- Hyperspectral satellites carry 400 TOPS of onboard processing power and begin serving overseas customers
- Move underscores a shift in commercial space from data collection to data services
A pair of hyperspectral satellites developed with AI capabilities from Chinese space startup Star.Vision (地卫二) has successfully entered orbit, marking another step in China’s effort to move more computing power into space and reduce reliance on ground-based processing.
The two satellites, known as Star AI-05 (东方星链05) and Star AI-06 (东方星链06), were launched from Haiyang in eastern China’s eastern Shandong Province on August 5.
According to the Hangzhou-based company, the satellites carry 400 trillion operations per second (TOPS) of onboard computing power, the highest level currently deployed on an operational hyperspectral satellite.
The launch also marks the latest milestone for Star.Vision, which completed China’s first overseas commercial delivery of an entire privately developed satellite in 2024.
Computing in space
The satellites are equipped with Star.Vision’s proprietary “Xian” (弦) AI computing module, allowing them to process information directly in orbit rather than sending raw data back to Earth.
That shift addresses one of the industry’s longstanding constraints. Because of limited transmission bandwidth and storage capacity, traditional remote-sensing satellites often discard between 80% and 90% of the data they collect.
By moving AI processing into orbit, Star.Vision says it can shorten response times from days to minutes while increasing data utilization rates from less than 10% to more than 80%.
Only half a day after entering orbit, the satellites transmitted their first processed images from the Greek port of Rafina, providing an early demonstration of their ability to capture and analyze data simultaneously.
Satellites with customers already waiting
Unlike many experimental spacecraft, both satellites entered service with commercial customers already lined up.
One mission, called Samarkand-2028, will support Uzbekistan’s cotton industry by monitoring crop conditions throughout the production cycle.
Another project, Lampung-1, will provide agricultural monitoring and disaster warning services in Indonesia’s Lampung province.
Founded in 2021 and headquartered in Hangzhou, Star.Vision says it operates in more than 40 countries and regions.
The company is also the only Chinese commercial space startup involved in both the country’s upcoming Chang’e-8 lunar mission and the Tianwen-3 Mars mission.
A different kind of AI cluster
Star.Vision is part of a group of emerging technology companies informally known as Hangzhou’s “New Eight Stallions” (新八骏), a label introduced earlier this year to describe a new generation of local deep-tech firms.
Unlike the better-known “Hangzhou Six Little Dragons,” a group largely associated with consumer-facing AI applications, the “New Eight Stallions” are concentrated in deeper areas of the technology stack, including space computing, in-memory AI chips, inference GPUs and dexterous robotic hands.
Together, they form an ecosystem spanning everything from computing infrastructure to end-user applications, highlighting Hangzhou’s broader shift from business-model innovation toward fundamental technological development.
From selling satellites to selling services
The successful launch points to a larger transformation taking place in the commercial space industry.
Competition is increasingly moving beyond the number of satellites companies can place into orbit and toward the value they can extract from the data those satellites generate.
For Star.Vision, the goal is to replace the traditional “capture first, process later” model with one in which satellites become intelligent computing nodes capable of analyzing information while still in space.
The company’s growing international footprint suggests that Chinese commercial space companies are also becoming increasingly focused on exporting services rather than hardware alone.
As more computing power moves into orbit, satellites may come to be valued not simply as observation platforms, but as data-processing infrastructure in their own right.
