Why does Elon Musk want to put data centres in space? | The Economist

Market Context & Motivation

• Rising costs of terrestrial energy and construction are driving industry interest in orbital data centers.
• Elon Musk merged XAI into SpaceX to consolidate capital flow, offsetting XAI's high training costs with SpaceX's Starlink revenue.
• Analysts suggest the merger and space-centric narrative may be strategic moves to increase valuations ahead of anticipated IPOs for SpaceX, OpenAI, and Anthropic.

Elon Musk's Strategic Plan

• Musk filed an FCC application to deploy a constellation of one million satellites, an order of magnitude larger than Starlink's 8,000–40,000 satellite footprint.
• Proposed architecture utilizes:
  • Sun-synchronous orbits for continuous solar power to run AI workloads 24/7.
  • Low-Earth orbits for lower-latency access to daytime users.
• Long-term roadmap includes:
  • Launching orbital data centers with annual capacity of 100–200 gigawatts (not cumulative), scaling to one terawatt-per-year from Earth.
  • Establishing manufacturing factories on the Moon to produce satellites.
  • Deploying "mass drivers" (railguns) on the Moon to launch satellites into Earth orbit.

Critical Pushback & Skepticism

• Sam Altman (OpenAI) characterized current orbital data center concepts as "ridiculous," citing prohibitive launch costs relative to terrestrial power prices.
• Altman highlighted the technical impossibility of repairing broken GPUs in space, noting that current failure rates make maintenance unfeasible.
• Industry consensus places large-scale orbital compute viability beyond the current decade.

Competitive Landscape & Emerging Pilots

• StarCloud: Already launched a GPU-equipped satellite in November to test cooling requirements, radiation shielding, and hardware durability in space; plans involve massive 2x2km orbital structures.
• Google: Partnering with Earth observation firm Planet to deploy orbital hardware next year.
• Jeff Bezos: Expressing interest in space-based AI factories as a precursor to broader space manufacturing, though estimating a one-decade timeline for feasibility.
• Eric Schmidt: Acquired Relativity Space last year specifically to secure launch capacity for anticipated orbital data center projects.
• Sam Altman (Pre-IPO Analysis): Previously declined to acquire space startup Stoke Space, deeming it a distraction until technical hurdles are resolved.