TakeMe2Space will launch its MOI-1A satellite on 1 October aboard SpaceX rideshare mission Transporter-18. The Hyderabad startup describes the spacecraft as India's first orbital computing satellite because it processes imagery in orbit and returns answers instead of raw files. Twenty-three customers have already booked capacity, from mapping and farming to mining, supply chains and insurance. For business, the flight tests whether AI inference can move from ground servers to orbit.
Rideshare slot after January loss
MOI-1A replaces MOI-1, lost on 12 January when the third stage of ISRO's PSLV-C62 rocket failed. After that failure the company took the first available slot on a SpaceX Falcon 9, ThePrint reported in May. The new payload will fly as part of Transporter-18, a rideshare flight that carries multiple small satellites at shared cost. An earlier test spacecraft, MOI-TD, launched from Sriharikota on 30 December 2024 and, according to the company, completed more than 20 in-orbit tests. Those trials included AI tasks run on applications uploaded from the ground, which prepared the commercial mission now scheduled.
MOI-1A weighs less than 50kg and runs on Nvidia Orin NX chips built for edge computing. Founder and chief executive Ronak Samantray told Reuters that customers load their own AI models onto the satellite. When the spacecraft passes over an area selected by a customer, those models analyse the observations on board. Only the processed result travels down to Earth, rather than large raw datasets. With about 150 watts of power available, the design functions as a small edge computer in orbit rather than a full data centre.
TakeMe2Space designs and builds most parts of its satellites in India, while buying propulsion systems, chips and solar cells from outside suppliers. The approach differs from projects pursued by Starcloud and SpaceX, which plan to fly more powerful Nvidia chips of the type used in ground data centres. MOI-1A therefore represents a lighter path focused on filtering data at the source. The company plans larger networked satellites that would serve many customers at once. It also aims to sell orbital data storage to banks, finance firms and defence clients as a backup to ground cloud providers.
What this means for data users
For companies that buy Earth observation, the model changes where analysis happens and how fast answers arrive. A farming business could request crop stress maps for specific fields, a mining operator could monitor pits, and an insurer could assess damage without waiting for full image downloads. Mapping firms and schools are among the 23 booked users, alongside supply-chain operators, and US space data firm Little Place Labs is named as a customer. Small teams gain access without building ground processing, while larger firms can test orbital inference alongside existing pipelines before committing wider budgets.
The constraints are physical and operational. Power near 150 watts and mass below 50kg limit model size and throughput, so heavy training stays on the ground. Rideshare timing, launch risk and satellite replacement remain factors, as shown by the January loss of MOI-1 on PSLV-C62. Buyers need to check revisit frequency over their areas, supported model formats for Orin NX, downlink latency, and how results are validated against ground truth. The mission alone does not prove persistent coverage, since continuous service would require the planned networked constellation.
The marker to watch is the 1 October flight and the first results from the 23 booked customers. Successful on-orbit inference on customer models would confirm demand for edge processing in space. Progress on larger networked satellites and orbital storage contracts with finance and defence clients would show whether the concept scales beyond a single payload.
