Samiksha Muskan
Background
The Union Cabinet approved the IndianAI Mission in March 2024 with a budget outlay of Rs. 10,371.92 crore as a national initiative to build an end-to-end artificial intelligence ecosystem in India. Officially, the mission seeks to democratise access to compute, improve data quality, develop indigenous AI capabilities, attract talent, support start ups, encourage socially relevant applications and promote responsible AI. The larger policy vocabulary surrounding the mission is significant: the Cabinet note explicitly links IndiaAI to “tech sovereignty”, indicating that the programme is not merely industrial policy but also a strategic response to concentrated foreign control over advanced AI infrastructure.
Compute sovereignty refers to a country’s ability to secure and govern access to the computational infrastructure, such as, high end Graphics Processing Units (GPUs), cloud capacity, storage and network architecture required to train and deploy advanced AI systems. Strategic autonomy, by contrast, does not require full self-efficiency but it aims to preserve decision making freedom while working within global technology markets, supply chains, and partnerships. IndiaAI sits at the intersection of these: it seeks to reduce structural dependence on a handful of foreign technology providers without assuming that complete technology autarky is feasible.
This framing matters because AI power today is increasingly shaped by access to compute. OECD analysis notes that countries need national compute planning not only for capacity but also for resilience, innovation effectiveness and sovereignty related concerns. A complete AI sovereignty is difficult because AI systems depend on globally dispersed supply chains, cloud infrastructure, models, and standards. For India, therefore, the core question is whether the IndiaAI Mission can build enough domestic capability to avoid a “compute divide” while still leveraging foreign hardware, platforms and partnerships when needed.
Functioning
The IndiaAI Mission is structured around seven pillars: IndiaAI Compute, IndiaAI Innovation Centre, IndiaAI Datasets Platform, IndiaAI Application Development Initiative, IndiaAI FutureSkills, IndiaAI Startup Financing, and Safe & Trusted AI. Together, these pillars form an integrated institutional architecture rather than a programme centred only on compute. While the compute pillar provides the foundational infrastructure, the other components ensure access to quality datasets, indigenous model development, applied solutions in key sectors, skill-building, startup support, and responsible AI governance. In this way, the mission links infrastructure with capability-building and public-interest outcomes under a single policy framework.
The compute pillar is the most strategically consequential. The 2024 Cabinet note states that IndiaAI will build a scalable ecosystem of 10,000 or more GPUs through public-private partnership and create an AI marketplace offering AI-as-a-service and pre-trained models.
By March 2025, the government had launched the AI Compute Portal, providing access to 10,000 GPUs, with another 8,693 to be added, and that eligible users could receive subsidies of up to 40% on cloud-based AI compute services. The portal is intended to widen access for startups, MSMEs, students, researchers, academia, and government agencies.
AIKosha, which is the datasets platform, was launched as a repository of ethically sourced, consent-based and non-personal datasets. The Innovation Centre is mandated to develop indigenous large multimodal and domain-specific foundation models, while the Application Development Initiative promotes AI solutions for public-interest sectors such as healthcare, agriculture, governance, climate and disaster management.
FutureSkills, Data Labs in Tier 2 and Tier 3 cities, fellowships and startup financing are to strengthen India’s human capital.
Performance
The mission was approved only in 2024, so the performance should be assessed as early-stage implementation. The government has launched the AI Compute Portal, AIKosha, competency frameworks, acceleration programmes and application challenges within a year. This is important because most AI related strategies in India have been mere vague ambitious documents.
There are also concrete early indicators of traction. According to the March 2025 PIB release, the compute portal became the largest component of the mission by receiving nearly 45% of the mission’s funding. It also states that a call of indigenous foundation models received 67 submissions within 15 days, while the IndiaAI Innovation Challenge attracted over 900 AI solutions.
At the same time, performance remains uneven when evaluated against the claim of compute sovereignty. The portal offers access to GPUs including NVIDIA H100, Intel Gaudi 2, AWS Trainium, among others, which demonstrates breadth of access but also reveals continuing dependence on foreign hardware and cloud ecosystems. In other words, India may be improving domestic access and bargaining power but it is not yet sovereign across the full stack of semiconductor design, fabrication, advanced accelerators, hyperscale cloud and frontier model infrastructure.
Impact
The most immediate impact of the IndiaAI Mission is distributive. By subsidizing access to expensive AI compute and making datasets and tools available through public digital platforms, the mission lowers barriers to entry for researchers, startups, students and smaller firms that would otherwise struggle to compete with large incumbents. In that sense, IndiaAI is trying to democratise not only AI use but also AI production capacity.
