India AI Stack: Powering Intelligence at Scale

1. At a Glance

• India AI Stack = an integrated national architecture stitching together compute, data, models, applications and safety layers to deliver AI at population scale, framed under the doctrine of "democratisation of AI / AI for Humanity" [1][2].
• Operationalised via the IndiaAI Mission under MeitY, with a Cabinet-approved outlay of ₹10,371.92 crore (March 2024) over 5 years [3].
• UPSC relevance: intersects GS-III (S&T, digital economy), GS-II (governance, e-services), ethics (GS-IV) and current affairs (AI Impact Summit 2026).

2. Why in the News

• PIB Backgrounder dated 4 February 2026 released as "India AI Stack: Powering Intelligence at Scale" outlining the layered stack [1][2].
• India AI Impact Summit 2026 (Feb 2026, New Delhi) issued a landmark global declaration with major AI investment commitments; VoicERA launched on BHASHINI infrastructure at the summit [4][5].
• Mission crossed 34,000 GPUs common compute capacity (mid-2025) and later 38,000 GPUs, far exceeding the original 10,000-GPU target [6][7].

3. Background & Evolution

• 2018: NITI Aayog's National Strategy for Artificial Intelligence ("AI for All").
• March 2024: Union Cabinet approves IndiaAI Mission, ₹10,372 cr, 5-year horizon, implemented by IndiaAI, an Independent Business Division (IBD) under Digital India Corporation, MeitY [3].
• March 2025: AIKosh (datasets/models platform) launched by MeitY [8].
• Feb 2026: India AI Impact Summit; India AI Stack backgrounder; sovereign foundation-model developers announced [1][4][9].

4. Core Static Facts

• Nodal Ministry: Ministry of Electronics & Information Technology (MeitY); implementing arm — IndiaAI IBD under Digital India Corporation [3].
• Outlay: ₹10,371.92 crore / 5 years (FY 2024-25 onwards) [3].

• Seven Pillars of IndiaAI Mission [3][7]: 1. IndiaAI Compute Capacity 2. IndiaAI Innovation Centre (Foundation Models) 3. IndiaAI Datasets Platform (AIKosh) 4. IndiaAI Application Development Initiative 5. IndiaAI FutureSkills 6. IndiaAI Startup Financing 7. Safe & Trusted AI

• AIKosh: 7,500+ datasets, 273 AI models, 20 industries [2].

• Bhashini: 350+ AI models for ASR, MT, TTS, OCR, language ID under National Language Translation Mission [2].
• Sovereign LLM developers selected: Sarvam AI, Soket AI, Gnani AI, Gan AI, Avatar AI, IIT Bombay-led BharatGen consortium, GenLoop, Zentieq, Intellihealth, Shodh AI, Fractal Analytics, Tech Mahindra Maker's Lab [2][9].

5. Multi-Dimensional Analysis

Scientific / Technological

• Stack layers: compute (GPUs) → data (AIKosh) → models (sovereign LLMs, Bhashini) → applications → safety [1][2].
• Compute is offered at subsidised, affordable rates through empanelled providers [6][7].

Economic

• Aims to catalyse AI startups via IndiaAI Startup Financing pillar and innovation grants [3].
• Application Development targets healthcare, agriculture, education, governance, manufacturing [1].

Social / Governance

• Democratisation of AI → multilingual access via Bhashini, lowering entry barriers for SMEs and academia [1][2].
• AIKosh sandbox + IDE allows non-elite developers and States to build solutions [8].

Ethical / Safety

• Safe & Trusted AI pillar → AI safety institute, deepfake detection, bias audits, watermarking tools [3].
• Aligns with India's Bletchley/Seoul/Paris AI safety commitments culminating in India AI Impact Summit 2026 Declaration [4].

Geopolitical / Strategic

• Push for sovereign foundational models trained on Indian data reduces dependence on Western/Chinese LLMs [9].
• India positioned as a Global South AI hub following Summit declaration [4].

