# AI Implications

By [DYLIT Chronicles](https://dylit.info/user/dylitmediabuzz)

[Everything AI - Beyond the Hype](https://dylit.info/pr/everything-ai-beyond-the-hype/6a9efac02e92664f4d50cf9d) > [AI Implications](https://dylit.info/ch/ai-implications/6a9efac02e92664f4d50cfc0)

Google Is Sending Its AI Chips to Space: Now Orbit Host AI Computing? What Google is testing Google is about to find out whether its AI chips can work in orbit. A prototype called MVP is scheduled to launch on 1 October 2026 on SpaceX's Transporter-18 rideshare mission, on a Falcon 9 from Vandenberg Space Force Base in California. Google built the satellite with Planet, the Earth-imaging company. It carries four Google TPUs, roughly the computing power of one data centre server, powered by about a kilowatt of solar panels. The flight is the first orbital step for Project Suncatcher, which Google introduced on November 4th, 2025 as a research effort to explore scalable machine-learning compute in space. What the test is meant to show The technical question is simple: will the chips survive? The mission will assess how the equipment handles launch stress, radiation and extreme temperatures in low Earth orbit. Operationally, Google will try out a cooling setup. It plans to use heat pipes and radiators, since conventional air cooling can't work in a vacuum. The strategic goal is learning. Google stressed that this first mission is meant to collect data and identify failure points, and that it isn't a ready-made orbital data center. In 2027, it plans to launch two satellites to test the high-speed laser links future clusters would need. How orbital AI computing could work The pitch starts with power. Google says satellites in low Earth orbit get near-constant sunlight and can generate up to eight times more solar power than on Earth. Satellites would fly close together and share work over lasers. The hurdles are serious. Heat can only escape through radiators. Radiation is another problem, and in Google's ground tests the high-bandwidth memory was the most sensitive part, with error rates the company called likely acceptable for inference, while effects on training still need study. Cost matters too. Google's researchers estimate launch costs could fall to around $200 per kilogram by the mid-2030s, but that's a projection. Repairs are nearly impossible once hardware is up there, and data still has to move to and from the ground. What it could mean for AI If space-based compute ever works at scale (and that's unproven), AI growth could become less tied to power grids and land on Earth. Pichai has suggested that ten years from now, data centers in orbit could be seen as normal. That's a hope. Nobody has shown yet that orbital compute can match ground data centers on price or reliability, and this flight won't prove it either. Opportunities and risks for other companies Google has company. Nvidia-backed startup Starcloud launched an H100 chip into orbit last November and has used it to run Google's Gemma model. Launch providers and satellite builders could see new demand if the idea catches on. Chipmakers might face pressure to design accelerators that tolerate radiation and shed heat without air. Rival cloud providers will have to decide whether to invest early or wait for results. The risks are just as real: high upfront costs, reliance on a small number of launch companies, orbital debris, and open regulatory questions. All of this hinges on tests that haven't happened yet. Sources https://blog.google/innovation-and-ai/models-and-research/google-research/google-project-suncatcher-facts/ https://thenextweb.com/news/google-project-suncatcher-first-tpu-launch Related YouTube videos Google's latest moonshot to put machine learning in space: https://www.youtube.com/watch?v=o1JK79jszqo Project Suncatcher - Google's Plan to Put AI Data Centers in Space: https://www.youtube.com/watch?v=XlSQZKY_gCg