Satellites typically operate for five to seven years, while AI and GPU technology cycles are considerably shorter
| Photo Credit:
ANI
India’s space-tech ecosystem is beginning to explore a shift from satellites that primarily collect and transmit data to intelligent computing nodes capable of processing information in orbit, as startups look to reduce bandwidth requirements and improve the speed of insights from Earth observation and other applications.
The emerging model involves using onboard AI to filter, classify and prioritise data before transmitting it to ground stations. This could be particularly relevant for defence, Earth observation, maritime surveillance, disaster response and space-domain awareness, where the value of information often depends on how quickly it can be acted upon.
“We see LEO satellites evolving from data collection platforms into intelligent nodes that can make first-order decisions in orbit,” said Krishna Teja Penamakuru, Chief Operations Officer and Co-founder, Dhruva Space. He said the earliest commercially viable applications are likely to be Earth observation, defence and government missions, where latency, bandwidth and data sovereignty are important.
Reduced Latency
The shift could significantly reduce the volume of data that needs to be transmitted to Earth. Sandeep Shah, Co-founder and Managing Director, Optimized Electrotech, estimates that onboard AI could achieve 50–90 per cent data reduction for suitable Earth-observation workloads, while bringing latency down from hours to minutes.
“LEO satellites are evolving from data collectors into intelligent sensing and decision nodes,” Shah said. “The transformation is from satellite imagery to actionable intelligence,” he added.
The opportunity is being driven partly by the growing volume of data generated by satellite constellations and constraints around ground-station access and downlink capacity. Processing data in orbit allows satellites to transmit insights rather than entire datasets.
Evolving Economy
However, the industry does not expect dedicated orbital data centres to become the dominant model immediately. Instead, specialised computing payloads are likely to be integrated into existing and new satellite constellations.
“Based on how the industry is progressing right now, the more likely path, at least within the three-to-five-year window, would be specialised compute payloads riding on existing and near-term Earth observation and IoT constellations,” said Ramesh Kumar V, Co-founder and CEO, Grahaa Space.
Power, thermal management, radiation tolerance and hardware refresh cycles remain challenges. Satellites typically operate for five to seven years, while AI and GPU technology cycles are considerably shorter, creating questions around the economics of replacing compute hardware in orbit.
Grahaa expects onboard processing and inference to become a default feature on next-generation Earth-observation satellites over the next three to five years. Dedicated orbital data-centre constellations, however, are likely to emerge later as the economics of space-based computing become clearer.
Published on August 18, 2026
