The NITI Aayog AI Mission story in 2026 is really a story about two connected but distinct government efforts: NITI Aayog’s original policy blueprint for artificial intelligence and MeitY’s IndiaAI Mission, which is now executing a rapid compute buildout on the ground. At the India AI Impact Summit 2026 in New Delhi, Union IT Minister Ashwini Vaishnaw announced that India would add 20,000 GPUs beyond its existing base of 38,000, with orders for a further 40,000 units to follow — a push officials are calling \”AI Mission 2.0.\” This explainer separates the policy groundwork from the compute rollout, lays out exactly what has been approved and deployed so far, and flags the practical questions that remain as India chases a 100,000-GPU target before the end of 2026.
NITI Aayog’s Role vs. MeitY’s IndiaAI Mission
It helps to be precise about which institution does what. NITI Aayog authored India’s foundational \”#AIforAll\” National Strategy for Artificial Intelligence back in June 2018, followed by a two-part Responsible AI approach paper in 2021 that proposed governance principles and enforcement mechanisms for AI systems. These documents set the policy direction — prioritising healthcare, agriculture, education, smart cities, and mobility, and recommending institutions like Centres of Research Excellence. Actual implementation of compute infrastructure, however, sits with the Ministry of Electronics and Information Technology (MeitY) through the IndiaAI Mission, approved by the Union Cabinet in March 2024 with an outlay of ₹10,371.92 crore (roughly $1.14 billion) over five years. NITI Aayog remains an advisory and strategy voice in India’s AI ecosystem, while MeitY and its IndiaAI Independent Business Division run the empanelment, subsidy, and deployment machinery for GPUs.
What Has Actually Been Approved and Deployed
As of mid-2026, the IndiaAI Mission’s common compute facility has onboarded 38,231 GPUs through empanelled private data centre operators across four completed empanelment rounds, a public-private partnership model designed to avoid the government building and owning data centres directly. These GPUs are made available to startups, academic researchers, and students at a subsidised rate of approximately ₹65 per GPU per hour — deliberately priced low so early-stage players are not squeezed out by global compute costs. Union Minister of State for Electronics and IT Jitin Prasada told the Lok Sabha that 190 projects have been approved under the mission: 78 with government entities, 46 with startups and MSMEs, 30 with early-stage startups, 27 with academia and researchers, five with students, and four with early-stage researchers.
Compute Expansion Timeline
| Milestone | Detail |
|---|---|
| March 2024 | IndiaAI Mission approved with ₹10,371.92 crore outlay over five years |
| Mid-2026 (current base) | 38,231 GPUs onboarded via common compute facility; 190 projects approved |
| February 17, 2026 announcement | 20,000 additional GPUs ordered within a week under \”AI Mission 2.0\” |
| Six-month rollout target | Over 50,000 new GPUs to be deployed; orders for a further 40,000 to be placed separately |
| End of 2026 target | National installed base to approach 100,000 GPUs, per IndiaAI Mission CEO Abhishek Singh |
Why the Compute Push Matters
India ranks as the world’s third-largest AI ecosystem behind only the United States and China, but access to affordable, high-performance compute has remained a persistent bottleneck, particularly for early-stage startups and academic labs that cannot afford global cloud GPU rates. By subsidising access through the common compute pool, the IndiaAI Mission aims to prevent compute scarcity from throttling India’s AI ambitions just as demand for training large language models and other foundation models accelerates. The mission has also selected 12 startups to build indigenous multimodal foundation models trained on India-specific datasets, and it is coordinating with the National Supercomputing Mission on indigenous AI processors and accelerators built on the open-source RISC-V architecture, tying the compute buildout to a broader push for self-reliance in AI hardware.
Limitations and Risks in the Rollout
Ambitious targets carry execution risk. Doubling GPU capacity within six months depends on global chip supply chains, import timelines, and the readiness of empanelled data centres to physically install and commission new hardware — none of which is fully within the government’s control. There are also open questions about how allocation will be prioritised as demand grows beyond the initial 190 approved projects, whether the ₹65-per-hour subsidised rate is sustainable at scale, and how power and cooling infrastructure for expanded GPU clusters will be financed. Because NITI Aayog and MeitY operate on separate tracks — policy versus execution — coordination gaps between long-term AI governance goals and near-term infrastructure decisions remain a structural risk worth watching as the mission scales.
How India’s Compute Push Compares Globally
India’s approach stands out for relying on a subsidised, government-empanelled common compute pool rather than purely hyperscaler-led buildouts, though both are happening in parallel. Private players including Reliance, Adani, and independent data centre operators such as Yotta have separately announced gigawatt-scale data centre ambitions, while global hyperscalers like Microsoft, Google, and Amazon Web Services continue expanding their own India-based cloud regions. What makes the IndiaAI Mission distinct is its explicit focus on keeping access affordable for startups and academic researchers who cannot compete with well-funded corporations for scarce GPU capacity, rather than leaving compute allocation entirely to market pricing. That said, India’s total installed base, even at a targeted 100,000 GPUs by the end of 2026, remains a fraction of the compute clusters being built by leading US and Chinese AI labs, underscoring that the current push is about closing an access gap for India’s own ecosystem rather than competing head-to-head on raw global compute scale. Export restrictions on advanced chips from the United States also continue to shape which GPU generations are available to Indian data centres, adding a geopolitical dimension to how quickly the expansion can proceed.
Frequently Asked Questions
Is NITI Aayog directly running the GPU compute expansion?
No. NITI Aayog set the original policy strategy for AI in India starting in 2018, but the compute infrastructure, GPU empanelment, and subsidy programme are executed by MeitY through the IndiaAI Mission.
How many GPUs does India have under the IndiaAI Mission right now?
As of mid-2026, the mission has onboarded 38,231 GPUs through its common compute facility, with 20,000 more already ordered and a further 40,000 planned within the next six months.
Who can access the subsidised compute?
Startups, academic researchers, MSMEs, and students can access GPUs at roughly ₹65 per hour through approved projects under the IndiaAI Mission’s common compute facility.
What is India’s total GPU target for 2026?
Officials, including IndiaAI Mission CEO Abhishek Singh, have indicated a goal of approaching 100,000 GPUs installed nationally by the end of 2026, nearly three times the mid-2026 base.
Bottom Line
The 2026 compute push shows India moving from AI policy papers to physical infrastructure at real speed, but the two should not be conflated: NITI Aayog laid the strategic groundwork years ago, while MeitY’s IndiaAI Mission is the body actually approving projects and onboarding GPUs today. The jump from 38,000 to a targeted 100,000 GPUs within a single year is aggressive by any global standard, and its success will depend on supply chain execution as much as government intent. Startups and researchers relying on subsidised compute should track empanelment rounds and project approval cycles closely, since access — not just national capacity — will determine who actually benefits from the expansion.
Sources
Primary sources
Related Topic Express coverage
Featured image: Photo via Unsplash (photo-1558494949-ef010cbdcc31); free to use under the Unsplash License. Illustrative only.
