SpaceX wants to put data centers in orbit, and Musk says it's no big deal

2026-06-09 · Source: The Decoder · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cloud Computing & IT Infrastructure, Emerging Technologies & Innovation · Depth: Intermediate, quick

Summary

SpaceX plans to deploy AI data centers in orbit, with Elon Musk asserting the engineering is "near-trivial" due to existing Starlink V3 satellite technology. The company aims for mass production by the end of 2027 from its Bastrop, Texas factory. The initial AI satellite is projected to deliver 150 kilowatts of peak power and 120 kilowatts of sustained compute, equivalent to a single Nvidia GB300 rack, utilizing space radiation for cooling and solar panels for power. However, the article notes that complex AI training, which requires tightly coupled GPUs with terabytes per second of bandwidth, remains challenging to replicate in orbit. Competitor Jeff Bezos estimates orbital data centers will not achieve cost parity with ground-based facilities for up to 20 years, casting a more cautious light on SpaceX's \$1.75 trillion valuation ahead of its IPO.

Key takeaway

For AI Architects evaluating future compute infrastructure, understand that while SpaceX proposes orbital data centers, current capabilities are limited. You should prioritize ground-based solutions for large-scale AI model training due to the need for tightly coupled GPUs and high bandwidth. Consider orbital options only for specific inference workloads with moderate latency requirements, acknowledging the 20-year cost parity projection from competitors.

Key insights

SpaceX aims to put AI data centers in orbit, but current technology limits them to inference, not complex training.

Principles

• Orbital AI compute faces significant coupling challenges.
• Cost parity for space data centers is decades away.
• Existing satellite tech supports basic orbital compute.

In practice

• Consider orbital compute for inference workloads.
• Evaluate latency/bandwidth needs for space deployment.
• Factor in cosmic radiation effects on training.

Topics

• Orbital Data Centers
• AI Inference
• AI Training
• SpaceX Starlink
• NVIDIA GB300
• Space Compute Costs

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

Related on AIssential

• Delian Asparouhov and Philip Johnston on making the case for orbital data centers — Orbital data centers are emerging as a viable, long-term solution for edge compute and inference, driven by space technology advancements.
• XAI's Radical Plan: Data Centers In Space — XAI plans orbital AI data centers and advanced automation, leveraging SpaceX's capabilities to overcome terrestrial compute limitations.
• Google is in talks with SpaceX on orbital data center partnership — Orbital data centers are emerging as a strategic evolution in data management and AI compute, driven by major tech companies.
• Space-Based Data Centres: The Future of AI Computing in 2025 — Orbital data centers offer sustainable, scalable AI computing by harnessing space's unique environment and unlimited solar energy.
• SpaceX: The AI Orbital Hyperscaler — A massive compute lease deal between Anthropic and SpaceX highlights critical AI infrastructure supply-demand imbalances.
• Philip Johnston of Starcloud at RAAIS 2026 — Orbital data centers can overcome terrestrial AI compute constraints by leveraging space-based power and cooling.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by The Decoder.