S02.1 · Healthcare & Life Sciences
A $1.6-1.7T innovator-drug market where AI is compressing R&D cost and eroding the large-R&D-org moat.
Biopharmaceuticals — the innovator drug developers, from big pharma majors to a long tail of clinical-stage biotech — is a $1.6-1.7T global market (2025E, IQVIA Institute), growing 5-8% forward through 2028-2029 on oncology, immunology and obesity/GLP-1 launches. The question that matters is not the growth rate. It is whether a large R&D organization is still worth paying for. The same AI that shortens time-to-molecule is compressing the discovery cost curve that historically justified large drug-developer valuations, and it is doing so exactly as patent-cliff economics force the majors to replace revenue.
Global medicine spending grew 5-7% from 2019 to 2024; IQVIA projects 5-8% forward (IQVIA Institute). The US books roughly 45-50% of revenue and carries virtually all of the list-price premium. The EU5 and Japan follow. Production splits across the US, EU, Puerto Rico and Ireland, with API manufacturing shifting steadily to India and China. The top 10 pharma majors hold roughly 35-40% of branded revenue, but the innovation pipeline runs through thousands of mostly pre-revenue, VC-backed clinical-stage biotechs. It is a barbell: a handful of commercial-scale players sitting downstream of a deep, fragmented discovery layer.
The chain runs from academic research and VC-backed biotech through CROs and CDMOs upstream, to payers, PBMs, distributors and providers downstream. On-patent economics are exceptional — 70-85% gross margins — and binary. Clinical and regulatory risk is existential at the molecule level, and reimbursement bodies (CMS, NICE, EU HTA) set realized price directly; list price does not. Any developer with a concentrated product base carries a structural revenue-replacement problem from patent-cliff exposure, however well that base performs today.
Agentic AI hits the discovery cost line first: target identification, trial design and regulatory-document drafting are becoming automated tasks rather than large-team functions. The compression is structural, not incremental. A large R&D organization has been treated as a moat; AI-native biotechs now run lean pipelines that previously required hundreds of scientists and clinical-operations staff. What falls is the fixed cost of reaching an IND filing — clinical and regulatory risk stays binary regardless of how the molecule was discovered.
The upstream layer is also changing how it gets paid. Where a platform's differentiated asset is an algorithm rather than a proprietary compound, licensing-first models are displacing the traditional own-and-develop model — a change in how value is captured, not a productivity tool bolted onto existing R&D. Target discovery and trial-matching are already there; AI-designed molecules dominating pipelines outright is a 5-10 year horizon. Big pharma's own push into CDMO and manufacturing capacity, companion diagnostics and digital health for adherence and real-world evidence is the mirror image of the same trade: control more of the value chain as the discovery layer becomes cheaper and more commoditized to access.