Latest News

SpaceX's orbital Data Centers could face the same challenges as Microsoft's abandoned underwater project

SpaceX's orbital Data Centers could face the same challenges as Microsoft's abandoned underwater project
SpaceX's orbital Data Centers could face the same challenges as Microsoft's abandoned underwater project

SpaceX filed an IPO on Wednesday that Elon Musk claims will fund an effort to transform the rocket manufacturer into an 'AI powerhouse. Launching up to 1,000,000 data-center satellites in orbit to bypass Earth's power and water limitations.

Microsoft had a similar ambition to escape land-based computing constraints in 2015, when it lowered ?a shipping-container-sized data center onto the seabed off Scotland, aiming to cut energy ?use through natural seawater cooling and tapping offshore wind and tidal power.

Microsoft's Project Natick, once hailed as a?potential breakthrough? for the data-center sector, has met all of its technical goals, but the underwater data centers have been abandoned over two years due to lack of demand from clients and unviable economics.

A Microsoft spokesperson responded to a question by saying: "While there are no datacenters currently in the water, Project Natick will be used as a platform for research to explore, test and validate new concepts about datacenter reliability.

Five data center experts told us that what happened to Microsoft should serve as a warning for SpaceX. Although the two projects are geographically far apart, they have many similarities. They both rely upon modular units, which are costly to deploy, and can't be repaired, upgraded, or expanded. These features are considered crucial by the AI industry.

Roy Chua of AvidThink said that "these problems will likely be more severe in space than in the ocean". He cited unresolved issues such as how to cool the data centers in orbit and high launch costs, along with the adverse effects of harsh space conditions on AI chips. SpaceX?did not respond to a comment request. SpaceX, the company that acquired Musk's AI start-up xAI in Feburary, could raise $75 billion if it goes public. This would make it possibly the largest IPO ever. xAI's holdings include the social media company X (formerly Twitter) and AI chatbot Grok.

MUSK'S SPACES? AMBITIONS FACE HURDS

The two people who were familiar with the project, but asked not to be identified due to the sensitive nature of the issue, said that although Microsoft had proven that undersea facilities could work, their customers did not want to scale them up. Instead, they wanted to expand conventional land-based installations that would allow cheaper and faster upgrades as AI technology accelerated.

The "locked-for life" sealed design, which?SpaceX will replicate in orbit, has limited flexibility because AI chips are improving rapidly every year. A satellite or an undersea information center, on the other hand, may only be replaced every five to seven (7) years.

Two people also said that the economics was a major obstacle. The cost of building data centers underwater was higher than on land. While the costs may have decreased at scale, it would still have taken tens or hundreds of millions of dollars to do so.

The cost of space will increase dramatically.

In a research note published in February, analysts at MoffettNathanson (an independent U.S. equity-research firm) said that Musk's plan of putting a million AI satellites into space would cost trillions.

Analysts say that in order to make data centers commercially viable, the launch costs will need to drop from their current low of thousands of dollars to a few hundred dollars per kilogram.

Tim Farrar is an independent satellite analyst with TMF Associates. "The issue is not whether or not something works, but if it makes economic sense compared to simply building more capacity at the ground level," he said. Musk claims he can overcome technical and financial obstacles, such as radiation exposure, heat control in a vacuum, and the need to replace hardware frequently, by lowering launch costs and developing more resilient AI chip.

Musk claims that demand will not be a problem, as Earth's resources will be quickly depleted by AI, which is required to support a future where robots will outnumber people, and all cars will drive themselves, and space travel will become routine.

Farrar stated that the idea that Earth's problems, such as power shortages and environment issues, can not be solved, is unrealistic. It makes Earth seem worse in order to make "everything" appear better from space.

Musk's argument hinges on Starship - SpaceX next-generation rocket. Starship is designed to be fully recyclable and can carry much larger payloads than SpaceX Falcon rockets. Starship has been years behind schedule and has experienced explosive failures on some of its 11 suborbital tests flights since 2023.

MoffettNathanson calculates that Musk's target would require 3,000 Starship launches per year, or 8 a day.

Blue Origin, Jeff Bezos’ space company, is also a supporter of orbital data centres. In March, the rocket company announced that its Project Sunrise concept will add AI computing capability in orbit by utilizing clean solar power and preserving data-center infrastructure on Earth.

Blue Origin has not responded to any further comments.

SPACE AI COULD be a niche business

Claude Rousseau is a Research Director at Analysys Mason, who monitors satellite markets. He believes that space data centers have a bright future.

Rousseau stated that "I am convinced that in the near future, space-based data centres will not be able to replace ground-based datacenters." He added that this would be more of a niche industry that serves infrastructure in orbit such as military satellite constellations or?space station.

The International Space Station, for example, already has experimental systems that?process data on orbit and reduce the reliance upon downlink bandwidth.

Nvidia's Chief Executive,?Jensen Huang, said on the All-In Podcast in February that the economics for space-based AI-data centers are still unattractive.

"We should work on the ground because we are already here," Huang said. He described orbital AI infrastructure more as a long-term engineering challenge than a short-term solution.

Chua said that plans to move data centres under the ocean or into space could create new challenges and make it harder to solve problems on Earth.

Chua stated that "there are many problems we can solve before going into space" pointing out improvements in AI chip performance, improved water recycling and the expanded use of solar energy and modular nuclear generation.

(source: Reuters)