🧠 Food for thought

Implications, context, and why it matters.

Orbital AI faces a steep cost gap

• Google is chasing space solar power as AI drives up electricity use. Its data center consumption rose from 2020 to 30.8 million megawatt-hours in 2024, more than double the earlier level 1.
• U.S. data centers could use as much as 12% of the nation’s electricity by 2028. That makes Sundar Pichai’s space power plan a long-range bet 1.
• The price still looks tough. One estimate puts a 1 gigawatt (GW) orbital data center at almost three times the cost of a similar site on Earth 2.
• The math only works if launch prices fall from about $3,600 per kilogram to less than $200. Project Suncatcher’s white paper puts that 18-fold drop in the 2030s 2.
• Server racks are heavy, and permanent space structures do not yet exist, said Matt Garman, CEO, Amazon Web Services (AWS) 1.

Orbital compute is already shaping ground infrastructure

• Google is not alone. SpaceX and Nvidia-backed startup Starcloud, a company developing space-based data center concepts, are also working on orbital compute ideas 1.
• That push is already affecting Earth-based capacity. Anthropic signed an agreement for all compute at SpaceX’s Colossus 1 in Memphis, which tops 300 megawatts across more than 220,000 Nvidia graphics processing units (GPUs) 3.
• Anthropic expects the deal to ease service errors and rate limits for Claude users, including Claude Pro and Claude Max subscribers 3.
• This approach mixes ground systems with future orbital capacity, giving providers room to add power on Earth before shifting workloads to space when limits tighten 2.