New Delhi: Space could soon become the cheapest place to power artificial intelligence at least that’s what Tesla CEO Elon Musk believes. In a recent conversation on the Dwarkesh Podcast, Musk predicted that AI systems could be moved to orbit within the next 36 months. During the discussion, he spoke about the economics of space-based data centers, the growing difficulty of scaling energy infrastructure on Earth and the challenges of manufacturing humanoid robots at high volumes in the United States.

Explaining why he sees space as a better option, Elon Musk said it is “harder to scale on the ground than it is to scale in space,” as he outlined his plans to build AI data centers beyond Earth.

“You’re also going to get about five times the effectiveness of solar panels in space versus the ground, and you don’t need batteries,” he said. Musk further pointed out that space does not “have a day-night cycle, seasonality, clouds, or an atmosphere”. “The atmosphere alone results in about a 30% loss of energy,” he added.

Musk Says Space Could Soon Become the Cheapest Home for AI

Backing his argument for space-based infrastructure, Elon Musk said, “So any given solar panel can do about five times more power in space than on the ground. You also avoid the cost of having batteries to carry you through the night. It’s actually much cheaper to do in space,” the world’s richest person said.

Musk predicted that space will soon become the most affordable location for artificial intelligence systems. “It will be space in 36 months or less. Maybe 30 months.”

When asked about GPU reliability during large-scale AI training, Musk downplayed concerns around hardware failures, suggesting they are not as significant as many believe. He reiterated that the future of AI infrastructure may ultimately lie beyond Earth.

Musk Downplays Concerns Over GPU Reliability

Addressing questions around hardware failures during large-scale AI training, Elon Musk said reliability largely depends on how new the GPUs are when they are deployed. According to him, most issues tend to arise in the early stages, when the chips are still going through initial testing and debugging.

He explained that once GPUs move past this phase whether they are made by Nvidia, Tesla, or other manufacturers producing TPUs or Trainium chips they generally become stable and dependable over time. Musk suggested that after the early debugging cycle, hardware failures are relatively rare, and ongoing servicing is not a major concern in his view.