6. Recent Developments (12-18 months)

• March 2024: Cabinet approval of IndiaAI Mission, ₹10,372 cr [3].
• March 2025: AIKosh platform launched [8].
• Mid-2025: Common compute crosses 34,000 GPUs [6].
• 2025-26: 24-month review: ecosystem foundation set; 38,000 GPUs operational [7].
• Feb 2026: India AI Impact Summit 2026 — global declaration & investment commitments [4].
• Feb 2026: VoicERA voice-AI platform launched on Bhashini [5].
• Feb 2026: Sovereign foundation model consortium expanded (BharatGen etc.) [9].

7. Prelims Hooks

• IndiaAI Mission approved by Union Cabinet in March 2024 with outlay ₹10,371.92 crore / 5 years [3].
• Nodal ministry: MeitY (not DST, not NITI Aayog); implementing arm: IndiaAI IBD under Digital India Corporation [3].
• IndiaAI Mission has 7 pillars including Compute, Innovation Centre, Datasets, Applications, FutureSkills, Startup Financing, Safe & Trusted AI [3].
• AIKosh = IndiaAI Datasets Platform launched March 2025; hosts 7,500+ datasets and 273 models [2][8].
• Bhashini sits under the National Language Translation Mission; 350+ AI models [2].
• Common compute capacity crossed 34,000 GPUs (2025) and later 38,000 GPUs, against initial target of 10,000 [6][7].
• BharatGen sovereign foundational model is led by an IIT Bombay-led consortium [2].
• India AI Impact Summit 2026 held in New Delhi (Feb 2026) — third in the global AI summit series after Bletchley & Seoul/Paris [4].
• VoicERA, launched at the Summit, rides on BHASHINI infrastructure [5].
• AI Stack layers: Compute → Data → Model → Application → Safety [1][2].

8. Mains Relevance

• GS-II: e-Governance, welfare delivery via AI-enabled public services.
• GS-III: Science & Technology — indigenisation of emerging tech; digital economy; cyber/AI security.
• GS-IV: Ethics — algorithmic bias, AI accountability.
• Probable stems: 1. "Discuss how the India AI Stack operationalises the principle of 'democratisation of AI'. What governance challenges remain?" (GS-III). 2. "Sovereign foundation models are as much a strategic imperative as a technological one. Examine in the Indian context." (GS-III). 3. "Evaluate the institutional architecture of the IndiaAI Mission with reference to its seven pillars." (GS-II/III).

9. Related Topics to Study Next

• National Strategy for AI, 2018 (NITI Aayog) — policy precursor.
• Digital Public Infrastructure (DPI) / India Stack — Aadhaar, UPI, DigiLocker as the substrate India AI Stack rides on.
• Bhashini & National Language Translation Mission — language-AI layer.
• Semicon India Programme — hardware leg complementing GPU compute.
• DPDP Act, 2023 — data protection regime governing AI training data.
• Global AI Governance — Bletchley (2023), Seoul (2024), Paris (2025), Delhi (2026) summits.
• National Quantum Mission — adjacent deep-tech mission.
• AI for Bharat / BharatGen / Sarvam AI — sovereign model ecosystem.

10. Common Errors / Trap Areas

• IndiaAI Mission is under MeitY, not NITI Aayog (NITI authored the 2018 strategy, not the mission).
• Implementing arm is IndiaAI IBD under Digital India Corporation, not a standalone statutory body — no parliamentary Act backs it.
• AIKosh ≠ Bhashini: AIKosh is the datasets+models repository; Bhashini is the language-AI platform.
• Outlay is ₹10,372 crore over 5 years, not annual.
• BharatGen is led by an IIT Bombay consortium, not by IIT Madras or IISc.
• The 10,000-GPU figure was the original target; actual deployment now exceeds 34,000–38,000 GPUs.