Microsoft's Abandoned Underwater Data Center Project Serves as Warning for SpaceX's Orbital AI Ambitions
When Microsoft lowered a shipping container-sized data center to the bottom of the Scottish seabed in 2015, it represented what many believed was a revolutionary glimpse into computing's future. A decade later, this ambitious project has been completely abandoned due to insufficient customer interest, with Microsoft confirming it no longer operates any underwater facilities. Now, as Elon Musk sets his sights even higher with plans for orbital data centers, industry experts fear SpaceX may be heading toward the same disappointing outcome, with potentially much higher costs of failure.
Musk's Bet on Space Data Centers and Microsoft's Underwater Experience
SpaceX filed for an initial public offering on Wednesday with Musk declaring that the proceeds will fuel launching up to one million data-center satellites into orbit. This ambitious plan aims to create a vast artificial intelligence computing network in space that would bypass the power and water constraints of Earth-based facilities. Between 2015 and 2022, Microsoft pursued Project Natick, an experimental effort to host data centers on the ocean floor that would be cooled naturally by seawater and powered by offshore wind and tidal energy.
However, this innovative project was ultimately shelved. Citing two sources with direct knowledge of the project, a Reuters report revealed that underwater data centers were abandoned primarily due to economic factors and a simple lack of client demand. This means that while the technology functioned successfully from a technical standpoint, customers showed no interest in scaling it for broader commercial use. Microsoft stated it would continue using Project Natick as a research platform despite discontinuing operational deployment.
Why Experts Believe Space May Present Even Greater Challenges Than Underwater Environments
Five data center specialists interviewed by Reuters indicated that Microsoft's experience provides a direct warning for SpaceX's orbital ambitions. Both projects rely on sealed, modular units that are expensive to deploy and cannot be easily expanded, repaired, or upgraded once in position. "These problems are likely to be more severe in space than under the sea," explained Roy Chua, founder of industry research firm AvidThink, highlighting the unique challenges of the space environment.
Beyond the fundamental difficulties of cooling data centers in the vacuum of space, experts point to the enormous cost of rocket launches required for such an ambitious project. To achieve Musk's stated targets, SpaceX may need to complete approximately 3,000 Starship launches annually, which translates to roughly eight launches every single day. This logistical challenge represents a significant barrier to economic viability and practical implementation.
Jensen Huang's Perspective: The Case for Ground-Based Solutions
Perhaps the most pointed skepticism has come from Jensen Huang, CEO of Nvidia, the company that manufactures chips powering virtually all of the world's leading artificial intelligence systems. Speaking on the All-In podcast in February, Huang argued that the economics of space-based AI data centers simply do not add up at present. "We should definitely work on the ground first because we're already here," Huang stated, describing orbital AI infrastructure as a long-term engineering challenge rather than anything approaching a near-term solution.
The technology executive emphasized that terrestrial solutions should take priority, given existing infrastructure and more manageable implementation challenges. Huang's perspective reflects broader industry concerns about the practicality and economic feasibility of moving critical computing infrastructure into space when ground-based alternatives continue to evolve and improve.
