The Indian government is developing a high performance AI compute system at its National Informatics Centre (NIC) in New Delhi with a capacity of around 1.1 exaFlops as a step towards reducing dependency on foreign AI companies.
In a written reply to questions raised by member of parliament (MP) Thiru Dr S Jagathratchakan, Union Minister of State for Electronics and Information Technology Jitin Prasada informed the parliament on August 5 that the union government has issued a purchase order for the high-performance AI compute system, which is expected to be ready this year.
MP Jagathratchakan had asked the details of initiatives undertaken by the Government to expand sovereign artificial intelligence infrastructure, strengthen domestic computing capacity, support development of indigenous platforms and reduce dependence on foreign technologies and service providers.
“India’s compute ecosystem currently draws on globally sourced GPUs procured through empaneled Compute Service Providers, consistent with prevailing international supply chains for AI hardware. Steps taken to reduce this dependence over time include the High-Performance AI Compute System taken up for establishment at the NIC Data Centre, Shastri Park, Delhi,” informed minister Prasada in a written reply to the parliament.
The National Data Center Shastri Park (NDCSP) Delhi under NIC was set up in 2011 and has around 475 Racks that hosts National Cloud as well as a number of Application/Services from various Government Departments in co-location mode.
A NIC tender document for cybersecurity staffing at the facility, reviewed by AI FrontPage, describes that the NDCSP is equipped with ICT equipment like Enterprise Class Storage arrays, LAN and SAN switches, Security devices, Distributed Denial of Services (DDoS), Network next-generation Firewall, SSL Offloader, Network Intrusion Prevention System (NIPS), Anti Advanced Persistent Threat (APT), Web Application Firewall (WAF), Antivirus tools, Vulnerability Assessment tools, Network Load Balancers, Blade Servers, Backup Infrastructure, Tape Library, VTLs, Linux and Windows Operating System, VMware, OpenStack and Microsoft Cloud Stack.
Meanwhile the minister informed that the government is pursuing a multi-pronged strategy under the IndiaAI mission and India semiconductor mission to build technological sovereignty and support development of indigenous platforms.
Under the IndiaAI mission, the government had identified 20 indigenous foundation proposals, including 12 large multimodal models and eight small language models for support from 506 applications received.
The intellectual property for these models will remain with the applicants. The supported projects include models developed by Sarvam AI, Gnani.AI, BharatGen and Avataar AI.
In order to expand computing capacity, the government has empaneled 15 compute service providers and approved 237 projects for subsidized compute support, sanctioning 93.18 lakh GPU hours.
According to the government the total data center capacity in the country has increased from about 375 MW in 2020 to around 1500 MW by 2025. To support AI development, about 38,231 GPUs have been onboarded through 14 empaneled service providers/data centers under the AI compute capacity framework.
Also Read: India Backs 20 Indigenous AI Models, Approves 58 Centres of Excellence: IT Ministry
