NSM  •  India AI Mission  •  AI Governance  •  Science & Technology Policy 

Study OAS Prism   |   11 August 2026

Computational capacity has emerged as a decisive factor of national power in the 21st century. Just as nations once competed for steel mills and nuclear reactors, today the race is for GPU clusters, exaflop systems, and AI infrastructure. When Prime Minister Narendra Modi inaugurated the “Param Pragya” AI-enabled high-performance computing facility at IIT Delhi’s Sonipat campus on August 8, 2026, it was not merely an institutional milestone — it was a statement of strategic intent.

Param Pragya sits within a layered national architecture built through two flagship programmes: the National Supercomputing Mission (NSM), which has been building India’s HPC spine since 2015, and the IndiaAI Mission (2024), which is focused on AI-specific GPU infrastructure and ecosystem development. Together, they represent India’s bid for computational sovereignty — the ability to train frontier AI models, run climate simulations, design defence systems, and drive drug discovery without depending on foreign infrastructure.  

For OPSC and UPSC aspirants, this ecosystem is essential preparation. The NSM, AIRAWAT, IndiaAI Mission, and the PARAM series appear regularly in both Prelims MCQs and Mains analytical questions under GS Paper III (Science, Technology & Environment).

2.1  Origin and Governance

The NSM was approved by the Union Cabinet in April 2015 with an outlay of ?4,500 crore over seven years. It is jointly implemented by two agencies — the Centre for Development of Advanced Computing (C-DAC), Pune, and the Indian Institute of Science (IISc), Bengaluru — under the dual administrative oversight of the Department of Science and Technology (DST) and the Ministry of Electronics and Information Technology (MeitY). This two-agency, two-ministry structure was designed deliberately: C-DAC manages infrastructure deployment and national rollout, while IISc anchors fundamental research integration.

The mission’s mandate is tri-dimensional: first, deploy a network of over 70 interconnected supercomputers at premier Indian institutions; second, build indigenous hardware capabilities including processors and high-speed interconnects; third, train a skilled HPC workforce to sustain and utilise these systems at scale.

2.2  Three Phases of Execution

2.3  The PARAM Family: 

2.4  Progress Against Targets (as of March 2025)

Only 34 supercomputers have been installed across 24 institutions against a target of 70+. Combined capacity stands at 35 petaflops against the 50+ petaflop goal. Yet utilisation metrics are strong: systems operate at 85–95% utilisation with over 95% uptime. Over 10 million compute jobs have been executed, enabling 10,000+ researchers and producing more than 1,500 peer-reviewed publications. Close to 22,000 individuals have received formal HPC training. The shortfall is not in demand but in supply — driven primarily by India’s 90–95% semiconductor import dependency, which is the single most critical structural vulnerability in India’s computing sovereignty agenda.

Param Pragya, inaugurated by PM Modi on August 8, 2026 during IIT Delhi’s 57th Convocation at its Sonipat campus, is an AI-enabled high-performance computing facility. It differs conceptually from the NSM’s standalone PARAM systems: while AIRAWAT and PARAM Rudra are national mission assets shared across a researcher community, Param Pragya is an institutional compute facility designed to serve IIT Delhi’s specific research ecosystem in AI, data science, advanced materials, and interdisciplinary computing.

The significance lies in scale: IIT Delhi executed sponsored R&D projects worth approximately ?600 crore in FY 2025–26. Its Sonipat campus SATHI (Sophisticated Analytical and Technical Help Institute) served over 4,000 national researchers in a single year. Param Pragya is designed to multiply this impact by transforming the campus into a node in India’s emerging national compute network rather than an isolated island of excellence.

The facility also reflects a broader policy shift embedded in the IndiaAI Mission: treating compute as shared strategic infrastructure — much like roads or power grids — rather than isolated institutional assets. This conceptual shift will appear in Mains questions on the governance of digital public infrastructure.

