High-performance computing is critical infrastructure for India's AI ambitions and climate resilience. Discuss.

High-performance computing (HPC) uses thousands of processors working in parallel on one problem. Its speed is measured in petaflops (PF), where 1 PF is 10¹⁵ operations per second. Under the National Supercomputing Mission (NSM, 2015), India now has 40 systems with a combined 68 PF [1]. Like power grids, HPC has become a base layer on which both AI and climate security depend.

HPC as the backbone of AI ambitions

• Data scale: India generates nearly 20% of the world's data [1]. Training AI models on this data needs dense, sovereign compute.
• AI compute: AIRAWAT ranked 75th in the TOP500 at ISC 2023 and is scalable to 790 AI-petaflops [2].
• Wider research access: NSM systems serve 10,000+ researchers from 200+ institutions, including in Tier-II/III cities. They support work in drug discovery and materials research [3].
• Strategic autonomy: the indigenous Rudra servers and software stack [2] reduce dependence on imported hardware that other countries can restrict.

HPC as the backbone of climate resilience

• Forecasting capacity: Arka (IITM Pune) and Arunika (NCMRWF Noida) raised the Ministry of Earth Sciences (MoES) capacity from 6.8 PF to 22 PF [4].
• Faster warnings: the Bharat Forecast System runs at 6 km resolution. Its run time fell from about 12 hours to 3–6 hours, which makes real-time forecasts possible [5]. This matters most for cyclones, heavy rain and heatwaves.
• Adaptation planning: climate modelling, flood mapping and disaster management all rely on HPC [1][3].

Gaps that weaken this infrastructure

• Scale: 68 PF is only about 55% of the 123+ PF target. The world's fastest single machine runs at about 2.19 exaflops (EF), roughly 2,190 PF [1].
• Chip dependence: India's own processor, AUM, is still being designed. It is based on Arm Neoverse V2 and built with a Japanese partner [6].
• Congestion: most systems run at over 85% utilisation, and many above 95% [3]. Long queues hurt smaller institutions that have no machines of their own.
• Fragmentation: NSM (run by DST and MeitY) and the MoES weather HPC systems are separate tracks [1][4].

HPC therefore connects India's data, AI capability and disaster preparedness in one chain. A mixed path would close the gap: a fresh NSM timeline, AUM-based servers backed by the India Semiconductor Mission, early exascale research, and new capacity where queues are longest. Treated as critical infrastructure, HPC can advance Aatmanirbhar Bharat and SDG 13 (Climate Action) together.