Agnikul, NeevCloud to launch India’s first orbital AI data centre platform

Space startup AgniKul Cosmos and Indian cloud and AI company NeevCloud plan to launch an AI-focused data centre platform in low-earth orbit (LEO), with a proof-of-concept mission targeted before the end of this year and commercial operations expected in 2027, according to their top executives.

The concept of space-based data centres has gained traction globally amid the growing energy demands of AI workloads. AgniKul and NeevCloud’s plan could position India at the forefront of this trend.

While companies such as SpaceX’s Starlink and major US technology firms have explored related ideas, no commercial deployments have yet been announced.


The initiative by AgniKul and NeevCloud, described by the companies as India’s first privately-led orbital AI inferencing infrastructure project, will see Chennai-based rocket maker AgniKul host NeevCloud’s AI-powered data centre on its extended upper stage: the top portion of its launch vehicle that typically releases satellites before being discarded.

“We had built our rocket’s upper stage as an extendable platform,” said Srinath Ravichandran, cofounder and chief executive of AgniKul Cosmos. “What we are doing now is allowing that upper stage to host their AI data centre in space. It becomes a platform on which their systems will run.”

A shared-hardware model

Unlike conventional satellite missions, where the upper stage ejects the payload and is subsequently discarded, AgniKul has developed patented technology to extend the life of its upper stage, effectively transforming it into a hosting station in orbit.

Ravichandran said this shared-hardware approach, where the rocket’s upper stage and the data centre operate as a single integrated system, could be a world first.

“Every other space data centre concept we are hearing about keeps the rocket and the satellite as separate bodies. Here we are sharing hardware. That makes it more cost-efficient, more compact and more controlled,” he said.

Under the partnership, AgniKul will deploy NeevCloud’s payload into LEO, potentially from Sriharikota, though final launch details will depend on orbit requirements and regulatory approvals.

The mission will launch from India, and Ravichandran described it as an “India story”, with both rocket and data centre hardware to be manufactured domestically.

The initial mission will carry a total payload of up to 500 kg, of which roughly 100 kg will account for the hosting platform itself.

NeevCloud’s first satellite configuration is expected to weigh 300 -350 kg and will be equipped with AI chips, storage and compute systems. The remaining mass allocation will go toward the company’s data centre modules.

The satellite will operate at an altitude of 350-500 km in LEO, circling the planet approximately 16 times a day.

Orbital inferencing layer

NeevCloud chief executive Narendra Sen, describes the project as an “orbital inferencing layer” designed to deliver low-latency AI services directly from space. The focus is not AI model training but real-time inferencing: the rapid processing of live data for applications requiring minimal latency.

“AI is growing very fast, but infrastructure is not coming with that speed,” Sen said. “We are solving the inferencing problem, not the training problem.”

The first configuration is expected to support some 500 high-performance AI chips. According to the company, a single satellite could handle up to 100,000 concurrent users or process around 10 million AI-driven calls per day, depending on workload intensity.

The companies aim to scale the deployment to 30-40 satellites by 2027, depending on demand and performance. Ravichandran said scaling to 30-40 satellites is feasible, while Sen indicated expansion would require additional capital.

The first satellite is estimated to cost about $1 million (₹8 crore-₹10 crore) and will be funded internally. NeevCloud is not currently raising capital but expects further funding will be required as the constellation expands.

The collaboration is structured as a long-term partnership, with each orbital platform expected to have an operational life of around five years before replacement or upgrades.

Energy and cooling advantages in space

Ravichandran said space offers structural advantages for running high-performance computing systems.

“Cooling data servers is much easier in space because you are exposed to near absolute zero temperatures,” he said. “At the same time, solar cells are more efficient since they can be deployed larger and get direct exposure.”

NeevCloud’s orbital system will draw power from solar panels and store energy in onboard batteries. However, continuous power supply remains a design challenge. A satellite in LEO spends about 60% of its time in sunlight and 40% in eclipse during each orbit.

“There is mission design involved to ensure maximum sunlight exposure…,” Ravichandran said.

Vijayakumar Nadar, head of engineering and product development (AI) at NeevCloud, said, “We need to store energy when the satellite is exposed to the sun and manage workloads intelligently when it moves into the dark side.”

AgniKul will provide the necessary power infrastructure according to customer specifications, including optimising solar panel orientation and energy storage to manage eclipse periods.

Space-hardened AI hardware

Beyond power and thermal management, the project requires qualifying terrestrial data centre hardware for the harsh space environment.

“Typical GPUs used on earth will have to be radiation hardened,” Ravichandran said. “You will have to take them through thermal cycles, radiation cycles and a full qualification process to make them space-ready.”

Vijayakumar said the system is being designed specifically for orbital conditions, including radiation tolerance, autonomous operations and energy-aware workload orchestration.

Once qualified, Ravichandran said operational costs could be significantly lower compared to terrestrial data centres built on large tracts of land with heavy power and cooling infrastructure.

Applications and insurance

The project is aimed at serving latency-sensitive applications in sectors such as defence, border security, maritime operations, oil and gas monitoring, disaster response, drones, robotics and remote healthcare, particularly in areas where terrestrial connectivity is limited.

Both launch and satellite insurance will be taken as mandated for space missions, Ravichandran said, with final structuring to be decided closer to launch. Sen acknowledged that insurance coverage for orbital assets remains limited, adding that risk evaluation is being undertaken in coordination with AgniKul and technology partners.

Boost to India’s private space ecosystem

Sen said India’s growing private space ecosystem, lower launch costs and returning aerospace talent have made such initiatives viable for domestic companies.

“Ten years ago, the ecosystem was not ready. Today, low-cost launch vehicles and technical capability are available,” he said. “Capital is not the constraint. Innovation and willingness to solve the problem are more important.”

Ravichandran described the mission as “another sign of victory for the privatisation of the space sector.”

For AgniKul, the partnership represents both a new revenue stream and a new class of customer beyond traditional imaging and communications satellites.

“If this succeeds, many will accept this as a better way to operate data centres. That means more launches and more opportunities for Indian companies,” Ravichandran said.

If successful, the project could position NeevCloud as an early mover in scalable, space-based edge AI infrastructure.