India’s $31 Bn Pharma Industry Rarely Discovers Drugs. That Needs an AI Fix

India's pharmaceutical exports crossed $31 billion in FY26, a milestone that, to many, confirms the country's status as the world's pharmacy. India is the world's largest manufacturer, supplying roughly half of global generic drugs and a fifth of all exported medicines by volume. 

However, the drug discovery and development, the high-margin engine that funds pharmaceutical innovation, has remained largely a Western domain.

That distinction matters because contract manufacturing is a shrinking advantage. As automation lowers production costs globally and biosimilars add complexity, the margin gap between making drugs and inventing them is widening. The question facing Indian pharma is whether it can use AI to close it, and whether it has such ambitions in the first place.

Why the Gap Persists

For decades, Indian pharma companies rationally avoided novel drug discovery. A new molecule takes 10–15 years and over a billion dollars to bring to market, with a roughly 90% clinical failure rate, according to the Tufts Center for the Study of Drug Development. 

Generics offered a far more predictable business: reverse-engineer a molecule whose patent has expired, navigate regulatory filings efficiently, and compete on cost. Contract manufacturing organisations (CMOs) like Sun Pharma, Cipla, and Dr Reddy's built global franchises on exactly this model.

The model has started compressing. US generic drug prices fell by more than 50% in some therapeutic categories over the past decade as competition intensified. Meanwhile, the most lucrative pharmaceutical categories—biologics, oncology, and rare disease drugs—are structurally harder to manufacture generically. 

"The CMOs are efficiency engines. India perfected the art of manufacturing over the last several decades. But now the movement is toward innovation, discovering, and developing new drugs rather than only manufacturing existing ones," Duraisamy Palani, Founder and CEO of life sciences-focused IT company Archimedis Digital, tells AIM.

What AI is Actually Changing

The most immediate application is in clinical trial design. As per a Bain & Co report, the number of AI-discovered molecules has grown by more than 40% per annum between 2018 and 2021.

Researchers at Harvard's Multi-Regional Clinical Trials Center describe two emerging approaches. One is digital twins—simulated patient models designed to predict disease trajectories and treatment responses. Then there is synthetic data, using artificially generated datasets that mirror the statistical properties of real clinical data and are applicable at every stage from trial design through to regulatory submission.

In practice, this means companies can stress-test a trial design computationally before committing to expensive human studies, catching protocol gaps and likely failure points in the design phase rather than after years of data collection. A 2025 study presented at the ESMO Congress, involving over 19,000 patients with metastatic breast cancer, found that AI models could generate synthetic datasets highly faithful to real patient populations, with strong agreement in survival outcome analyses. Think of it as a compiler for clinical trials: errors caught in design, not after two years of data collection.

Crucially, this isn't a regulatory shortcut. The US Food and Drug Administration's 2025 draft guidance makes clear that AI tools used in drug development must be validated and transparent before influencing any submission. What synthetic data does is reduce the cost of reaching a trial that's actually worth running.

Beyond trial design, AI is being used to predict molecular structures, evaluate toxicity profiles, and narrow the field of candidate compounds before reaching labs. Drug discovery has always been a filtering problem—eliminating bad candidates as cheaply as possible. The 2024 Nobel Prize in Chemistry, awarded for AlphaFold's breakthrough in protein structure prediction, put a formal stamp on how fundamentally AI is reshaping this early filtering stage.

“It is thus imperative for Indian companies to either invest in building in-house AI capabilities or collaborate with global technology vendors to stay competitive. Indian startups are increasingly leveraging AI-based technologies for advanced drug discovery; however, there is tremendous potential to scale this up,” Bain & Co noted in its report.​​

Inside the Enterprise Shift

Biocon, one of India's largest biopharmaceutical companies, offers a useful window into how the transition is being structured internally. Mandar Ghatnekar, the company's CTO, describes two distinct tracks of AI adoption.

"The first is productivity-led AI, where automation and real-time analytics are already driving improvements in efficiency and resilience across functions such as regulatory, finance, procurement, supply chain, and manufacturing at scale," he says.

The second track is harder to build and more consequential: embedding AI directly into scientific decision-making. "Several of these capabilities are currently being piloted as proof of concepts within our R&D programmes," Ghatnekar remarks, describing AI systems that "connect fragmented data and knowledge, enabling scientists to identify patterns and insights that can accelerate development timelines."

Biocon is also establishing what it calls an "Office of AI (Augmented Intelligence)"—a governance structure designed to scale these capabilities responsibly across the enterprise. The creation of dedicated AI governance is significant: it signals that pilots are maturing beyond experiments into infrastructure.

The Domain Problem

Then there’s the challenge of regulatory traceability. A general-purpose model that generates a plausible-sounding clinical insight is not useful if it can't explain, step-by-step, how it got there. Regulators require that every inference in a submission trail be auditable, applying the “if you can’t document, it didn’t happen” standard.

This is pushing life sciences companies toward smaller, domain-specific models trained on scientific and regulatory datasets—models that trade breadth for accountability. Isomorphic Labs, a subsidiary of Alphabet launched from DeepMind, utilises a highly specialised, AI-first technology stack built on Google Cloud Platform to reimagine drug discovery. Meanwhile, Insilico Medicine launched Science MMAI Gym, a specialised training model designed to train LLMs for vetting molecular design.

It's a meaningful architectural divergence from the foundation model approach, which requires meeting strict regulatory standards. But it has also given rise to pharma-focused AI firms and large model providers licensing into the sector.

Google Cloud’s MedLM is fine-tuned for AI-driven diagnostics in healthcare and life sciences. Meanwhile, NVIDIA’s BioNeMo drug discovery platform enables researchers to design molecules and predict protein structures.

AI adoption across Indian pharma is also uneven. Larger companies like Biocon and Sun Pharma have the capital and talent pipeline to run serious AI programmes. Most mid-tier and smaller manufacturers don't—and they represent the bulk of India's pharmaceutical output.

Government policy is beginning to move in the right direction. The Indian Drug Manufacturers' Association has pushed for investment in digital infrastructure. Recent policy signals like the proposed BioPharma SHAKTI Initiative, with a proposed outlay of ₹10,000 crore, specifically target biopharmaceutical research, biosimilars, and AI-powered drug discovery. 

But infrastructure support and regulatory clarity around AI-generated clinical data will be necessary if the sector's adoption is to go beyond a few headline companies.