Chinese and Indian companies are taking AI processing into space. Whether this widens access to useful satellite services will depend on decisions made on Earth.
On 14 October, a Chinese rocket carried eight computing satellites into orbit for Star Compute, ADA Space’s planned orbital computing network. Their developer plans experiments ranging from advanced cooling to training artificial-intelligence models in space. Meanwhile, India’s TakeMe2Space advertises an orbital laboratory aboard its MOI-1A satellite, which customers can task for US$375 an orbit.
The announcements differ in scale but point towards a consequential possibility: universities, smaller businesses and public agencies might obtain useful space capabilities without building their own satellites.
But renting computing time is not the same as acquiring the ability to use it. Governments now face two connected questions: which orbital services deserve public support, and what would allow more countries to benefit from them?
Satellites already process data onboard. The emerging proposition is to let customers upload their own applications and purchase processing as a service. Analysing observations before transmitting them could reduce data traffic and deliver findings sooner. For an agency responding to a flood, the useful product might be a map of affected roads rather than a collection of images awaiting analysis.
That is a more realistic immediate proposition than moving a terrestrial data centre into orbit. Running applications on a satellite and training large AI models across a network of spacecraft involve different demands for power, communications and heat management. Calling both “orbital computing” can obscure their differences.
TakeMe2Space advertises customer-controlled imaging and processing, with results delivered within minutes. MOI-1A launched on SpaceX’s Transporter-18 rideshare mission in early October. That confirms the launch, not reliable operation of customers’ applications. The advertised service still needs independently documented performance results.
China’s approach also extends beyond launching experimental hardware. ADA Space announced a strategic agreement with Tencent Cloud in June and a constellation partnership with SenseTime in July. This latter agreement envisages a commercial network combining ground and space computing by 2028. These announcements connect spacecraft development with established cloud and AI businesses, although the published reports disclose neither guaranteed purchases nor the full financial commitments behind deployment.
There is also explicit public-policy interest. In April, Zhao Ce, a deputy director at China’s Ministry of Industry and Information Technology, advocated local trials, research into satellite computing components and applications, including real-time remote sensing and emergency communications. He also explicitly supported international cooperation.
That is a more useful starting point than assuming a technological contest. The statement signals support; it does not by itself fund Star Compute or award a service contract. The official emphasis on practical applications suggests a test that governments elsewhere should also apply: what problem does processing in orbit solve better than existing arrangements?
Governments can help establish whether these services are useful by purchasing trials against defined needs. A contract could require timely flood mapping or infrastructure monitoring, with results compared against existing satellite and ground-based services.
That would test more than whether a computer functions in orbit. It would establish whether the service delivers a better answer, sooner or at lower cost. Public support should reward that result. A successful launch is a milestone, not a measure of public value.
The second test is who can use the service. An affordable booking price could open satellite experimentation to universities, smaller businesses and public agencies. But users still need software, expertise and the ability to check the answers returned. A cheap orbit is of limited value to a potential user unable to prepare a task or interpret its results.
This is where international cooperation could make a lasting difference. Partnerships could combine service access with training, scientific collaboration and opportunities to develop applications. Customers would also need clear terms governing their data and software, including their rights if the service changes or ends.
A regional trial could bring together a satellite operator, university researchers and a disaster-management agency, with the agency defining the problem. Could processing observations in orbit identify flooded transport routes quickly enough to inform an evacuation? The comparison should include the time needed to acquire imagery, check the result and deliver it to emergency staff. Faster computation alone would not establish a better service.
Who sets the task is also important. If international cooperation consists only of selling access to a platform, participating countries may gain information without developing much expertise. Giving local researchers and public agencies a role in designing applications and evaluating results would make the partnership more valuable. It would also help distinguish capabilities worth retaining locally from services that can sensibly be purchased abroad.
Australia could build on existing regional collaboration. In February 2026, Indonesian Geospatial Information Agency staff trained at the University of Sydney with support from Geoscience Australia, learning to process satellite imagery using Indonesia’s Piksel platform. That partnership relates to ground-based analysis but offers a model for jointly evaluating orbital processing. Australia could help regional partners design trials, develop applications and publish independent results. The diplomatic value would lie in building expertise that remains useful beyond any single commercial platform.
China’s launch and India’s advertised laboratory make these questions timely. They do not yet establish a mature market, nor do they represent two complete national models. They offer an opportunity to shape an emerging service around the people expected to use it.
The policy ambition should be larger than putting more computers in space: enabling more people on Earth to ask useful questions of them – and understand the answers.
Richard de Grijs is Professor of Physics and Astronomy at Macquarie University and former Executive Director of the International Space Science Institute–Beijing. He is a frequent media commentator on developments in international space programs, with a focus on the Asia–Pacific region.
This article is published under a Creative Commons Licence and may be republished with attribution.
Photo: “The Soyuz MS-19 crew ship above southeast Asia” by NASA Johnson, CC BY-NC-ND 2.0
