Elon Musk — "In 36 months, the cheapest place to put AI will be space” - Dwarkesh Podcast Recap

Podcast: Dwarkesh Podcast

Published: 2026-02-05

Duration: 2 hr 50 min

Summary

Elon Musk predicts that within 36 months, the most cost-effective way to deploy AI infrastructure will be in space, driven by energy availability and efficiency. His argument hinges on the limitations of terrestrial resources compared to the advantages of operating in a space environment.

What Happened

In this episode, Elon Musk discusses the future of AI infrastructure and the unique challenges posed by energy demands. Musk points out that the total cost of ownership of a data center is largely influenced by energy costs, which only account for 10 to 15% of the total. He argues that as electrical output stagnates globally, particularly outside of China, the need for alternative energy sources becomes critical. Space, with its abundance of solar energy and minimal regulatory hurdles, emerges as a viable solution for future AI operations.

Musk elaborates on the benefits of solar energy in space, asserting that solar panels in orbit can be up to five times more effective than those on the ground due to the absence of atmospheric interference. He notes, "because you don't have a day-night cycle or seasonality, clouds, or an atmosphere in space, ... it's actually much cheaper to do in space." He predicts that the costs associated with running AI operations in space will drastically decrease, making it the most economically feasible option for AI deployment within a short timeframe.

The conversation also touches on the logistical challenges of servicing GPUs in space and the reliability of these chips after initial debugging. Musk expresses confidence about GPU reliability, stating, "...once they start working and you're past the initial debug cycle... their actual reliability... they're quite reliable past a certain point." He emphasizes that as the demand for AI continues to increase, the only feasible way to meet that demand at scale will be in space, where energy harnessing can reach unprecedented levels.

Key Insights

• The total cost of ownership for data centers is primarily driven by energy costs, which are becoming increasingly constrained on Earth.
• Space provides a unique environment where solar panels can operate more efficiently and without the limitations posed by Earth's atmosphere.
• Musk believes that the logistics of servicing GPUs in space will not be an issue due to their reliability after initial testing.
• In the next 36 months, Musk predicts that the most economic place to put AI infrastructure will be in space, driven by the need for scalable energy solutions.