Examine how the IndiaAI Mission's seven-pillar architecture seeks to build sovereign AI capacity. Illustrate with recent indigenous models.

Q. Examine how the IndiaAI Mission's seven-pillar architecture seeks to build sovereign AI capacity. Illustrate with recent indigenous models. (15 marks, 250-350 words)

Sovereign AI capacity means the ability to build, own and govern AI systems on domestic compute, data and talent. The IndiaAI Mission, approved by the Union Cabinet in March 2024 with an outlay of ₹10,371.92 crore over five years under MeitY [1], pursues this through seven interlocking pillars rather than a single flagship model.

Compute sovereignty — the hardware layer - The IndiaAI Compute Capacity pillar aggregates GPUs in PPP mode; the common pool crossed 34,000 GPUs in 2025 [2] and 38,000+ GPUs, offered to startups and academia at a subsidised ₹65/hour [3]. - This removes the single biggest entry barrier — foreign cloud dependence for training frontier models.

Model and data sovereignty - The Innovation Centre (foundation models) pillar supports 12 companies building indigenous multimodal models on Indian languages and data [3]. - The Datasets Platform pillar streamlines access to quality non-personal datasets, so models reflect Indian contexts rather than imported ones.

Diffusion — skills, startups, applications - FutureSkills funds PhD, PG and UG fellowships; Startup Financing backs early-stage deep-tech ventures; Application Development targets health, agriculture and governance use cases [3].

Guardrails - The Safe & Trusted AI pillar funds work on bias mitigation, explainability and auditing — critical as generative video raises deepfake risks [3].

Illustration: Varya, launched on 12 June 2026 in New Delhi by Avataar with MeitY's support, is India's first distilled video-generation model. Using distillation it compresses inference from 50 steps to 4, generating video at ₹0.48 per second — up to 10× cheaper than leading global models — for educators, MSMEs and content creators [4]. It shows the pillars working together: public compute plus startup financing yielding a privately built, indigenous frontier model.

The architecture is therefore an ecosystem strategy, not a subsidy scheme — sovereignty emerging from compute, data, talent and trust simultaneously. Sustaining it will require deepening domestic semiconductor capacity and operationalising the Safe & Trusted pillar, so that affordability and accountability advance together towards a genuinely self-reliant, inclusive AI ecosystem.

(~325 words)

Sources: 1. Cabinet Approves Over Rs 10,300 Crore for IndiaAI Mission (PIB, March 2024) — outlay of ₹10,371.92 crore, five years, MeitY 2. India's Common Compute Capacity Crosses 34,000 GPUs (PIB) — 2025 GPU pool figure 3. India Accelerates AI Self-Reliance: 12 Companies Developing Foundation Models Using 38,000 GPUs at ₹65/Hour (PIB) — 38,000+ GPUs, ₹65/hour, 12 foundation-model teams, pillar-wise details (FutureSkills, Startup Financing, Safe & Trusted AI, Datasets Platform) 4. India AI Mission launches Indigenous Varya, a Video Story Generating AI Model (PIB, 12 June 2026) — Varya's launch, developer Avataar, 50→4 steps, ₹0.48/second, 10× cost efficiency, user groups