Global AI in Drug Discovery Market Outlook to 2032

Executive Summary

AI in drug discovery has moved out of its proof-of-concept phase. The integrated category, spanning AI software platforms, computational drug-discovery services, AI-native biotech R&D, and pharma in-house AI deployment, grows from US$2.5 billion in 2024 to US$22 billion by 2032 at 30% CAGR.

Three forces drive the trajectory. The technology stack passed a credibility threshold in 2024-25: AlphaFold 3 (May 2024) extended structure prediction to small molecules and nucleic acids; Isomorphic Labs raised US$600 million Series A (March 2025); Insilico Medicine's rentosertib Phase IIa results were published in Nature Medicine (June 3, 2025) as the first peer-reviewed clinical proof-of-concept for an AI-discovered drug. Big-pharma capital migrated from speculative bets to scaled multi-target collaborations: Isomorphic Labs' January 2024 Lilly (US$1.7 billion) and Novartis (US$1.2 billion) deals, Generate Biomedicines' Novartis (over US$1 billion, September 2024) and Amgen (over US$1.9 billion, expanded January 2025), the NVIDIA-Lilly US$1 billion co-innovation lab (January 2026). And regulatory frameworks converged: FDA Modernization Act 2.0 (December 2022) and 3.0 (Senate-passed December 2025), EMA September 2024 reflection paper, EU AI Act (August 2026 enforcement).

Market Overview

The AI in drug discovery scope captures four flow components: AI software platforms (NVIDIA BioNeMo, Schrödinger, Atomwise, Iktos), computational services (Recursion, Insilico, Isomorphic offered to pharma partners), AI-native biotech R&D, and pharma in-house deployment (Lilly-NVIDIA co-lab, Pfizer IPC, AstraZeneca, Roche pRED, Novartis Basel). We exclude downstream commercial-stage drug revenues and contract research that is not AI-augmented.

The category sits at the intersection of three forces. Pharmaceutical R&D productivity has been declining for two decades; AI enters a market that has been searching for an inflection-point technology since the 1990s. The biotech capital cycle compressed materially in 2023-24, making partnered cash flows from big-pharma deals a structural feature of the category. And the binding constraint shifted from compute, capability, or capital to clinical translation; Phase II success rates remain near historical industry averages even for AI-originated assets.

Geopolitically, North America anchors 56% of category value (US-headquartered AI-native biotechs and pharma in-house). Europe (Isomorphic Labs London, AstraZeneca, Roche, Novartis) holds 22%. China at 11% includes Insilico's Phase IIa for rentosertib run across 21 Chinese sites; the NMPA's 2024 AI/ML guidance supports cross-jurisdictional programmes.

Market Size and Growth Outlook

The 2024-28 phase compounds at approximately 33% as the FDA Modernization Act 3.0 framework opens annual non-clinical-testing spend (US$3-5 billion) to AI/ML competition, the Insilico rentosertib Phase IIa proof-of-concept catalyses pharma willingness-to-pay, and the NVIDIA-Lilly US$1 billion co-innovation lab sets the integrated pharma-GPU-vendor capex template. From 2029 onward growth moderates to 25% as the category reaches scale and consolidation completes.

Market Segmentation

By Discovery Workflow Stage

Hit discovery and lead optimisation lead at 38% because this is where AI's capability advantage is most directly substitutable for traditional medicinal chemistry FTE-time. The fastest-growing subsegment is clinical-trial design enhancement (10% currently, expected to reach 17% by 2032 at 38% CAGR) as digital-twin, synthetic-control-arm, and trial-eligibility-scoring methods cross regulatory acceptance.

By Buyer Cohort

Top-20 pharma in-house deployment leads at 40% and is the structurally most important cohort because it captures both internal AI capex and downstream willingness-to-pay for partnered platforms. The Lilly-NVIDIA US$1 billion co-innovation lab (January 2026) is the template; pharma and GPU-vendor co-investments at five-to-ten-year duration replace single-target partnered-platform deals.

By Therapeutic Area

Oncology dominates at 41% because of rich multi-omic substrate, highest unmet-need-to-revenue economics, and target-class diversity AI can exploit. The fastest-growing therapeutic area is rare / genetic disease (10% to 16% by 2032) as orphan-drug economics combine with AI's strength on small-data target validation.

Trends and Developments

First clinical proof-of-concept and pipeline credibility

Insilico Medicine's rentosertib (ISM001-055 / INS018-055) Phase IIa results published in Nature Medicine on June 3, 2025 became the first peer-reviewed clinical proof-of-concept for an AI-discovered drug. The 71-patient placebo-controlled trial across 21 Chinese sites showed an FVC improvement of 98.4 mL at 60 mg in 12 weeks versus a 62.3 mL decline on placebo. Insilico's broader 2025 pipeline carries 8 clinical programmes; Phase IIb proof-of-concept study targeted for 2025-26.

