The SpaceX IPO and Data Centers in Space

· Source: Stratechery by Ben Thompson · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Emerging Technologies & Innovation, Robotics & Autonomous Systems · Depth: Advanced, long

Summary

SpaceX is pursuing a \$2 trillion valuation in its IPO, despite reporting \$18.67 billion in revenue and \$4.9 billion in losses last year, with growth slowing to 33%. This valuation is underpinned by an estimated \$28.5 trillion total addressable market, heavily skewed by a speculative \$26.5 trillion in AI. The company's strategy mirrors Tesla's "brand halo" effect, where Musk's vision and scale transform markets, as seen with Starlink's adoption by American Airlines for inflight Wi-Fi on over 500 narrowbody aircraft by Q1 2027, promising up to 1 Gbps per antenna. The article explores the feasibility and use case for data centers in space, proposing individual satellites as "rack-satellites" capable of housing AI compute, addressing terrestrial power and zoning constraints. These space-based data centers would primarily serve agentic inference, a potentially massive market.

Key takeaway

For investors or AI/ML directors evaluating long-term compute strategies, recognize that SpaceX's ambitious \$2 trillion IPO, while financially unconventional, signals a potential paradigm shift towards space-based data centers. You should consider the long-term viability of agentic inference workloads in space as a solution to escalating terrestrial power and zoning challenges. This vision, if realized, could establish a monopoly provider of marginal compute capacity, warranting a re-evaluation of traditional investment metrics for such ventures.

Key insights

Musk's companies succeed by changing market rules through scale and visionary "all-in" bets, often defying conventional financial logic.

Principles

In practice

Topics

Best for: Investor, Director of AI/ML, Consultant

Related on AIssential

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Stratechery by Ben Thompson.