3.1  Connection to IndiaAI Mission and International Partnerships

Param Pragya connects to the IndiaAI Mission’s compute expansion effort. Under the mission, India’s national GPU capacity has been scaled from an initial 10,000 GPU target to 38,000 GPUs and 1,050 TPUs, with plans to add 20,000 more units. Internationally, Abu Dhabi’s G42, along with MBZUAI, Cerebras Systems, and C-DAC, has partnered to deploy exaflop-scale AI infrastructure in India — a significant development that signals the internationalisation of India’s compute strategy.

The Union Cabinet approved the IndiaAI Mission in March 2024 with a total outlay of ?10,372 crore. Union Budget 2024–25 made an initial allocation of ?551.75 crore. The mission addresses the full AI ecosystem stack through six interlocking pillars:

The mission’s architectural logic creates a compute-to-application pipeline: infrastructure → datasets → models → applications → skills. Param Pragya and AIRAWAT feed this pipeline at its base. A student who understands this pipeline can answer both MCQs on individual components and Mains questions on India’s overall AI strategy.

5.1  The Existing Framework

India’s approach to governing AI currently rests on three pillars. NITI Aayog’s National Strategy for Artificial Intelligence (2018) established the “AI for All” vision across five priority sectors: healthcare, agriculture, education, smart cities, and transport. The Digital Personal Data Protection Act (DPDP Act), 2023, provides the foundational data rights framework — granting data principals rights of consent, correction, and erasure — though enforcement remains nascent. Sector-specific guidelines from the RBI (for financial AI) and the Ministry of Health (for telemedicine AI) address domain-specific risks. In November 2025, the government released IndiaAI Governance Guidelines, covering transparency norms, human oversight requirements, and accountability mechanisms for AI in public services.

5.2  The Critical Gap: Absence of a Dedicated AI Regulator

Unlike the EU’s AI Act (2024), which mandates a risk-based classification system and bans “unacceptable risk” AI systems outright, India has no dedicated AI regulatory authority. The proposed “AI Regulatory Authority of India” (AIRAI) — envisioned as a multi-stakeholder body with technologists, ethicists, lawyers, and civil society members — remains at consultation stage. This is a significant governance gap for a country deploying AI in credit scoring, judicial support, agricultural advisory, and welfare distribution. Automated decision-making affecting fundamental rights operates without any statutory review mechanism — an accountability vacuum with no EU or UK equivalent.

5.3  India and the World: Governance Approaches Compared

India’s “middle path” — risk-informed but innovation-promoting — is defensible for a developing economy that cannot afford to stifle transformative technology with regulatory overreach. But the absence of algorithmic accountability legislation means AI decisions affecting fundamental rights operate without judicial review provisions. This gap narrows only when AIRAI moves from consultation to statute.

6.1  Semiconductor Import Dependency

India imports 90–95% of its semiconductor requirements. Every HPC system in the NSM — including Param Pragya — ultimately depends on chips manufactured abroad. The India Semiconductor Mission (ISM) is attempting to address this through domestic fabrication units, but TSMC-equivalent domestic capacity remains a 10–15 year horizon. Any geopolitical supply chain disruption can stall India’s entire compute expansion agenda overnight.

6.2  Brain Drain and the Workforce Gap

NSM has trained nearly 22,000 HPC professionals. Yet the same talent pipeline feeds Silicon Valley and European research centres, where compensation structures are incomparably better. Without dedicated retention mechanisms — HPC fellowship programmes, competitive salaries at C-DAC and IISc, and industry-linked research stipends — India’s HPC workforce will remain thin and perpetually dependent on institutional training without institutional retention.

6.3  The Access-Utilisation Paradox

NSM systems run at 85–95% utilisation, which appears excellent. But this masks deep structural inequality: only 24 institutions out of 1,000+ universities nationwide host supercomputing systems. The universities that most need compute access — state universities, Tier-II engineering colleges, regional research centres — are the ones least represented. NSM’s model rewards pre-existing excellence but has not yet democratised access.

