India’s push for sovereign artificial intelligence infrastructure is taking an unexpected turn: towards space. Space-tech startup Pixxel and AI firm Sarvam AI are collaborating on a project that aims to place computing power in orbit. Their goal is to train and run large language models (LLMs) directly in space.
At the centre of this initiative is a satellite called The Pathfinder, a 200-kg platform designed to function as an orbital data centre. Slated for launch by the end of 2026, it is expected to carry high-performance GPUs similar to those used in Earth-based AI facilities. Sarvam AI plans to use this infrastructure not just for deployment, but also to train its models in orbit—potentially marking India’s first homegrown LLM developed beyond Earth.
Read More | India’s $15 Billion AI Data Centre in Visakhapatnam Signals Big Tech Push
Why are companies looking at space for computing?
The idea of moving data centres to space is no longer fringe. It’s being driven by a very real constraint on Earth: energy. AI systems are becoming increasingly power-hungry. As companies scale up infrastructure to support more complex models, electricity consumption, land use, and environmental impact are becoming major bottlenecks.
- Global data centre capacity could reach 200 GW by 2030
- India’s capacity alone may grow from 1.8 GW to 10.5 GW by 2031
This rapid expansion is already facing pushback in some regions.
Space offers a theoretical workaround:
- Continuous access to solar energy without atmospheric loss
- Proximity to space-based data sources like Earth observation satellites
- Reduced the need to transmit massive datasets back to Earth
These advantages have caught the attention of global tech players such as Google, SpaceX, and Meta, all of whom are exploring aspects of orbital computing or space-based energy systems.
Read More | AI Job Fears Rise—Mayor Sadiq Khan Appoints Taskforce Chief to Safeguard Workers
Even the Indian Space Research Organisation (ISRO) has signalled interest in enhancing onboard data processing capabilities in future missions.
What makes The Pathfinder different?
Traditional satellites are built with low-power, radiation-hardened chips designed for longevity. The Pathfinder flips that model. It is being designed to host cutting-edge computing hardware—closer to what you’d find in modern AI data centres.
Key aspects of the mission include:
- Testing AI workloads in space conditions
- Managing heat dissipation in a vacuum
- Evaluating power efficiency of high-performance GPUs
- Running real-time data processing and AI inference in orbit
Pixxel will build, launch, and operate the satellite from its upcoming Gigapixxel facility, which is intended to scale production to 100 satellites.
Why this matters for India’s AI strategy
For Sarvam AI, the project goes beyond experimentation. Running models in space aligns with a broader goal, that is, to build a fully indigenous AI stack. By reducing dependence on foreign cloud infrastructure, the company is also addressing concerns around data sovereignty.
If successful, this approach could reshape how and where AI systems are developed, especially for countries looking to retain control over critical digital infrastructure.
The big challenges: cost, maintenance, and viability
Despite the promise, orbital data centres remain a high-risk bet. Cost is the biggest hurdle. Launching hardware into space is still expensive. For large-scale deployment to make sense, launch costs would need to fall dramatically, from thousands of dollars per kilogram today to a few hundred.
Read More | Lost Your Job to AI? China Says That’s Not Legal—Here’s Why
Maintenance is another major concern. Unlike Earth-based facilities, you can’t easily replace or upgrade faulty GPUs in orbit. Any failure could mean significant losses.
There are also broader questions around:
- Long-term reliability of high-performance hardware in space
- Efficiency gains versus total lifecycle cost
- Scalability of orbital infrastructure
A long-term bet on the future of computing
Even with these uncertainties, the Pixxel–Sarvam initiative reflects a growing belief that the future of computing may not be confined to Earth.
If the Pathfinder succeeds, it could validate a new model, one where data is processed closer to where it is generated, and where the lines between space infrastructure and digital infrastructure begin to blur. For India, it’s an early but significant step in exploring what sovereign AI could look like in the decades ahead.
