Partnership · 1 March 2026NeevCloudAI inference infrastructure provider
PartnershipAI & Deep Tech
NeevCloud partners with Agnikul for orbital data centre
By Abha Lohia · Editor ·
- NeevCloud builds AI inference systems that process data without training models
- The company signed an MoU with Agnikul to deploy this tech in low Earth orbit satellites
- This aims to cut latency below 10ms by avoiding terrestrial network delays and bottlenecks
What happened
- NeevCloud and Agnikul signed a MoU in February 2026 to collaborate on space-based computing.
- The partnership focuses on deploying AI inference infrastructure within Low Earth Orbit satellites.
- The system will operate at approximately 300km altitude to provide ultra-low latency services.
- Target latency is sub-10ms, significantly faster than the 100-200ms typical of terrestrial networks.
- The orbital data centre uses solar power and vacuum thermal management for sustainability.
- Custom ASIC chips hardened against radiation will handle all inference tasks onboard.
- Initial power capacity is set between 10-15kW, scalable up to 200kW or more.
- Satellite modules have a lifespan of 4-5 years before de-orbiting and replacement.
Why it matters
- Terrestrial data centres face long build times of 24 months due to land and regulatory hurdles.
- India's current data centre capacity exceeds 1GW, projected to grow past 2GW by 2026.
- Global demand drives a 14% CAGR through 2030, reaching ~200GW total capacity.
- Orbital deployment bypasses traditional fibre and power grid constraints entirely.
The business
- NeevCloud specializes exclusively in inference-only workloads, excluding model training processes.
- The technology relies on radiation-hardened custom ASIC chips for reliable space operation.
- Failed modules are designed for de-orbiting and replacement rather than in-orbit repair.
- The constellation plan involves launching up to 600 satellites over three years.
- Power scalability allows expansion from initial specs to 50kW or 200kW+ per unit.
- The setup avoids terrestrial bottlenecks like land acquisition and fibre laying delays.
The founders
NS
Narendra Sen
CEOSR
Srinath Ravichandran
CEO- He leads Agnikul, a startup headquartered in Chennai.
In numbers
sub-10msLatency target
~300kmAltitude
10-15kWInitial Power
200kW+Max Scalable Power
4-5 yearsLifespan
up to 600 satellitesConstellation Size
24 monthsBuild Time (Terrestrial)
~200GWGlobal DC Capacity (2030)
>2GWIndia DC Capacity (2026)
What happens next
- The companies plan to launch a constellation of up to 600 satellites over three years.
Up against
SpaceXAxiom SpaceStarcloudGalaxEye
Words to know
- MoU
- Memorandum of Understanding; a formal agreement outlining intentions to cooperate.
- Inference
- Using trained AI models to make predictions on new data, distinct from training.
- ASIC
- Application-Specific Integrated Circuit; custom chips designed for specific high-performance tasks.
- De-orbit
- Deliberately lowering a satellite's orbit so it burns up in the atmosphere.
The numbers around it
AI & Deep Tech in March 2026: 11 rounds with a disclosed amount, $79.8 million in all; the biggest was Deccan AI ($25 Mn). Every sector this year →
Read next
FundingSatSure raises ₹60 crore from Antariksh fundSatSure is a Bengaluru startup that turns satellite images and AI into insights for banks, insurers, infrastructure firms and government agenciesFundingUnsiloed AI raises $7.5M seed roundUnsiloed AI builds software interfaces that convert complex documents like PDFs into clean structured data for AI agents and enterprise toolsFundingCyient Semiconductors gets $10M from GlobalFoundriesCyient Semiconductors is a Hyderabad-based chip firm making custom ASICs, with strengths in analog mixed-signal and power chips
All startup news on 1 March 2026 →
Is this NeevCloud? Get your “Featured on StopDown” badge →