Chennai-based Agnikul Cosmos plans to launch a prototype of an artificial intelligence (AI) data centre into orbit by the end of 2026, with the goal of making it commercially viable by 2027, co-founder Srinath Ravichandran said. AI data centres are typically massive facilities used to train AI models and process vast amounts of data.
Prototype to be hosted on rocket upper stages
Agnikul’s prototype will be hosted on the extendable upper stage of its rockets. The data centre will primarily be used for AI inference—where a trained model analyses new data and makes predictions—which requires significantly less energy and infrastructure compared to model training.
While Agnikul Cosmos will handle the launch, the AI data centre will be developed by Bengaluru-based NeevCloud, an AI SuperCloud platform, as announced by the company on February 12.
Rising demand for data centre capacity
With the rapid expansion of AI applications, the volume of data that needs to be stored and processed has increased significantly. According to global consulting firm McKinsey & Company, global demand for data centre capacity could grow annually between 19 per cent and 22 per cent from 2023 to 2030, reaching 171 GW to 219 GW per year.
Traditional data centres require enormous amounts of power and water for cooling, making it increasingly difficult to find suitable locations. Companies such as SpaceX, Google and Axiom Space are exploring ways to deploy data centres in space.
Why put data centres in space?
Ravichandran said space offers nearly unlimited solar energy and highly efficient cooling due to exposure to extremely low temperatures. He also noted that an orbital data centre would be physically safer and harder to access, enhancing data confidentiality.
The system will not function as a single satellite. Instead, a constellation of satellites—or upper stages of different rockets—in low Earth orbit will host separate data centre modules.
The network will operate through intra-satellite and satellite-to-ground communication. During the eclipse phase, when satellites pass through Earth’s shadow and lose direct sunlight, they will operate in low-power mode for tasks such as data transmission. They will switch to high-power mode when exposed to sunlight. This operational framework already exists for imaging satellites.
Radiative cooling and commercial plans
The satellites will use radiative cooling to reduce temperatures, a process in which heat is lost through thermal radiation. Agnikul Cosmos claims to have worked extensively on this technology.
Although the prototype launch by the end of 2026 will not immediately commercialise the data centre, Ravichandran said it will demonstrate the company’s capability to build AI infrastructure in space.
“From 2027, we will be in commercialisation mode, which will open up a new class of customers and significantly increase our market,” he said.
ALSO READ: School children Aadhaar biometric update: UIDAI covers 1.2 crore students across 1 lakh schools in 6 months