India's AI Stack For Population-Scale Impact

India's ₹10,300 Crore Push for AI Compute: Ambition Meets Infrastructure Challenges

On February 14, 2026, the India AI Impact Summit in New Delhi spotlighted India’s ambitious AI deployment strategy, emphasizing practical applications in agriculture, health, education, and environmental forecasting. At the heart of this vision lies a ₹10,300 crore allocation for AI compute—promising shared access to 38,000 GPUs and 1,050 TPUs—alongside a ₹76,000 crore Semiconductor Mission for indigenous processors. These numbers tell a promising story, but beneath the surface lies a set of uncomfortably complex questions about scale, equity, and institutional capacity.

Breaking Free of Abstract AI Policies

India’s approach to AI, reflected in this Summit and the Economic Survey 2026, shows a clear departure from the overgeneralized rhetoric dominating earlier debates. The shift toward prioritizing human welfare, economic inclusion, and context-specific innovation is important. Tools like PadhaiWithAI for personalized school math coaching and Qure.ai for resource-poor healthcare diagnostics illustrate AI being deployed not as a luxury, but as a necessity targeting fundamental social outcomes. This sets India apart from countries like China, where AI investments disproportionately favor surveillance technologies and export-driven industries.

The focus on Agriculture is particularly significant. Neoperk’s soil health monitoring, CottonAce’s pest detection, and Niqo Robotics’ real-time weed control represent scalable solutions for farming crises commonly worsened by climate volatility. This is not just theoretical. In Tamil Nadu’s Cauvery delta, specific AI tools have already reduced pesticide use by nearly 22% while boosting yields by 19%—a concrete proof point underscoring the utility of India-specific models.

The Machinery Behind India's AI Stack

Much of this ambition is housed within coherent institutional frameworks. The IndiaAI Mission supports 12 domestic AI models with a 25% cost offset for computational expenses, while BharatGen builds highly contextualized foundation models tuned explicitly for local needs. The Bhashini initiative hosts 350+ models for speech recognition and translation, essential in a multilingual democracy where linguistic inequities persist.

Underpinning these efforts is the National Supercomputing Mission, which has already deployed over 40 petaflops through institutions such as IITs and IISERs, bolstered by systems like PARAM Siddhi-AI and AIRAWAT. On the physical infrastructure side, India’s 3% share of global data centre capacity is projected to jump to 9.2 GW by 2030, supported by rapidly expanding optical fiber and 5G penetration covering 85% of the population.

Legislatively, the SHANTI Act—India’s flagship framework on nuclear power—directly targets low-carbon energy solutions essential for compute-intensive AI applications. It underscores the interconnected nature of AI scale and energy sustainability, with peak demand already at 242.49 GW and negligible shortages (0.03%) as of FY 2025–26.

The Data Versus the Reality

The Summit's numbers are impressive, but the data beneath them reveals mounting inequities. Despite advances like modular AI models through BharatGen, many state governments struggle with basic infrastructure gaps. For instance, less economically advanced regions such as Bihar and Jharkhand have adoption rates for AI health diagnostics below 9%, compared to over 70% in Karnataka and Tamil Nadu.

Similarly, while tools like Rocket Learning’s AI literacy programs have boosted primary school outcomes in several districts, state capacity to integrate these services remains uneven. In Uttar Pradesh, DIKSHA platform utilization rates lag significantly behind states like Kerala, raising questions about regional AI-readiness despite the government’s insistence on inclusivity.

Moreover, while the government’s push for indigenous processors (SHAKTI, VEGA) is laudable, India remains critically behind fabrication-heavy countries like Taiwan. The ₹76,000 crore Semiconductor Mission is ambitious, but projections suggest only partial substitution of imports by 2030, far short of self-sufficiency. This exposes a strategic vulnerability precisely at a time when geopolitical supply chain disruptions grow more common.

The Questions Nobody Wants to Ask

Skepticism remains, especially concerning implementation capacity. Can India sustain such compute-heavy AI ambitions given its significant regional and institutional disparities? The optical fiber network rollout is commendable, but poorer districts in remote regions still languish without reliable energy or internet—conditions wholly incompatible with advanced AI deployment.

Additionally, regulatory clarity is lacking. For instance, the IndiaAIKosh repository claims to host 5,722 datasets, but its access mechanisms remain centralized, raising concerns about equitable sharing. Many start-ups and local institutions cannot afford advanced compute, resulting in an increasingly narrow distribution of AI resources despite claims of democratic inclusion.

The broader issue of AI governance also looms large. While frameworks like the SHANTI Act address compute sustainability, cyber security and misuse risks from generative AI models (e.g., Sarvam Vision outperforming global names like ChatGPT) remain inadequately addressed. India lacks comprehensive AI legislation akin to the EU's AI Act, leaving several critical gaps for ethical deployment.

What India Can Learn from South Korea

South Korea offers a carefully targeted comparison. In 2018, Korea introduced context-focused AI implementation through its Smart Farming initiative, which deployed AI-guided irrigation and pest control localized to individual farming villages. Similar to India's CottonAce, predictive algorithms were tailored to microclimates rather than relying on generic datasets. However, South Korea paired this with significant local capacity-building programs—installing user training hubs directly in rural areas rather than relying entirely on tech companies for dispersed rollouts. India’s AI stack remains heavily centralized, undermining such community-specific scaling. The inherent lesson lies in placing capacity-building alongside technological rollout, not behind it.

Exam Integration

Practice Questions for UPSC

Prelims Practice Questions

Frequently Asked Questions