India aims for the stars with orbital datacentres

India is embarking on an ambitious journey to establish orbital datacentres, a move that could redefine the landscape of artificial intelligence (AI) infrastructure. With the increasing demand for AI workloads and the pressing need for low-latency solutions, traditional terrestrial datacentres in India are facing significant limitations. This initiative is not only a response to these challenges but also aligns with the country's data sovereignty requirements across critical sectors such as defence, manufacturing, maritime, and healthcare.

The Rise of Orbital Datacentres

The concept of orbital datacentres is gaining traction as a revolutionary alternative to ground-based facilities. A notable partnership was formed between NeevCloud, a sovereign AI cloud infrastructure company, and Agnikul Cosmos, a spacetech startup, which was announced in February 2026. Together, they aim to launch India’s first indigenous AI-powered datacentre in space, with a pilot project slated for completion by the end of 2026. Other startups, like TakeMe2Space, are also exploring the potential of building orbital datacentres to facilitate AI model operations directly on satellites.

The Indian Space Research Organisation (ISRO) and the Department of Space (DoS) are reportedly assessing the feasibility of deploying physical datacentres in orbit to handle satellite and communications data. This raises the question of whether orbital datacentres represent a new frontier in edge computing or signify a more monumental shift in data processing.

Narendra Sen, founder and CEO of NeevCloud, emphasizes three key advantages of orbital datacentres: energy, latency, and sovereignty. He points out that space offers unlimited access to solar power, allowing for efficient heat radiation without the water and energy constraints faced by terrestrial facilities. This stark contrast highlights the limitations of land-based datacentres, which are increasingly transitioning to power-neutral and carbon-neutral models due to energy becoming a critical bottleneck.

Under the new partnership, Agnikul will supply the launch vehicle and orbital hosting platform using its extendable upper-stage rocket architecture, which eliminates the need for dedicated satellites for orbital payloads. NeevCloud’s compute modules will harness solar power in orbit and integrate AI chips directly into the architecture. Initial deployments are projected to require between 10kW and 15kW of power, with future iterations potentially scaling up to 50kW, 200kW, or more.

Overcoming Terrestrial Limitations

Moin SPM, Agnikul’s co-founder and COO, highlights the inherent constraints of terrestrial datacentres, which must be situated near population centers and power grids. These facilities struggle to provide services to remote areas and applications that require global low-latency coverage. While underwater facilities offer efficient cooling solutions, they remain bound by geographical limitations and are maintenance-intensive.

Priya Krishnamurthy, director at Altos India, notes that orbital datacentres present compelling opportunities, particularly in accessing near-limitless solar energy and reducing reliance on terrestrial infrastructure. They also promise to lower latency for specific satellite-based services and minimize land and water usage compared to traditional facilities.

Sen further explains that orbital datacentre nodes can achieve response times of under 15 milliseconds, a significant improvement over the 100 to 200 milliseconds typical for terrestrial datacentres. This capability is crucial for real-time AI inference applications, which cannot afford delays.

The strategic implications of this initiative are also noteworthy. Punit Badeka, co-founder of EON Space Labs, argues that the future of warfare may be in space, underscoring the need for countries to develop robust infrastructure in this domain.

To compete with terrestrial datacentres, traditional metrics such as power usage effectiveness (PUE) must be reevaluated. Sen points out that PUE was originally designed for environments where power, cooling, and land were predictable. However, as AI demands grow, the limitations of PUE become apparent. A space-based datacentre, powered entirely by solar energy and devoid of mechanical cooling, could achieve PUE figures approaching 1.0, maximizing power efficiency for computing tasks.

Challenges and Considerations

Despite the potential advantages, the environmental impact of launching datacentre systems into space is a significant concern. Sen acknowledges the need to address the launch footprint and the consumption of Earth's resources during deployment. Agnikul’s upper-stage model aims to mitigate this by repurposing hardware that would otherwise contribute to space debris.

The challenges of operating in space are substantial. Building an orbital datacentre entails addressing issues such as radiation exposure, component degradation, thermal management in a vacuum, and limited maintenance options. Krishnamurthy notes that while the absence of atmospheric heat in space may reduce the need for traditional cooling, thermal management remains a critical design challenge.

Moreover, the difficulty of human intervention in space complicates maintenance. Repairs or upgrades would be infrequent and costly, often relying on robotics or future in-orbit servicing capabilities. Instead of traditional maintenance, operators may focus on hardware refreshes, opting to de-orbit outdated modules and replace them with new ones.

As the pilot launch approaches, NeevCloud and Agnikul are validating their architecture and systems before scaling up. Sen envisions this project as a stepping stone towards democratizing AI, particularly for regions where access to terrestrial datacentres is limited. He estimates that the space datacentre sector could be valued at $3 billion to $5 billion within the next few years, potentially growing to a $50 billion global market by 2040.

Sen concludes that India’s initiative is not merely about infrastructure; it is about ensuring that the future of global AI intelligence remains independent and resilient, particularly in a world where geopolitical tensions could disrupt access to vital computing resources.