➔ India's Rising Supercomputing Ecosystem National Supercomputing Mission (NSM) & Indigenous Deep-Tech Stack (UPSC/RAS/PSI)

• Nodal Agencies- Jointly steered by the Department of Science and Technology (DST) and the Ministry of Electronics and Information Technology (MeitY).
• Implementing Partners- Centre for Development of Advanced Computing (C-DAC), Pune, and Indian Institute of Science (IISc), Bengaluru.

Macro Profile, Global Compute Context & Mission Architecture

1. From PARAM 8000 to NSM- India's indigenous high-performance computing (HPC) trajectory began with C-DAC’s PARAM 8000 in 1991 (operating at 1 GigaFLOPS) and evolved through PARAM Yuva (54 TeraFLOPS) into a distributed national grid under the NSM.
2. National Data Imperative- India generates nearly 20% of the world's data, driving the strategic necessity for sovereign supercomputing clusters to process large datasets without foreign compute dependence.
3. Compute Units & Global Speed Context-

a.    Supercomputing processing speed is measured in FLOPS (Floating Point Operations Per Second)-1 TeraFLOPS (TF) = 10¹² operations/sec; 

b. 1 PetaFLOPS (PF)= 10¹⁵ operations/sec;

C. 1 ExaFLOPS (EF)= 10¹⁸ operations/sec.

d.   The global supercomputing frontier reached approximately 2.19 ExaFLOPS in 2026.

4. The Three-Tiered 'Build Approach'- NSM executes an indigenous capacity lifecycle structured into three concurrent phases- (1) Component Assembly, (2) Sub-system Manufacturing, and (3) Complete Indigenous Design & Manufacturing Support.
5. Target vs. Deployed Metrics-
   

a. Target- Deploy 50 supercomputing systems with an aggregate compute power exceeding 123 PetaFLOPS.

b.Operational Delivery- Deployed 40 supercomputer systems delivering 68 PetaFLOPS of cumulative compute across national research and higher educational institutions.

6. Alignment with Viksit Bharat & ESTIC-2026- HPC infrastructure features as a core pillar at the CSIR-organized Emerging Science, Technology and Innovation Conclave (ESTIC-2026) at Bharat Mandapam (theme- "Creative Minds, Catalyzing Science, Empowering Technologies").
   
   

System Classification & Deployed Grid Footprint

1. High-End Compute Tier (>1 PF)- 13 systems commissioned to support advanced foundational research in atomic physics, molecular dynamics, and climate modeling.
2. Mid-Range Compute Tier (500 TF to 1 PF)- 12 systems deployed across premier Indian Institutes of Technology (IITs), National Institutes of Technology (NITs), and CSIR research centers.
3. Entry-Level Compute Tier (<500 TF)- 15 systems installed to decentralize computational access for regional universities and tier-II technical institutes.
4. Distributed Network Backbone (NKN)- All deployed systems are interconnected via the National Knowledge Network (NKN), a high-speed multi-gigabit national data highway enabling remote resource pooling across geographical boundaries.
5. Execution Volume- The national supercomputing grid has processed over 1.5 crore compute jobs and supported >1,990 peer-reviewed research publications.
6. Researcher Reach- Provides compute access to over 16,000 researchers, including more than 2,900 doctoral (PhD) scholars across 400+ participating institutions.
   
   

Indigenous Hardware & Software Stack- Moving Beyond Assembly

1. Rudra Server Architecture- Indigenously designed by C-DAC to provide processing nodes for complex mathematical simulations and high-density AI/ML workloads.
2. Commercial Electronics Manufacturing Transfer- Intellectual property and manufacturing technology for Rudra servers were transferred to domestic Electronics Manufacturing Services (EMS) partners under Make-in-India guidelines.
3. PARAM Rudra Supercomputers- Systems built using Rudra server blades and an indigenous system software stack; 6,000 Rudra servers are active on ground, with an additional 1,500 servers under fabrication.
4. High-Speed Interconnects (Trinetra Network)- Indigenously designed and validated inter-node low-latency communication fabrics operating at 100 Gbps and 200 Gbps to prevent data bottlenecks in parallel processing clusters.
5. Indigenous Direct Liquid Cooling- Engineered direct-contact liquid cooling hardware to maintain thermal efficiency in high-density server racks, lowering Power Usage Effectiveness (PUE).
6. PARAM Shavak ("Supercomputing-in-a-Box")- A compact, self-contained desktop supercomputer designed and manufactured in India to deliver entry-level HPC/AI capabilities to engineering colleges without specialized cleanroom data center infrastructure.
   
   

Strategic Applications of National Interest

Human Capital Development & Institutional Convergence

1. Cross-Disciplinary Faculty Upskilling- Joint collaboration between NSM and the All India Council for Technical Education (AICTE) to train non-computer science faculty in HPC, Machine Learning, and parallel programming.
2. Dedicated Pedagogical Frameworks- Annual EduHPC Workshops provide structured HPC curriculum kits, alongside specialized High-Performance Scientific Computing courses hosted on SWAYAM/NPTEL.
3. Centralized Knowledge Repositories- Deployed the HPC Shiksha Portal, granting open access to lecture materials, containerized algorithms, and virtual coding sandboxes.
4. Grassroots Bootcamps & Hackathons- Organizes regional deep learning bootcamps and student hackathons focused on applying Generative AI and HPC to rural development, linguistics, and agricultural logistics.
5. NSM Users Forum- An inter-institutional collaborative platform connecting users across 40 nodes to exchange system optimizations, parallel scripting practices, and bug resolutions.
6. Institutional Departmental Convergence- Computational modules directly feed predictive operations across statutory agencies, including the India Meteorological Department (IMD), Central Water Commission (CWC), Central Pollution Control Board (CPCB), and the Ministry of AYUSH.

Strategic Significance

1. Advancing Strategic Autonomy- Developing indigenous processors, interconnects, and cooling stacks mitigates the risk of external technology embargoes (such as the historic denial of Cray supercomputers in the late 1980s).
2. National Data Sovereignty- Prevents sensitive spatial, geological, genomic, and seismic datasets from being routed through or processed on foreign commercial cloud platforms.
3. Enabler for Sovereign AI (BharatGen / IndiaAI)- Large Language Models (LLMs) and Multimodal AI require tens of thousands of compute hours; domestic supercomputing infrastructure provides the compute backbone for national AI strategies.
4. Progress Across UN SDGs- Advanced compute modeling directly accelerates progress across 11 UN Sustainable Development Goals, spanning climate resilience (SDG 13), clean water access (SDG 6), good health (SDG 3), and industrial innovation (SDG 9).
5. Challenges to Address- Closing the remaining gap to achieve the planned 123 PF target, transitioning from sub-system assembly to indigenous fabrication of silicon GPUs/accelerators, and improving energy efficiency across high-density facilities.

Source: PIB