Another lens to assess the impact is strategic. A domestic compute layer, even if built through partnerships and cloud procurement, can reduce vulnerability to price shocks, platform gatekeeping, export controls or restrictions in accessing critical AI infrastructure. Resilience is a key component of the national mission and interdependence is unavoidable but can be minimised through diversification and institutional design. IndiaAI, thus, strengthens strategic autonomy by ensuring that Indian innovators and public institutions are less exposed to external control.
The mission does not intend on framing AI only as a national security or corporate productivity project; it is also embedding AI into a wider developmental goal in public-sector skilling and sectoral transformation in areas of healthcare, governance, agriculture, learning disabilities and climate resilience. This combination could become a distinctive Indian model in enhancing digital public infrastructure and inclusion given that implementation remains broad based and accountable.
Emerging Issues
Several issues complicate the narrative of compute sovereignty. First, access is not the same as sovereignty. If the GPUs, cloud stack and much of the enabling hardware remain foreign, India gains operational control of the underlying technological chokepoints. China’s unsuccessful attempt is an example that complete sovereignty is unrealistic and a more pragmatic goal is managed dependence.
The absence of strong domestic capability in chips, data governance, model evaluation and public procurement can create a shallow ecosystem. Compute capacity alone is insufficient unless linked to skills, innovation institutions, policy effectiveness and resilience. IndiaAI recognizes this in design, but implementation gaps can still emerge if the compute layer outpaces the growth of research quality, domestic model-building or sector-specific adoption.
There are certain governance concerns. The mission’s ambition to use datasets at scale, build foundation models, and accelerate deployment in governance and public services raises questions about privacy, dataset quality bias, accountability, security, and the institutional meaning of “safe and trusted AI.” The mission’s legitimacy will depend not only on infrastructure volumes but also on transparent rules, public safeguards and measurable public value.
Way Forward
India’s most realistic objective should be strategic autonomy through layered compute sovereignty, not complete self-sufficiency. This approach is necessary because IndiaAI still faces governance and implementation challenges; access does not mean equal ownership, reliance on foreign hardware and cloud stacks persists, and compute capacity alone cannot guarantee responsible or inclusive AI. The policy response should focus on building domestic capacity where dependency is most risky, like in public-interest compute access, secure datasets, model evaluation, public-sector AI tools, and some indigenous models, while continuing to diversify foreign partnerships hardware and cloud services.
To address implementation gaps, India should adopt a medium-term national compute plan with benchmarks for capacity, utilization, affordability, and regional access and publish periodic data on portal usage, allocation criteria, subsidy beneficiaries, research outputs and model-development outcomes. This would improve transparency and allow the mission to be assessed against its stated goals rather than broad announcements.
To address governance risks, India should institutionalize independent auditing for datasets, foundation models, and public-sector AI deployments under the Safe & Trusted AI pillar. Public-private partnerships should also be designed to strengthen domestic learning, local capability, and portability across vendors, so that India avoids long-term lock-in to a small set of platforms. The broader lesson is that IndiaAI will succeed not by proving total sovereignty, but by creating enough national capacity and policy flexibility for India to make meaningful AI choices on its own terms.
References
Carnegie Endowment. (2026, June 17). Early Lessons in the Pursuit of Sovereign AI. Carnegie Endowment for International Peace. Retrieved July 19, 2026, from https://carnegieendowment.org/research/2026/06/early-lessons-in-the-pursuit-of-sovereign-ai
Government Of India. (n.d.). INDIAai | Pillars. Retrieved July 19, 2026, from https://indiaai.gov.in/
Government of India. (2024, March 7). Press Release:Press Information Bureau. Press Release:Press Information Bureau. Retrieved July 19, 2026, from https://www.pib.gov.in/PressReleaseIframePage.aspx?PRID=2012355®=48&lang=2
Government of India. (2025, March 6). Press Information Bureau. Press Information Bureau. https://www.pib.gov.in/PressReleasePage.aspx?PRID=2108961®=48&lang=2
Organisation for Economic Co-operation and Development. (n.d.). AI compute. OECD. Retrieved July 19, 2026, from https://www.oecd.org/en/topics/sub-issues/ai-compute.html
Organisation for Economic Co-operation and Development. (2023, February 28). A blueprint for building national compute capacity for artificial intelligence. OECD. Retrieved July 19, 2026, from https://www.oecd.org/en/publications/a-blueprint-for-building-national-compute-capacity-for-artificial-intelligence_876367e3-en.html
About the Contributor
Samiksha Muskan is a Research and Editorial Intern at IMPRI. She is currently pursuing a BA (Honors) with Research in Political Science from Lady Shri Ram College for Women at Delhi University.
Acknowledgement
I would like to express my sincere gratitude to IMPRI for giving me this opportunity to write Policy Update and its intellectual encouraging environment. I would especially like to acknowledge Pallavi and Shivani Chauhan for giving valuable feedback and review.
