Google Launches Satellite to Test AI Chips in Space as Part of Long-Term Project

Oct 1, 2026

AI Summary

Google has successfully launched a prototype satellite to test its Tensor Processing Unit in space, marking a significant step in its Project Suncatcher initiative aimed at developing orbital data centers. The company anticipates that achieving cost-effective launches will require SpaceX's Starship to conduct around 1,800 flights over the next decade.

Google launched a prototype satellite aboard a SpaceX rocket to test its Tensor Processing Unit (TPU) in space.
The satellite will operate in 15-minute bursts to manage power and thermal systems while running various models.
Project Suncatcher aims to create large-scale compute clusters in orbit, with plans for a network of 81 satellites processing data in parallel.
Google released a peer-reviewed white paper analyzing orbital data centers, discussing the future of space access and cost reduction.
The paper suggests that SpaceX could reduce launch costs to around $200 per kilogram by 2035, requiring Starship to complete approximately 1,800 launches to achieve this.
Google believes its chips can withstand space radiation, with a low error rate for typical inference operations, although challenges remain for large-scale training tasks.

googlespacexspace data centersadvanced chiporbit

Google Launches Satellite to Test AI Chips in Space as Part of Long-Term Project

Related Stories

OpenAI Revises Revenue Projections, Delta Air Lines Misses Earnings Estimates, and Starbucks Considers Chipotle Acquisition

Revolut CEO outlines plans to expand digital banking services in the US

Bending Spoons CEO Highlights Acquisition Opportunities Amid Software Valuation Declines

AI Competition Shifts Focus to Financing Strategies Amid Technology Developments