AI for Biopharma, Part I

For some time, “healthcare” has been the reflexive answer of AI optimists pressed on the promise of AI. This series tests that instinct, from the perspective of a fellow optimist.

Today’s post is about AI in drug development: what is happening, what has been produced, and what it might mean. A subsequent piece will take up why progress might look slower than expected, and what changes are (or should be) on the horizon.

Where are we now?

In 2026, AI is standard tooling in drug R&D.

One sign of how routine it has become: AlphaFold’s open structure database has been used by more than three million researchers across 190 countries.

Drugmakers spend roughly $300 billion a year on R&D, and all of the world’s top biopharmas now use AI somewhere in the pipeline.

It’s visible from the regulatory side as well. FDA drug and biologic submissions with an AI or machine-learning component jumped from a handful in the late 2010s to 132 in 2021 - approximately a tenfold leap in a single year - and passed 500 cumulatively by 2025.

Capital has followed. Private investment in AI-focused drug discovery has reached tens of billions over the past decade, and dozens of AI-associated programs have entered human testing since 2020 (counts vary depending on the definition of an “AI drug”).

All this raises the question of what exactly it means for AI to be involved in drug development. Activity is not spread evenly: it clusters upstream (discovery) and successes there have truly been notable.

AlphaFold predicts a protein’s 3D structure from its amino-acid sequence at near-experimental accuracy. AlphaGenome (2025) predicts how single-letter DNA variants change gene regulation across the non-coding genome. RFdiffusion generates novel protein backbones from scratch with a diffusion model, while Chai and AlphaFold3 predict how proteins bind small molecules, DNA, and one another. Generative-chemistry models propose new drug-like scaffolds.

Downstream use - in trials, approval, and surveillance - is quieter, but not absent: PK/PD modeling, digital pathology, trial enrichment, and pharmacovigilance already use mature computational methods. In December 2025, FDA qualified AIM-NASH, its first AI drug-development tool, for AI-assisted liver-biopsy scoring in MASH trials.

This pattern is not accidental: discovery, relative to wet-lab experimentation, is among the more tractable problems for current ML architectures. Structure prediction has a clean, gradable objective (molecular accuracy against known structures) as well as a large labeled dataset in the Protein Data Bank. Molecular design offers fast in-silico feedback. As Demis Hassabis has put it, he has long “thought of biology as an information processing system at a fundamental level.”

It also fits how the industry innovates: much early experimentation happens in small, nimble biotechs that adopt new methods fast, then license or sell to big pharma. What follows is a brief survey of how these capabilities have been built and deployed - not exhaustive, but enough to make the point.

Figure 1 — AI capabilities across development stages (2018-2026). Interactive; hover any dot, or open the full chart ↗.

What has it yielded to date?

Plenty of activity in the pipeline, and very little in the market - largely what you would expect at this point.

Clinical trials consume the majority of drug-development capital and time: about 69% of R&D cost, and on average close to eight years of testing versus under three for preclinical work.

No de novo molecule whose target or molecular design originated principally on an AI platform has reached the market. (Of note, AI-assisted repurposing is different: marketed examples exist, including Igalmi.)

The table below is a curated, non-exhaustive set of programs in which AI materially affected target discovery, screening, design, optimization, or repurposing. Those roles are labeled separately.

Figure 2 — AI-discovered and AI-designed drugs in the clinic. Open the full table ↗.

A few observations worth stating plainly.

• “Zero approvals” is not a verdict. The first AI-designed molecules entered trials around 2020, and first-in-human to market runs about eight years on average. The leaders - Insilico’s rentosertib (Phase III for pulmonary fibrosis) and Relay’s RLY-4008 (an FDA decision expected around September 2026) - are close, not done.
• The misses are varied. Exscientia’s DSP-1181 (OCD), BenevolentAI’s BEN-2293 (eczema), and the ulotaront schizophrenia program suffered setbacks across different diseases and mechanisms. The sample is too small and heterogeneous to reveal an AI-specific pattern. The only clear conclusion is that human translation remains a central risk.
• It is not entirely clear what an “AI drug” is. Definitions span “AI helped screen,” “AI designed the molecule against a known target,” and “AI found the target and designed the molecule” - rentosertib’s claim.

So: real molecules, real momentum, no approvals yet.

What does this all mean for public health?

A harder question - and I’ll afford myself some authorial license to oversimplify. Let’s say for a moment that the social value biopharma delivers is a function of three things:

• Diversity & efficacy - distinct diseases for which an effective therapy exists.
• Access - affected populations’ ability to receive and afford the treatment.
• Clinical deployment - how effectively the treatments we have are actually used.

AI’s clearest effect is on the first: more targets, more hits, more leads, faster.

But that is also the slowest lever to reach a patient. Clinical development and review still run close to a decade between a designed molecule and a market authorization.

Access is, for now, largely untouched. Affordability depends principally on pricing, coverage, diagnostics, distribution, and healthcare-system capacity, rather than approved-drug volume. Faster discovery does not obviously lower prices.

Regulatory change and lab automation could move both - a faster FDA, cheaper production - though the net direction is not obvious.

Clinical deployment may be the most underrated of the three.

We have more than 20,000 approved prescription drug products. Most are still prescribed using population-average guidelines. The largest near-term gains may come not from a new molecule, but from using existing ones better: matching drug to patient through richer biological data. A topic for a later post.

So, will AI open the floodgates?

Very likely, eventually. Leading labs are explicit about the ambition: Isomorphic Labs was founded to “solve disease” with AI, and Hassabis has publicly predicted the technology will compress drug discovery timelines dramatically.

Initial results have been impressive, as attested above. The pipeline has gone from near-zero to 150-plus programs in five years. Every major biopharma is now a buyer.

But “eventually” is doing some work in the above. Part II is about the path to drug abundance: stay tuned.