Vertical AI for India, and from India to the world
AI as substrate, not vertical. The moat is data, regulation, and frugal cost structure.
The 2024-era venture consensus held that horizontal AI infrastructure was the place to be. The sharper bet is vertical AI built in India, for verticals where Indian data, Indian regulatory positioning, or Indian cost structure is the unfair advantage.
I. The market shift
Scroll to walk through the three structural shifts. The chart on the right updates as each step comes into view.
AI moved from vertical to horizontal
Artificial intelligence has stopped being a category of company you can invest in and started being the substrate every company is built on. The implication for venture is not 'invest in AI' — that's the 2023 take, and it's already mispriced. The implication is that AI collapses the cost and time required to build software by an order of magnitude, which in turn changes who can credibly compete and where the durable margins live. Distribution, proprietary data, regulatory positioning, and physical-world integration matter more than ever, because the act of writing code matters less.
The sharper version: vertical AI in India, from India
We have rotated correctly toward vertical AI as a category. We have not rotated incompletely enough. The companies that compound for the next decade are the ones where the moat compounds with depth — proprietary domain data, regulatory standing, customer integration depth — and where the cost structure is built on Indian engineering rather than Silicon Valley. Healthcare-AI built on Indian clinical data. Legal-AI built on Indian case law. Mid-market accounting automation that's only economically viable when the AI plus a small Indian human-in-the-loop costs 10% of the Western alternative.
Why India specifically — the unfair-advantage stack
India produces approximately 1.5 million engineering graduates annually — more than the United States and China combined. The average annual cost of a software engineer in India is around $10,000 versus $110,000 in the US. UPI processed over 10 billion transactions in a single month in 2023 — proof that Indian customers adopt advanced digital infrastructure faster than Western markets. Stack these together and you get a structural cost-and-data moat that no US-only or China-only company can replicate.
II. Sub-themes within the pillar
Indian-data-native verticals
Healthcare, fintech, legal, and supply chain in India generate datasets that no Western incumbent can access. A clinical-AI company training on de-identified Indian hospital data, or a legal-AI company built on Indian case law, is not competing with OpenAI for the same prize — it is building a moat OpenAI cannot trespass on. The data is the durable asset; the AI is the engine that turns it into product.
Cost-arbitrage workflows
Workflows that exist globally but are too expensive to automate at Western unit economics — mid-market accounting, claims processing, contract review for SMBs, customer service in long-tail languages — become viable when the AI plus a small Indian human-in-the-loop costs 10% of the Western alternative. The buyer is global; the cost structure is Indian; the margin profile is unique to companies that can credibly run that combination.
Frugal-engineered AI infrastructure
Not the foundation models, but the picks-and-shovels for everyone deploying them — observability, model routing, fine-tuning ops, evaluation frameworks — built by engineers who grew up optimizing for constraint and who can serve a cost-conscious global market. The buyer is anyone running LLMs in production; the value is reducing their compute spend by orders of magnitude.
III. Where we see the wedge
Within this pillar, we look for three specific shapes of company. Each is a different way of expressing the same conviction.
Industries where Western consultants charge $200/hour for work an AI + Indian-human-in-the-loop system can do at $20/hour, with quality that approaches the senior consultant.
Healthcare, legal, and regulatory tech where the data is geo-specific and inaccessible to Western AI labs. Moats compound with every customer.
Existing vertical SaaS workflows — utility billing, sales call management, product analytics — where AI fundamentally changes the cost structure and unlocks markets that were previously priced out.
IV. Why we have the right to win in this pillar
Our angel portfolio includes Carestack — backed in 2013, built on Indian dental data, $145M+ subsequently raised. Our Fund I AI bets (Bynry, Runo, Orca AI, Skydda, Whatmore) are already operational evidence of what works. The Hurun India network gives us the canonical view of which Indian companies will produce the proprietary datasets the next decade of Vertical AI compounds on.
V. Companies in our portfolio expressing this pillar
The bets we have already made are the proof that the conviction is operational, not theoretical. 6 companies in Fund I currently sit in this pillar.
VI. Outlook
We expect 5–7 of Fund II's positions to land in this pillar — the largest single conviction. The companies will be split roughly evenly between Indian-data-native verticals and global-cost-arbitrage workflows, with one or two infrastructure plays.
The four pillars sit inside one thesis. Read the cornerstone essay, or see how Fund II operationalizes them.