FOR MOST OF PHARMACEUTICAL HISTORY, drug discovery has been a process of educated guessing at enormous scale. The traditional approach — high-throughput screening, in which automated systems test hundreds of thousands of existing compounds against a biological target and identify those that show activity — is fundamentally a search through known chemical space, filtered by biological intuition built from decades of medicinal chemistry experience. It is costly ($2.6 billion on average from discovery to approval, by 2020 estimates), slow (10 to 15 years from initial target identification to patient access), and prone to failure at every stage (approximately 90 percent of drug candidates that enter clinical trials fail to reach approval). Artificial intelligence promised to change this calculus. After years of bold predictions and modest reality, 2025 delivered the first genuine clinical validation that AI-designed drugs can work in humans — and 2026 is shaping the next chapter.
The AI drug discovery market was valued at approximately $1.94 billion in 2025 and is projected to reach $2.6 billion in 2026, growing toward $16.49 billion by 2034 at a compound annual growth rate of 27 percent. More than 200 AI-designed drug candidates are now in clinical development globally, with 15 to 20 entering pivotal Phase III trials in 2026. These numbers represent a qualitative shift: AI is no longer a laboratory curiosity or a back-office analytical tool. It is the primary engine of discovery at a growing number of pharmaceutical organisations.
The 2025 Breakthrough: First-in-Human Validation
The defining event of 2025 in AI drug discovery was the publication of positive Phase IIa results for ISM001-055, a TRAF2- and Nck-interacting kinase (TNIK) inhibitor designed entirely by Insilico Medicine’s Chemistry42 platform for the treatment of idiopathic pulmonary fibrosis (IPF). IPF is a devastating progressive lung disease with limited treatment options; the positive Phase IIa data — demonstrating both safety and early efficacy signals — marked the first time an AI-designed molecule had shown clinical benefit in a controlled human trial. This is the proof-of-concept moment the field has been waiting for since the first AI drug discovery programmes launched in the mid-2010s.
The same year saw AstraZeneca report that over 90 percent of its small molecule discovery pipeline is now AI-assisted, with significant improvements in lead optimisation accuracy. The Recursion-Exscientia merger created an end-to-end discovery platform integrating phenomic screening with automated precision chemistry. Insilico also demonstrated AI’s capability beyond simple small molecules, using Chemistry42 to design a first-in-class PROTAC (Proteolysis-Targeting Chimera) with a dual-action cancer mechanism targeting the challenging oncology target PKMYT1. 2025 also saw the highest single-year jump in Investigational New Drug filings for AI-originated molecules, driven by companies including Insilico, Recursion, BenevolentAI, Absci, and Generate Biomedicines.
The Computational Revolution: New Molecular Tools
Underlying the clinical progress is a rapid evolution in the computational tools available for molecular design. Boltz-2, an open-source biomolecular foundation model developed by MIT and Recursion, delivers near-physics-level accuracy for binding affinity predictions at speeds up to 1,000 times faster than traditional free-energy perturbation simulations — making large-scale virtual screening of vast chemical spaces practical for early-stage research teams that previously could not afford the computational resources for such analyses. Chai Discovery’s Chai-2, backed by OpenAI and Anthropic, introduced zero-shot antibody design with hit rates of 16 to 20 percent, representing a 100-fold improvement over previous computational benchmarks. OpenFold3 released an open-source alternative to AlphaFold3 with commercially usable protein-ligand co-folding capabilities, democratising access to structural biology tools that were previously available only to the largest pharmaceutical companies.
AI-native biotechs are showing materially higher Phase I success rates while shortening timelines by 40 to 50 percent compared with traditional discovery programmes. The Capgemini 2026 report on AI in biopharma found that 41 percent of R&D leaders are now planning to automate entire discovery workflows using agentic AI — systems that do not merely analyse data on demand but autonomously propose hypotheses, design experiments, evaluate results, and iteratively refine molecular candidates in a continuous computational loop that operates around the clock.
The Agentic Transition in Drug Discovery
The transition from AI as a tool to AI as an agent in drug discovery is the most significant near-term development in pharmaceutical science. Traditional AI drug discovery requires a scientist to frame a problem, submit a query, evaluate the AI’s output, and decide on the next step. Agentic drug discovery inverts this workflow: the AI system maintains the strategic objective — identify a potent, selective, synthesisable candidate against Target X with acceptable ADMET properties — and autonomously executes the design-make-test-learn cycle, instructing robotic synthesis platforms to make the compounds it predicts will be best, directing automated assay systems to test them, interpreting the results, updating its molecular models, and generating the next round of candidates, all without requiring a human to mediate each step. The self-driving laboratory concept, long discussed theoretically, is now operational in multiple major pharmaceutical research facilities.
The regulatory implications are significant and only partially resolved. The FDA published draft guidance in January 2025 providing a risk-based credibility assessment framework for AI models used to support regulatory decision-making. Final guidance is expected in mid-2026. The EU AI Act’s high-risk provisions take effect on 2 August 2026, potentially classifying certain drug development AI applications as high-risk and imposing compliance requirements around model transparency, training data governance, and ongoing performance monitoring. Pharmaceutical companies using AI in regulatory-critical workflows face the simultaneous challenge of moving faster with AI while building the documented evidence base that regulators will require to accept AI-influenced submissions.
India’s Strategic Position
India’s pharmaceutical industry — the world’s third largest by volume — stands at a crossroads in its relationship with AI drug discovery. As a dominant generics manufacturer, India’s competitive advantage has historically rested on process chemistry and regulatory competence rather than novel discovery. AI potentially changes this equation: computational drug discovery requires intellectual capital, not physical infrastructure, and India’s deep talent pool in chemistry, biology, and data science is an underutilised resource. Initiatives like the Biotechnology Industry Research Assistance Council’s support for AI-bio startups and IIT and IISc research in computational chemistry are promising, but the scale of investment needed to compete with dedicated AI drug discovery platforms requires engagement from large Indian pharmaceutical companies willing to transition R&D budgets toward computational approaches.
– Sri Harsha
