Biopharma's AI investment brings results and new challenges

The biopharmaceutical industry faces record revenue losses from expiring brand exclusivity, while its research and development (R&D) engine for replacing that revenue is still slow, expensive, and high-risk. In our analysis of early clinical-stage compounds, candidates from AI-native biotechs have success rates that are more than 40% higher than those reported by the industry overall.

The clinical evidence that manufacturers, R&D service providers, and investors have been waiting for seems to be arriving. Findings are still nascent, but the question they raise is one the industry has not had to ask before: If AI helps more viable candidates enter and clear early clinical development, will the capital and portfolio capacity exist to advance them? And how will the broader R&D ecosystem evolve in response?

Biopharma revenue pressures increase the need for R&D productivity

Biopharma's top line is under simultaneous pressure from pricing scrutiny on innovation, a record wave of expiring exclusivity, and faster erosion once that protection ends. Industry loss-of-exclusivity will total roughly $440 billion in worldwide product sales between 2028 and 2032 — more than double the prior five years — spanning immuno-oncology, cardiovascular disease, metabolic disease, and immunology.

These dynamics underscore the urgency of improving pipeline productivity. The average drug takes 7.8 years from first human dose to approval, roughly 90% of candidates fail, and development averages $2.67 billion per asset. Industry returns reached a post-COVID high of 7% internal rate of return (IRR) in 2025, driven by GLP-1 assets. Excluding them, 2025 IRR was 2.9%. The industry desperately needs a structural solution to the persistent productivity gap.

AI investment in drug discovery surges as clinical evidence emerges

Biopharma has already placed extensive bets on AI as a route to higher productivity. Investment in AI and machine-learning-based drug discovery and R&D licensing partnerships grew rapidly from $12 billion in 2024 to $43 billion (total committed funds) in 2025. The first half of 2026 added another $46 billion.

While AI has gained the most traction in discovery and early preclinical research, investment is now following molecules downstream into protocol optimization, site selection, and trial execution. Yet, despite the speed at which investors and sponsors have moved, meaningful evidence of the impact on R&D outcomes has been lacking. AI-discovered molecules still account for only about 1% to 2% of the clinical pipeline, according to our analysis. But as more programs have recently reached Phase I and Phase II, a decisive question can be asked: Are these candidates advancing through the clinic at higher success rates?

AI-enabled drugs have higher success rates in early clinical trials

We identified AI-native biotechs — companies built around embedded AI discovery or R&D platforms — as a relatively clean proxy for AI-enabled assets in development. Using reported development data, we tracked each compound going back to 2016, confirming status as programs advanced. Our analysis identified 118 total programs in Phase I, 66 of which were active/ongoing. For Phase II, we identified 46 programs in total, with roughly half active/ongoing. The sample size for determining Phase II outcomes is notably limited, reflecting the relatively early stage of AI investment and use. Exhibit 1 shows the status of all identified AI-enabled programs by trial phase.

The early evidence indicates that AI-enabled drugs have significantly higher success rates in early clinical trials compared with historical industry rates. Our analysis shows a success rate of 74% for Phase I programs and 45% for Phase II based on progression to the next phase — which rises to 81% and 55%, respectively, using phase completion as the threshold. Compared with success rates reported by the Biotechnology Innovation Organization (BIO) in their Clinical Development Success Rates study, the AI-enabled program success rate exceeds the industry overall for both Phase I (76% versus 52%) and Phase II (45% versus 29%). As Exhibit 2 illustrates, leveraging AI as a core technology improves success rates by more than 40% in early clinical trials and by more than 55% in Phase II.

AI-native biotechs are, by definition, early movers. They are built from the ground up around AI platforms, with the discipline and data infrastructure to deploy them consistently. Their clinical outcomes should be read as a leading indicator rather than a ceiling. As the broader industry continues to embed AI tools into R&D workflows, improved phase-transition rates should extend beyond this initial cohort. The dynamic is the same whether deployed by a Series B biotech or a top-10 pharma: better target validation, higher-quality molecular candidates, and more rigorous go/no-go decisions entering the clinic.

How AI could reshape drug development, investment, and R&D services

These findings indicate that ongoing application of AI in discovery and early development — whether by AI-native companies or the industry more broadly — will result in more drugs advancing to bigger-budget, later-stage clinical trials. Here are several other important implications:

• The binding constraint in drug development shifts from science to capital. More programs advancing through Phase I means more decisions about funding expensive proof-of-concept and pivotal trials. Sponsors that cannot fund all worthy programs internally will need more creative partnering and financing approaches — or risk halting development on some attractive assets.
• Stranded assets become a tailwind to industry business development and licensing (BD&L). Midsize companies without big pharma balance sheets will look to monetize programs they cannot advance. This increases the volume of de-risked, late-stage assets available for licensing and co-development — a meaningful additional source of pipeline for large pharma trying to offset revenue at risk from loss of exclusivity.
• The net effect for tools and services vendors is additive. Conversations with pharma R&D leaders suggest AI-driven efficiency gains are unlikely to translate into smaller R&D budgets. More programs advancing through each development phase against a stable or growing spending base means greater aggregate demand for contract research organization, lab, bioanalysis, diagnostics, real-world evidence, and contract development and manufacturing organization services across the development continuum.
• There’s a need for more proactive, disciplined portfolio planning. An increase in attractive late-stage assets will only exacerbate the persistent portfolio-management challenges facing biopharma companies. Companies that lack the decision-making infrastructure to prioritize across a larger set of viable programs risk misallocating capital — funding less differentiated assets while breakthrough programs stall. As we have written before, the industry overall does a poor job of operationalizing the principles needed to manage both pre- and post-commercial portfolio decisions throughout product lifecycles.
• Winners in tools and services will be prepared for an increasingly AI-driven R&D environment. Success factors include the ability to scale for increased trial volumes, manage larger and more geographically distributed studies, and integrate automation into core service offerings — particularly for data management, medical writing, pharmacovigilance, and lab workflows that may face per-unit pricing pressure even as volumes rise.

As AI improves the prospects of early-stage drug candidates, the challenge for biopharma shifts from scientific productivity to capital discipline and portfolio decision-making. Companies that build that infrastructure now will be better positioned to capture the value these advances create.