Introduction
AI in healthcare has progressed from conference-deck aspiration to clinical pipeline reality. The transition is measurable: a drug discovered and designed with generative AI is now in a Phase 3 trial, the FDA lists more than 1,600 AI-enabled medical devices, and every major pharma company has built, bought or licensed AI capabilities for drug discovery, clinical trial optimization, and manufacturing. At JPM 2026, the narrative shifted decisively: AI is no longer a differentiator that companies trumpet in investor presentations. It is infrastructure that companies are expected to have.
AI in Drug Discovery
The most commercially significant application of AI in healthcare is drug discovery, where AI platforms can compress target identification and lead optimization timelines from years to months. The landmark proof point is Insilico Medicine's rentosertib (formerly ISM001-055), a TNIK inhibitor for idiopathic pulmonary fibrosis whose target and molecule were both found with generative AI. Insilico went from target nomination to preclinical candidate in under 18 months versus the traditional 4-5 years, published positive Phase 2a results in Nature Medicine in 2025, and dosed the first patient in a 320-patient Phase 3 trial in September 2026. Public investors bought the story: Insilico's December 2025 Hong Kong IPO raised HK$2.28 billion (about $290 million), the largest biotech listing in Hong Kong that year.
- AI-Native Biotech
A biotechnology company that uses AI/ML as its core platform for drug discovery and development, rather than layering AI onto traditional discovery workflows. Companies like Recursion Pharmaceuticals (which absorbed Exscientia in 2024), Insilico Medicine, insitro, and Isomorphic Labs (DeepMind) have built proprietary datasets and models that generate novel molecular structures, predict binding affinity, and optimize ADMET (absorption, distribution, metabolism, excretion, toxicity) properties computationally before synthesizing physical compounds. The investment thesis is that AI-native platforms generate more candidates, faster, at lower cost, and with higher probability of clinical success.
Major pharma companies are licensing and acquiring AI capabilities rather than building everything from scratch. The platforms consolidated first: Recursion's all-stock combination with Exscientia, valued at about $688 million when announced in August 2024, closed that November and joined two leading AI discovery engines. The 2026 deals are larger on paper. Eli Lilly licensed preclinical oral programs from Insilico for $115 million upfront and up to about $2.75 billion in total, and joined NVIDIA in committing up to $1 billion over five years to a shared AI lab. AstraZeneca's October 2026 collaboration with insitro pays $40 million upfront against about $1.46 billion in milestones. These partnerships create a growing pipeline of advisory work for healthcare bankers covering AI drug discovery companies.
AI in Medical Devices and Diagnostics
Devices are where AI is most commercially mature. The FDA has authorized over 1,600 AI-enabled medical devices, and 2025 set a record for new authorizations. Radiology (automated image analysis for screening and diagnosis) accounts for roughly three in four, with cardiology (ECG interpretation, arrhythmia detection) a distant second and digital pathology (cancer grading from tissue slides) a smaller but strategically important niche.
For healthcare bankers, the AI diagnostic space presents both M&A advisory and valuation opportunities. Diagnostic AI companies with FDA-cleared algorithms and established clinical adoption are acquisition targets for large device and diagnostics companies (medtech tuck-in strategy) and health IT platforms. The clearest example is Roche's agreement to acquire PathAI in May 2026 for $750 million upfront plus up to $300 million in milestones, folding a partner since 2021 into Roche Diagnostics; even here, nearly 30% of the headline value is contingent. The key valuation driver is whether the AI product can achieve reimbursement, since many AI diagnostic algorithms lack dedicated CPT codes and rely on bundled payment or institutional purchasing.
The next article examines the $236 billion patent cliff and why it is driving a biopharma M&A supercycle.
