The Next Decade of AI in India: What I'm Betting On

The Starting Point

India's AI story in 2025 is early innings. The large language model wave of 2022–2024 was primarily a US phenomenon, with Indian adoption concentrated in enterprise use cases and English-dominant products. The next phase — AI that works for India's actual population, in India's actual languages, on India's actual hardware — is just beginning.

These are the bets I'm making on where that next phase goes, based on what I'm seeing from the ground in Ahmedabad.

Bet 1: Vernacular AI Becomes the Main Event

Today's successful AI products in India are almost all English-primary. This describes a subset of the market — educated, urban, relatively affluent. The opportunity in the next decade is AI that works natively for Hindi, Tamil, Telugu, Marathi, Gujarati, and the other major Indian languages.

The infrastructure is improving rapidly. Indic-BERT, IndicTrans, and the multilingual models from IITM and AI4Bharat are producing language capabilities that weren't available 3 years ago. The gap between English AI quality and Indic language AI quality is closing — not closed, but closing.

My bet: the category winners in Indian consumer AI over the next decade will be vernacular-native, not English-first-with-translation. The businesses that build the multilingual infrastructure now will have a compounding advantage when the vernacular user base scales.

Bet 2: Healthcare AI Has the Largest Impact

India has 1 doctor per 1,457 people, compared to the WHO recommendation of 1 per 1,000. This gap is not closable with more medical schools in any reasonable timeframe — but it might be addressable with AI-augmented healthcare that makes existing providers significantly more effective.

The near-term opportunities: AI-assisted diagnosis for common conditions (TB, diabetic retinopathy, anemia) where AI diagnosis has achieved near-specialist accuracy and can be deployed at the point of care; AI for medical record transcription and coding (a persistent bottleneck in Indian healthcare administration); and drug adherence monitoring for chronic disease management.

This sector has regulatory complexity, but the regulatory environment is evolving. CDSCO is actively developing AI/ML device guidance. The companies that engage with the regulatory process now will have cleared those hurdles when market timing is right.

Bet 3: Agriculture Gets Smarter Than Anyone Expects

India is the world's second-largest agricultural producer. The sector employs approximately 42% of the workforce. AI for agriculture — crop disease detection, yield prediction, soil health monitoring, weather-adapted planting advice — has genuine product-market fit in ways that Western markets, with their consolidated large-farm agriculture, don't fully appreciate.

The deployment model that works in India: phone-based (not web-based, since rural connectivity often supports voice calls better than web browsing), vernacular-language, and advice-oriented rather than data-visualization-oriented. A farmer who receives an audio message in Gujarati saying "your field in sector 3 shows signs of early blight — apply copper fungicide within 48 hours" acts on that information. A map interface showing NDVI values doesn't.

Bet 4: Edge AI Matters More Here Than Anywhere

India's connectivity profile — excellent in urban cores, variable in Tier-2 cities, poor in rural areas — makes edge AI more valuable here than in markets with uniformly good connectivity. Models that run on-device, inference that works offline, AI that doesn't require a cloud round trip — these aren't just technical differentiators in India. They're product requirements for serving the majority of the population.

The companies and engineers who develop expertise in edge AI deployment — model compression, TensorRT optimization, on-device inference — are positioning for a market where this expertise is worth more than in highly-connected markets.

What I'm Building Toward

At Sunbots and through Xwits Developers, the bets we're making are consistent with the above: deepen multilingual capabilities across SmartON and the management platform, develop healthcare-adjacent AI for assistive technology (SmartON's user base overlaps significantly with accessible healthcare use cases), and continue investing in edge AI architecture that works in India's actual connectivity environment.

The next ten years in Indian AI are more interesting than the last five. The infrastructure is better, the talent pool is deeper, and the problem space is clearer. I'm optimistic — not because AI is magic, but because the hard foundational work is far enough along that the application layer can start to scale.