6.4  Energy Consumption and Sustainability

Each HPC system consumes 5–7 megawatts of power. India’s NSM network, even at its current 34-system scale, represents a significant and growing energy load. There are no mandated green computing standards for HPC infrastructure, even as India has committed to net-zero by 2070. As NSM scales toward its 70+ system target, this energy gap must be planned for, not resolved retroactively.

Odisha is not a passive observer of India’s AI and HPC transformation. The state has, over the past two years, assembled one of India’s most structured sub-national AI and semiconductor ecosystems — combining formal policy, institutional infrastructure, and industrial investment.

7.1  Odisha AI Policy 2025 and Governance Framework

The Cabinet-approved Odisha AI Policy 2025 establishes a two-tier governance architecture:

• AI Taskforce: Strategic oversight at cabinet level; chaired by the Chief Minister
• AI Cell under OCAC: (Odisha Computer Application Centre) — operational implementation and technical backbone
• Odisha AI Mission: A whole-of-government structure coordinating departments, academia, industry, and startups

 

7.2  Institutional AI Research Ecosystem

AI research labs are being established at four premier institutions: IIT Bhubaneswar, NIT Rourkela, IIIT Bhubaneswar, and NISER Bhubaneswar — focused on applied innovation in healthcare, agriculture, disaster management, and administrative services. A NASSCOM State AI Centre of Excellence operates in Bhubaneswar on a hub-and-spoke model, connecting government, academia, industry, and startups. Odisha has signed an MoU with BHASHINI for Odia language corpus expansion and speech-to-text systems — a critical step for AI-driven governance in a state where Odia is the primary language of administrative interaction.

7.3  Strategic Investments in AI and Semiconductors

 

7.4  Human Capital and Policy Ambitions

Odisha has set ambitious human capital targets: 15% of STEM graduates specialising in emerging technologies by 2029, rising to 55% by 2036; and 75% of state government officials trained in digital and AI fundamentals. The state has adopted an “open-first” approach, establishing a State Open AI Hub for open-source AI tools and supporting contributors through hackathons and micro-grants — an innovative governance mechanism rare at the sub-national level.

Odisha’s semiconductor policy offers India’s lowest power subsidy at ?2 per unit, making the state highly competitive for energy-intensive chip fabrication. The Digital Fab Lab at STPI Bhubaneswar provides VLSI and electronics design infrastructure, creating a near-complete semiconductor value chain from design to fabrication support within the state. The C-DAC partnership through SFAL (Semiconductor Fabless Accelerator Lab) further anchors Odisha in India’s national semiconductor value chain.

• Accelerate India Semiconductor Mission: Domestic fabrication capacity must be built before the next NSM phase begins. A 90–95% import dependency that holds every installed supercomputer hostage to global supply chains is not a tolerable long-term position for a country claiming HPC sovereignty.
• Legislate AIRAI: Move the proposed AI Regulatory Authority of India from consultation papers to statute. A statutory framework covering algorithmic accountability and automated decision-making rights is not merely desirable — it is constitutionally necessary once AI systems begin determining welfare eligibility, credit access, and judicial outcomes.
• Democratise HPC Access: A “Compute-as-Service” model — allowing state universities and regional colleges to access NSM systems remotely — should be piloted under the IndiaAI Mission. Odisha’s NIT Rourkela should be prioritised in the next round of NSM deployments, making it a regional HPC hub for eastern India.
• Mandate Green Computing Standards: Require renewable energy sourcing for all HPC systems above 1 petaflop and establish mandatory carbon footprint reporting — aligned with India’s 2070 net-zero commitment. Green supercomputing is a policy gap that must be closed before India scales to 70+ systems.
• Scale Odia Language AI: The BHASHINI-OCAC partnership must produce a full-scale Odia language foundation model. State-level AI-driven governance in agriculture, health, and administration requires AI that operates in Odia, not English. This is Odisha’s most strategically important AI investment.
• HPC Workforce Retention: Create dedicated HPC fellowship programmes at C-DAC and IISc with industry-competitive stipends. Without retention mechanisms, NSM’s training investment exits the country the moment it matures.