Musk wants SpaceX IPO to fund AI space data centers. Microsoft’s underwater retreat sounds a warning.

Register Here.

Asked for comment, a Microsoft spokesperson said: “While we currently have no datacenters in the water, we will continue to use Project Natick as a research platform to test, evaluate and validate new ideas about datacenter reliability and stability.”

Five data center experts told Reuters that what went wrong for Microsoft is a matter of caution for SpaceX because although both projects are worlds apart geographically, they share the same things: both rely on modular components that are expensive to use and cannot be expanded, repaired or improved – features that are considered important by the AI ​​industry.

“These problems may be more severe in space than under the sea,” said Roy Chua, founder of industry research firm AvidThink, pointing to unsolved questions about how to cool data centers in orbit, the high cost of launching a rocket and the effects of a harsh environment on AI chips.

MUSK’S SPACE THINGS TO DEAL WITH PROBLEMS

Although Microsoft has proven that undersea data centers can work, customers have not been interested in scaling them up, instead growing conventional land-based infrastructure that has allowed for cheaper, faster upgrades as AI development accelerates, two people with knowledge of the project said, asking not to be named because of the sensitivity of the matter.

A proven, “locked-for-life” design – which SpaceX would carry out in orbit – is limited, as AI chips improve rapidly every year, while a satellite or undersea data center can be replaced every five to seven years.

The economy was also an obstacle, the two people said. Running data centers under the sea was more expensive than building on land, and while those costs could come down significantly, doing so would require tens of billions of dollars in investment.

The place will be very expensive.

Analysts at MoffettNathanson, an independent American equity research firm, said in a February report that Musk’s plan to put a million AI satellites in space would cost billions of dollars.

For on-site facilities to be commercially viable, startup costs will need to drop from today’s low thousands of dollars per kilogram to the low hundreds of dollars per kilogram, analysts say.

“The issue is not whether something will work, but whether it makes economic sense versus building more capacity on the ground,” said Tim Farrar, an independent satellite industry analyst at TMF Associates.

The need will not be an issue, Musk says, because the Earth’s energy resources will quickly run out as AI is needed to support a world where robots outnumber humans, all cars drive themselves and space travel becomes the norm.

“The idea that we can’t solve the world’s problems, like the lack of electricity and environmental issues, strikes me as something wrong with the Earth trying to make everything seem better in space,” Farrar said.

Musk’s case hinges on Starship, SpaceX’s next-generation rocket, designed to be fully reusable and carry larger payloads than SpaceX’s Falcon rockets. But Starship is behind schedule and has had multiple problems on some of its 11 suborbital test flights through 2023.

MoffettNathanson estimates that achieving Musk’s goal would require 3,000 Starship launches per year, or eight per day.

Jeff Bezos’ space company Blue Origin also supports orbital data centers. The rocket company said in March that its Project Sunrise concept would add AI computing capabilities to orbit, fueling clean solar power while maintaining Earth’s data center infrastructure.

Blue Origin did not respond to a request for further comment.

THE AI SPACE COULD BE A NEW BUSINESS

Space data centers have a future, but there is potential to complement ground-based data centers, said Claude Rousseau, director of research at Analysys Mason who tracks satellite markets.

“I firmly believe that in the future there will be no way that space-based data centers can replace data services,” said Rousseau, adding that it will be a very good industry serving orbit facilities, such as military constellations and space stations.

For example, the International Space Station already hosts test systems designed to process data in orbit and reduce dependence on downlink bandwidth.

“We need to work on the ground first because we’re already there,” Huang said, describing orbital AI architecture as a long-term engineering challenge rather than a near-term solution.

Chua said that plans to move data centers under the sea or in the danger of space to try to escape the problems of the Earth and create new and difficult challenges.

“There are many problems we can solve on Earth before space,” Chua said, pointing to the success of the AI ​​chip, efficient water recycling, and the expanded use of solar energy and modular nuclear power.

Our standards: The Thomson Reuters Trust Principles.opens a new tab