OpenAI CEO Sam Altman has publicly questioned the feasibility of orbital data centers, suggesting underwater infrastructure could prove easier to build and cool than computing systems deployed in space. The disagreement highlights a fundamental split in how tech leaders envision solving AI's massive infrastructure challenge as demand for computing power continues to skyrocket.

Why Are Tech Leaders Exploring Unconventional Data Centers?

The AI boom is straining traditional infrastructure in ways the industry never anticipated. Power grids, chip supply chains, and cooling systems are all buckling under the weight of training and running increasingly sophisticated AI models. Companies like Amazon, Microsoft, and Google continue expanding land-based facilities, but the appetite for computing power keeps growing faster than conventional solutions can accommodate. This desperation has pushed visionary leaders to explore alternatives that would have seemed purely speculative just a few years ago.

Elon Musk's SpaceX and Google have both publicly committed to orbital computing projects. Google CEO Sundar Pichai previously stated that the company would "send tiny racks of machines and have them in satellites, test them out, and then start scaling from there," adding that within a decade, space-based data centers would become "a more normal way to build data centers." Musk has claimed that space will be the cheapest place for AI data centers within three years. Google is reportedly in talks with SpaceX for a rocket-launch deal to deploy orbital data centers, according to recent reports.

What's Altman's Alternative Vision?

During an appearance on the technology podcast TBPN, Altman offered a more grounded perspective. While he wished Musk "luck," he argued that underwater infrastructure presents fewer engineering obstacles than orbital systems. The OpenAI CEO acknowledged that ocean robotics researchers have not yet focused on deploying computing systems underwater, but he suggested the concept could ultimately prove more feasible than constructing space data centers.

"It feels easier than constructing space data centers," Altman stated, noting that while space offers freedom from land and regulatory constraints, energy generation and infrastructure remain formidable challenges.

Sam Altman, CEO at OpenAI

Altman's reasoning centers on practical engineering realities. Space-based systems face extreme obstacles: generating power in orbit, managing thermal dissipation in a vacuum, launching and maintaining equipment at orbital altitudes, and dealing with radiation and micrometeorite damage. Underwater facilities, by contrast, benefit from natural cooling properties of ocean water and avoid many regulatory hurdles associated with land-based construction.

How to Evaluate Competing Infrastructure Strategies

• Energy Efficiency: Underwater systems can leverage ocean water for cooling, reducing energy consumption compared to traditional air-cooled data centers, while orbital systems must generate power and manage heat in the vacuum of space.
• Regulatory Complexity: Space-based infrastructure requires coordination with space agencies and international treaties, whereas underwater facilities face different but potentially less restrictive environmental and maritime regulations.
• Scalability and Maintenance: Land-based data centers can be expanded and serviced relatively easily, while orbital systems require expensive rocket launches for deployment and repairs, and underwater systems require specialized robotics and submarine access.
• Proven Technology: Traditional data centers use mature cooling and power infrastructure, whereas both space and underwater computing remain largely experimental with unproven economics at scale.

The disagreement between Altman and Musk reflects broader industry uncertainty about where future AI infrastructure will actually be built. Analysts note that investment ties between AI and aerospace are deepening. Deloitte projects U.S. defense spending on AI will grow 3.5 times by 2029, with contracts awarded to leading AI firms. Altman himself acknowledged this convergence, observing that "everyone has diversified into every company at this point, basically".

The tension between visionary projects and practical engineering hurdles remains unresolved. While Musk and Pichai champion space-based solutions, Altman's skepticism suggests that the path forward may involve multiple approaches. Some companies may pursue orbital systems for specific use cases, while others invest in underwater or other unconventional infrastructure. The winner will likely be determined not by ideology but by which approach first achieves reliable, cost-effective operation at scale. For now, the race continues, with each leader betting on a different vision of how humanity will power the next generation of artificial intelligence.