Pennsylvania. And while space is a frigid minus 455 degrees Fahrenheit, it is also a vacuum. That means there is no air to transfer the heat from the A.I. chips. To cool the chips, the data centers would instead require large radiator panels to disperse the heat. Such hurdles have not stopped people like Mr. Musk, who leads SpaceX and the artificial intelligence start-up xAI. Mr. Musk started engaging with others about space data centers in November on X, the social media platform he owns, saying that "serious AI scaling" had to "be done in space." In another post, he mused about building 300 gigawatts of space data centers, which would require more than half of the power that the United States uses in a year. Bret Johnsen, SpaceX's chief financial officer, said in a letter to shareholders last month that the company would explore an initial public offering next year partly to raise money for projects including "A.I. data centers in space." SpaceX and Mr. Musk did not respond to requests for comment. Tom Mueller, a former SpaceX executive who believes humans will hit the limits of terrestrial energy sources by 2040, said part of the reason Mr. Musk and other A.I. leaders were talking about space data centers was the financial opportunity. "The hottest thing to invest in right now is A.I., and the second-hottest thing is space," Mr. Mueller said. "Now they're converging."
[2]
Even the sky may not be the limit for AI data centres
AI leaders believe Earth may soon lack the power and space for data centres, pushing them to look skyward. Big tech figures are backing the idea of building computing hubs in orbit, despite huge costs and doubts. Supporters say it is inevitable, while critics call it unrealistic. If the architects of the artificial intelligence boom are right, it is only a matter of time before data centres -- the giant computing facilities that power AI -- will float in orbit and be visible in the night sky like planets. The science-fiction-like dream is being driven by AI and space industry leaders who are growing increasingly worried that data centres will eventually require more energy and land than are available on Earth. So one solution -- perhaps the only solution, they say -- is to start building them in space. Google announced in November that it was working on Project Suncatcher, a space data centre project that would begin test launches in 2027. Elon Musk said at a recent conference that space data centres would be the cheapest way to train AI "not more than five years from now." Others pledging support for the idea include Jeff Bezos, the founder of Amazon and Blue Origin; Sam Altman, the chief executive of OpenAI; and Jensen Huang, the CEO of Nvidia. "It is not a debate -- it is going to happen," said Philip Johnston, the CEO of Starcloud, a space data centre startup. "The question is when." The notion has gained traction as the AI race hits a fever pitch, fuelling fears of a potential bubble. Meta, OpenAI, Microsoft, Amazon and other big tech companies are investing hundreds of billions in data centres worldwide, with OpenAI alone committing $1.4 trillion to such projects. Saudi Arabia and other nations are also pouring money into these efforts, while smaller companies pile up debt and take on financial risks to join the frenzy. Yet earthbound data centres are increasingly running into limits. In many places, the projects lack enough available power for the computing needs. Local opposition has also flared over whether data centres are driving up utility bills and exacerbating water shortages. That has led to more creative -- some might say wishful -- thinking with space data centres. Technologists and scientists have researched the idea and concluded that some version of these projects may be possible in the next few decades. But sceptics said the proposals flew in the face of physics and would be astronomically expensive.
Share
Share
Copy Link
Google announced Project Suncatcher for 2027 test launches while Elon Musk predicts space data centers will be the cheapest AI training option within five years. With OpenAI committing $1.4 trillion to data center projects and terrestrial facilities hitting energy limits, AI and space industry leaders see orbital computing hubs as inevitable—though critics call the physics and economics unrealistic.
The artificial intelligence boom is pushing computing infrastructure to unprecedented scales, and AI and space industry leaders now argue that Earth may not have enough energy and land resources to sustain the growth. Google announced in November that it's developing Project Suncatcher, a space data center initiative with test launches planned for 2027
1
. Elon Musk declared at a recent conference that space data centers would become the cheapest method for training AI "not more than five years from now," while other prominent figures including Jeff Bezos, Sam Altman, and Jensen Huang of Nvidia have pledged their support for building data centers in orbit1
2
.
Source: NYT
Philip Johnston, CEO of space data center startup Starcloud, stated bluntly: "It is not a debate -- it is going to happen. The question is when"
1
. This confidence reflects growing concerns about AI's insatiable demand for resources as the technology race accelerates.Meta, OpenAI, Microsoft, Amazon, and other major tech companies are investing hundreds of billions in AI data centers worldwide, with OpenAI alone committing $1.4 trillion to such projects
1
. Saudi Arabia and other nations are similarly pouring money into these efforts, while smaller companies accumulate debt and accept financial risks to participate in the expansion.
Source: ET
Yet terrestrial data centers increasingly face critical bottlenecks. Many projects lack sufficient available power for their computing needs, while local opposition has intensified over concerns that these facilities drive up utility bills and worsen water shortages
1
2
. These constraints are forcing companies to consider what some call wishful thinking: computing hubs in orbit that could bypass Earth's limitations entirely.Space data centers would differ dramatically from football-stadium-size terrestrial facilities. Mock-ups from companies like Starcloud depict large satellites with server clusters housing semiconductors at the center, surrounded by miles of solar panels. The primary advantage is abundant solar energy, with nearly 24/7 sun access and no clouds to obstruct power generation, Johnston explained
1
. Fewer environmental regulations and no neighbors to oppose construction present additional benefits.These orbital facilities would appear in the sky at dawn and dusk, visible from Earth at about a quarter the width of the moon, and would require complete rebuilding every five years when AI chips typically need replacement
1
. However, the high cost of space launches remains the primary obstacle. Currently, launching materials costs around $8,000 per kilogram, with SpaceX offering the cheapest rate at approximately $2,000 per kilogram—and individual server racks can weigh over 1,000 kilograms1
.Phil Metzger, a physics professor at the University of Central Florida and former NASA physicist, suggests the economics will make sense when launch costs drop to around $200 per kilogram, which he predicts could take a decade
1
. Google's November research paper on Project Suncatcher similarly predicted costs could decline to that level "in the mid-2030s," comparing the timeline to its driverless robot taxis, which required 15 years to develop1
.Related Stories
Feasibility depends not only on cheaper launches but also on solving technical challenges like radiation exposure and cooling in the vacuum of space. While technologists and scientists have researched the concept and concluded some version may be possible in coming decades, skeptics argue the proposals defy physics and would be astronomically expensive
1
2
.Pierre Lionnet, a space economist and director at trade association Eurospace, said recent comments from tech luminaries like Musk about space data centers are "a magnitude larger than what current research suggests is possible." He called the proposals "completely nonsensical"
1
. Others question whether launch costs will decline as rapidly as proponents predict.Despite skepticism, the concept has historical precedent. NASA introduced the idea in the 1960s, and "data repositories" in space appeared in 1980s science fiction
1
. The modern vision of space data centers capable of AI scaling emerged only in the last decade, driven by the explosive growth of artificial intelligence and its unprecedented computational requirements.Metzger acknowledged the uncertainty but maintained cautious optimism: "As a business case, it's plausible. It's been an evolving discussion"
1
. Whether Blue Origin, SpaceX, or other companies can reduce launch costs sufficiently remains the critical variable that will determine if satellites hosting AI infrastructure become a common sight in the night sky—or remain confined to speculative presentations and research papers.Summarized by
Navi
11 Dec 2025•Technology

04 Nov 2025•Technology

02 Dec 2025•Technology

1
Business and Economy

2
Policy and Regulation

3
Technology
