On 1 October 2026, two projects will launch on SpaceX's Transporter-18 mission to test the operation of artificial intelligence directly in orbit.
The Indian startup TakeMe2Space is sending up the MOI-1A satellite, weighing less than 50 kilograms. The craft is equipped with Nvidia Orin NX processors and is designed to run AI models in space. It will process Earth observation data even before transmitting it to ground stations.
This is India's first orbital satellite for AI-based computing. Launching on a SpaceX rideshare mission opens up access to orbit for small companies without their own rockets.
At the same time, Google is launching a satellite prototype as part of the Suncatcher project. On board are four TPU tensor processing units of the company's own design. The satellite, built by the firm Planet, will remain in orbit for about a year.
The main task is to test how the chips withstand cosmic radiation and temperature swings. Before launch, Google already conducted ground tests in a thermal vacuum chamber simulating the conditions of space.
In the future, Google plans to create an entire constellation of such satellites for data centers in space. In 2027, the company intends to launch two more craft to test laser communication between them.
Both projects show how commercial AI chips are gradually moving beyond Earth's laboratories. What will change in data processing when computing moves closer to the source?
The Transporter-18 mission is becoming a platform for such experiments. It allows new technologies to be tested without the enormous costs of dedicated launches.
