The Future of India’s Healthcare Market

Healthcare is the only sector where the customer does not want to be a customer. Nobody wakes up excited to buy an MRI. That single fact explains most of what is broken about Indian healthcare, and it also explains why the most valuable thing this country owns has been sitting unexploited for twenty years.

I spent two years building in digital health, first through a masters in MedTech Innovation and then as a founder trying to turn skin health into something a person could actually measure and track. I now spend my days as a venture scout and early-stage operator looking at preventative healthcare and AI in health. What that period taught me had almost nothing to do with technology. It was about plumbing. Who pays. Who refers. Who trusts whom. A beautiful product with no payer and no clinician behind it is a science project.

But there was a second thing, and it took me longer to see. Every time we tried to validate anything, the answer came back different depending on where the user was. Water hardness changed what happened to skin. Cooking oil changed lipid panels. Humidity changed everything. Altitude, staple grain, pollution load, endogamy, language of instruction. Move 100 kilometres in India and the inputs to a human body change materially. Move 500 and you are effectively in a different country.

That is not a footnote. I think it is the whole thesis.

‍

The numbers, briefly

India’s healthcare market is routinely cited at around $400 billion in FY26. That figure is real but not useful, because roughly 35 to 40 percent of it sits with informal, unorganised providers who will never appear on a cap table. The organised, investable market, meaning hospitals, diagnostics, pharma, devices, and health insurance, is closer to $220 to $250 billion, growing 8 to 12 percent annually.

The supply side is thin. India has roughly 1.3 to 1.9 million hospital beds, about 1.3 to 1.5 physicians per 1,000 people including AYUSH practitioners, and 2.0 to 2.3 nurses per 1,000. The capital is arriving anyway. Private equity and venture capital deploy $2 to $3 billion a year into Indian healthcare, and hospitals are absorbing the largest share. Manipal filed a DRHP in March 2026 for an offering of up to roughly $1.17 billion. Global funds are bidding for a 25 percent stake in Cloudnine at around ₹10,000 crore. Quality hospital assets are transacting at high-teens EV/EBITDA.

Meanwhile the public rails are further along than most investors realise. Over 78 crore digital health IDs have been generated under the Ayushman Bharat Digital Mission as of February 2026, and eSanjeevani has handled roughly 37.2 crore teleconsultations. GenomeIndia sequenced 10,000 whole genomes across 83 communities, and initial analysis surfaced large numbers of variants that are rare or absent in global variant databases.

Stack these facts next to each other and the pattern is uncomfortable. Identity is solved. Consent architecture exists. Genomic reference data exists. Capital is abundant. And almost nobody is connecting any of it to what actually happens to patients over time.

‍

Why this isn’t another India-data essay

The standard version of this argument goes: India has 1.4 billion people, therefore India has a data goldmine, therefore build AI. People have been writing that sentence since 2012 and the mine is still unexplored. There is a reason, and it is not incompetence.

Scale was never the asset. China has scale. The United States has better instrumented scale. If population size were the input, this would have been solved by someone with more capital than us a decade ago.

The asset is variance. India is not one population, it is several thousand endogamous ones layered on top of a dozen climatic zones, a dozen dietary regimes, and one of the widest income spreads on earth. That combination does not exist anywhere else at this resolution. And variance is precisely the input that every serious clinical model, drug label, and risk algorithm in the world is currently starved of.

The thesis of this piece is that India’s healthcare opportunity is not building better care delivery for Indians, though that will happen. It is that India owns the only population diverse enough to break Western-cohort models, it is currently exporting that asset for free through academic collaboration and importing the models built on it, and the companies that build the collection loop to capture it will end up licensing to the rest of the world rather than selling to Indians.

That is the trade. Everything below is how you actually run it.

‍

The 100 kilometre problem

Here is the mechanism, stated plainly.

Almost every clinical threshold used in an Indian hospital today was calibrated on a population that does not look like the patient in front of the doctor. The BMI cutoff for metabolic risk had to be revised downward for Asian populations because Indians accumulate visceral fat and develop insulin resistance at body weights that would be considered unremarkable in Europe. Cardiovascular risk calculators built on Western cohorts systematically misprice Indian patients. Drug dosing, dermatological baselines, nutritional reference ranges, growth charts, all of it carries an imported prior.

Now add the internal variation. A patient in coastal Kerala, a patient in Punjab, and a patient in the Northeast differ in staple grain, cooking medium, salt load, water mineral content, sun exposure, ambient humidity, air quality, alcohol and tobacco patterns, and genetic substructure. The same intervention produces different outcomes, and nobody can currently say by how much, because the study that would tell you has never been run at that granularity.

This is the sentence I keep coming back to. A model trained in Bengaluru will perform differently in Bhagalpur, and right now nobody can prove it, disprove it, or price it. That is not a gap in the literature. That is an entire asset class that has never been assembled.

GenomeIndia is the cleanest illustration. India built the reference layer, sequenced 83 communities, found variants the world had never catalogued, and the project’s own documentation notes that linking that genomic data to clinical and health outcomes remains future work. We have the map with no territory attached to it.

‍

The stack: collection loop, preventive layer, licence

The opportunity is a sequence, and the order is not negotiable. Run it backwards and you end up with a dataset nobody validated and a platform nobody wants.

Layer one: the collection loop. The loop cannot be funded by the data itself. Nobody pays for collection on day one. It has to sit on top of a transaction the customer was already making, so that the data is a byproduct and the byproduct is the asset. A diagnostic panel. A pharmacy refill. An insurance renewal. A corporate health check. A wearable someone bought for sleep scores.

The uncomfortable truth is that the best-positioned players in India today did not choose this. The organised diagnostics chains are sitting on a decade of samples, results, and pincodes, and almost none of it is linked longitudinally to what happened to the patient afterwards. They have the loop and have never closed it.

