Google is preparing to launch its first experimental satellite dedicated to orbital AI computing aboard a SpaceX rocket. The mission, part of Project Suncatcher, marks the tech giant’s initial step toward testing whether space can host large-scale AI workloads.
What Happened
The satellite, known as MVP, is scheduled to fly next week on SpaceX’s Transporter-18 rideshare mission, with a slated launch date of October 1. The refrigerator-sized spacecraft carries four of Google’s Tensor Processing Units (TPUs), the same chips used in its ground-based data centers. Google accelerated this test by integrating its custom chips into a satellite rather than waiting for the originally planned 2027 launch of two dedicated custom satellites.
According to a Google blog post, the MVP satellite’s solar panels supply approximately one kilowatt of power. The TPUs will run Gemini models in bursts of roughly 15 minutes before shutting down to allow radiators to shed heat. Google confirmed that its Trillium TPUs survived vibration testing simulating sustained launch loads of up to 10g, with individual components enduring 50 to 100g, and handled radiation doses exceeding what a five-year mission would deliver.
Why It Matters
This launch underscores the deepening ties between Google and SpaceX. The companies’ relationship became public in May when reports indicated Google was in talks with SpaceX for a launch deal tied to orbital data centers. Google currently holds a stake of roughly 6% in SpaceX. While both companies are pursuing the concept of orbital compute, they operate on vastly different timelines and scales.
SpaceX’s own orbital compute program centers on the AI1 satellite, a 70-meter structure derived from Starlink V3 hardware designed for 150 kW of peak compute—about 150 times the power draw of Google’s MVP. SpaceX is constructing a Gigasat factory in Bastrop, Texas, aiming for an annualized production rate of 1 GW of space compute by the end of 2027. Elon Musk recently stated on X that “the amount of compute in space will obviously round up to 100% of all compute.”
In contrast, Google has adopted a more cautious public stance. Its research estimates that launch prices must fall below $200 per kilogram for orbital data centers to compete with ground facilities on energy cost, a threshold the company believes may not be reached until the mid-2030s. The Suncatcher team expects the effort to remain a research project rather than a commercial product for several years.
The Bottom Line
Google’s decision to fly its first orbital AI test on SpaceX’s Transporter-18 mission places its experimental hardware on the rocket of the company with the most aggressive timeline in the field. While SpaceX pushes for massive scale and Google focuses on validating technology and economics, this launch serves as a critical proof-of-concept for the emerging orbital data center market.