Big-pharma deal commitments compound

Cumulative committed deal value across AI-native biotech and big-pharma collaborations exceeded US$10 billion in upfront and milestone potential between January 2024 and February 2026. Anchor deals: Isomorphic Labs' January 2024 Lilly (US$45 million upfront and US$1.7 billion milestones) and Novartis (US$37.5 million and US$1.2 billion), Isomorphic's February 2025 Novartis expansion, Generate Biomedicines' September 2024 Novartis (US$65 million upfront and over US$1 billion in milestones), Amgen's January 2025 expansion (US$370 million sixth-programme milestones), and the NVIDIA-Lilly US$1 billion co-innovation lab (January 2026).

Standalone AI-native biotech model consolidates

Recursion's all-stock acquisition of Exscientia (announced August 2024, closed November 2024) marked the beginning of category consolidation. The combined entity carries approximately 10 clinical and preclinical programmes, a phenomic dataset of more than 60 petabytes, and more than US$20 billion in milestone potential. BenevolentAI's December 2024 second restructuring in two years and Euronext Amsterdam delisting demonstrated the alternative path. Schrödinger cut software-revenue guidance from 10-15% to 8-13% in mid-2025 reflecting biotech-capital-cycle drag.

Other relevant developments include compute substrate concentration to NVIDIA BioNeMo (named adopters as of Q1 2026 include Lilly, Amgen, Novo Nordisk, Merck KERMT, Thermo Fisher, Basecamp, Dyno, Terray), pharma in-house AI capability building accelerating beyond Lilly to AstraZeneca, Pfizer, Roche, and FDA companion diagnostic approval acceleration in 2024-25. Domain-specific foundation models (AlphaFold-derived from Isomorphic Labs, NVIDIA BioNeMo's protein-folding suite) extend the generative-AI capability frontier into life-sciences workflows.

Competitive Landscape

NVIDIA BioNeMo's 18% understates strategic importance because the platform sits underneath most other top-10 operators' compute stack. Top-5 application-layer operators (Recursion, Schrödinger, Isomorphic, Insilico, Generate) collectively control 41% of category value; we expect this to rise to 52% by 2032 as operator consolidation compounds. Recursion is the largest standalone application-layer operator post-Exscientia merger. Schrödinger has the highest-quality physics-based moat. Isomorphic is the most capitalised AI-native operator. Insilico has the highest clinical-credibility position post-rentosertib.

Key Policies and Regulatory Environment

FDA Modernization Act 2.0 and 3.0

FDA Modernization Act 2.0 signed December 29, 2022 removed the 1938 mandatory animal-testing floor for new drug applications. April 2025 FDA roadmap focused initially on monoclonal antibodies. FDA Modernization Act 3.0 passed the US Senate by unanimous consent December 2025. Estimated annual non-clinical-testing spend opened to AI/ML competition: US$3-5 billion through 2030.

EMA Reflection Paper and CHMP qualification opinion

EMA's reflection paper on AI in the lifecycle of medicines was adopted by CHMP and CVMP and published September 9, 2024. In March 2025, CHMP issued the first qualification opinion on an AI-based innovative development methodology, moving AI from reflection-paper status to formally accepted methodology in the European regulatory toolkit.

Other relevant frameworks include the EU AI Act high-risk classification for medical AI (full enforcement August 2026; 8-14% compliance overhead), China NMPA AI/ML guidance (2024) which enabled Insilico's 21-site Chinese trial, IndiaAI Mission healthcare track integration with India's existing generic-pharma R&D infrastructure, the BIOSECURE Act spillover effects on US-China life-sciences cooperation, and the FDA AI/ML SaMD framework establishing precedent for quality-management-system requirements that increasingly extend into discovery-side toolchains.

Future Outlook

The AI in drug discovery market in 2032 reaches approximately US$22 billion. Pharma in-house AI deployment accounts for 38% of category value (up from 22% in 2024). AI-native biotech R&D for 32% (down from 40% as operator consolidation concentrates spend). AI software platforms for 17%. Computational drug-discovery services for 13%. The structural shift through 2032 is migration from external AI vendors selling discrete tooling toward integrated pharma-platform partnerships that make AI a first-party R&D function.

The competitive landscape consolidates further. Top-5 application-layer operators expand from 41% to 52% of category value by 2032. NVIDIA BioNeMo's substrate position expands from 18% to 22% through compounding pharma customer lock-in. Active AI-native biotech operator count consolidates from approximately 40 today to 18-22 by 2032.

The biggest risk is a translation-failure cluster in 2026-28. The 15-20 AI-discovered Phase II readouts due over the next 24 months will determine whether AI changes Phase II outcomes or only Phase I throughput. If readout cluster reverts to historical Phase II baselines, the back-half forecast (2029-32) compresses by 25-35%.