Two design decisions matter more than anything else, and neither is retrofittable. First, collect to a licensable standard from the first sample: known provenance, clean consent under DPDP, standardised assay method, and above all linked outcomes over time. A million uncontrolled records license for nothing. Fifty thousand records with clean consent and five years of follow-up license to pharma, to device makers, and to regulators. Most Indian health data plays die on quality, not on scale. Second, tag geography as a first-class variable rather than a shipping address. If you cannot stratify by district, you have not built the asset described in this piece.

Layer two: the preventive layer. This is where the data starts paying for itself, and it is also where founders consistently misprice the customer.

Nobody pays for the heart attack that did not happen. The consumer will not, sustainably, and the hospital has no reason to. The only entities structurally motivated to fund prevention are the ones carrying the risk: insurers with a loss ratio, employers with a benefits bill, and the state with a scheme liability. Prevention in India is not a consumer product. It is an underwriting product that happens to reach a consumer.

Which makes the payer shift the most important thing that happened in this sector in the last eighteen months and the one nobody wrote about. The 56th GST Council meeting exempted individual health insurance policies, including family floater and senior citizen plans, from the 18% levy with effect from 22 September 2025. Health premiums grew 27.17% year on year in January 2026, with retail health up 27% against group health at 10% . Retail health growing at nearly three times group is exactly the cohort you need: individually underwritten, individually incentivised, and reachable.

Every point of insurance penetration creates a buyer for risk stratification that did not exist the year before.

Layer three: the licence. This is a year seven outcome, and any piece that pretends otherwise is selling something. But it is the reason the first two layers are worth building at a standard higher than the domestic market currently demands.

The realistic year-three revenue is enterprise. Risk stratification sold to insurers. Cohort access and site identification sold to pharma for trials, where India’s diversity is a genuine competitive input rather than a talking point. Calibration and validation data sold to device and AI companies who need Indian evidence for CDSCO approval and, increasingly, need non-Western validation to make any generalisability claim at all.

The year seven version is that Indian platforms license outward. Qure.ai is the existence proof that this is not fantasy: built on Indian scans, selling globally, valuable precisely because the training distribution was not American.

‍

Where the loop can actually attach

Read the current market map as a list of candidate hosts for the collection loop rather than as a set of independent segments.

Diagnostics is the strongest attachment point in the country and the most underrated. Organised chains run 25 to 35 percent EBITDA margins on an asset-light base, already touch the patient at the moment of measurement, and investment focus within the segment has shifted decisively toward genomics, oncology, and molecular testing. They have the sample, the result, and the location. They lack the follow-up.

Hospitals own the outcome, which is the scarce half of the equation, but they own it episodically. High-acuity specialties such as oncology, cardiology, and neurology are where both the clinical value and the deal activity are concentrating. A hospital that systematically linked its outcomes back to a diagnostics loop would be building something nobody in India currently has.

Insurance and the payer stack is where the loop gets monetised, per the argument above.

Medtech is the sharpest illustration of the cost of not owning your own data. India’s device market was around $15.2 billion in 2025 and is credibly projected to reach roughly $50 billion by 2030, yet 70 to 80 percent of requirement is still imported, the FY26 import bill touched nearly ₹89,000 crore, up 17 percent in a year, and Indian manufacturers spend under 1 percent of revenue on R&D against 6 to 8 percent globally. We import the device and we import the reference ranges baked into it.

Digital health raised about $7.25 billion between 2014 and 2024 and roughly $310 million in H1 2026. The correction was correct. The money still moving is going to defensible IP, regulatory pathways, and measurable outcomes, which is precisely the profile of the thesis described here.

‍

What I would fund, and what I would avoid

Fund: businesses where a high-frequency existing transaction generates clinical data as a byproduct, and where the founder designed for licensability before scale. Preventive products sold to a risk-carrying payer rather than to a consumer. Anything that closes the loop between a measurement and an outcome twelve months later. Doctor-founder DNA over D2C DNA. Geography-native design: vernacular, district-stratified, built for Tier 2 and Tier 3 from the first version rather than ported down from a metro product.

Avoid: data plays with no underlying transaction, because collection funded by a fundraise stops when the fundraise stops. Prevention sold direct to consumers on willpower. Volume without provenance. Anything whose moat is described as “we have a large dataset” without a sentence explaining who is contractually allowed to use it and for what. And the Indian-market-size pitch that treats 1.4 billion people as the asset, because scale was never the thing we uniquely own.

What India 2030 looks like

1. The payer, not the patient, becomes the primary customer for most of Indian healthtech.
2. Diagnostics chains realise the sample archive was the business and the test was the customer acquisition cost.
3. At least one Indian company licenses a clinical model or dataset internationally at a valuation nobody in this market currently believes is possible.
4. Regional stratification becomes a standard clinical expectation rather than an academic curiosity, and imported reference ranges start being challenged in Indian practice.
5. Insurance integration determines who owns the mass market, exactly as it does in wellness.
6. The most valuable healthcare companies own longitudinal relationships, not procedures.

India has spent two decades trying to be a cheaper version of a Western healthcare system. Cheaper drugs, cheaper devices, cheaper surgery, cheaper software. It worked, and it capped out.

The thing we actually own is not cost and it was never scale. It is the fact that this country contains more human variation per square kilometre than any comparable landmass on earth, that the variation is medically consequential, and that nobody, including us, has bothered to write it down.

We are sitting on the mine. The question is not whether it is there. The question is whether the people who build the collection loop over the next five years will be Indian companies, or whether we will do what we did with genomic data and clinical trial sites, which is hand over the input for free and buy back the output at a licence fee.

That is the India I would like to see us build. One that stops apologising for its complexity and starts charging for it.