Interview, Podcast
Why does Elon Musk want to put data centres in space? | The Economist
- Industry projections anticipate orbital data centers as the optimal construction location in the very near future, with specific architectural plans including a one-million-satellite halo in sun-synchronous orbit and a middle band for daytime AI capacity, all powered by continuous sunlight.
- SpaceX targets a scale increase from current thousands or tens of thousands to one million satellites, with launch capabilities planned for 100 to 200 gigawatts per year, potentially expanding to one terawatt per year through lunar factories and mass drivers if a moon-based manufacturing economy is established.
- Strategic financial expectations include Initial Public Offerings for SpaceX, OpenAI, and Anthropic over the next few years, with a projected combined AI IPO boom and a valuation of $1.25 trillion for SpaceX following these maneuvers.
- Skeptical viewpoints from industry leaders like Sam Altman and Jeff Bezos characterize current space data centers as economically unviable due to launch and power costs, with Altman predicting they will not matter at scale this decade and Bezos estimating a decade-long timeline before they make sense compared to lunar factory utility.
- Technical and operational challenges identified include the unresolved difficulty of repairing broken GPUs in space and the logistical need to build massive structures, such as a two-by-two kilometer StarCloud facility containing GPU shipping containers, before orbital computing becomes practical.
- Corporate actions and partnerships reflect anticipation of future launch capacity, including Google's plans to fly projects with Planet next year, Eric Schmidt's acquisition of Relativity Space to secure launch capabilities, and Sam Altman's potential future interest in acquiring Stoke Space.