2026: the year AI stops being optional in drug discovery

In July 2026, the first Phase III trial of a fully AI-designed drug began enrolling patients with idiopathic pulmonary fibrosis, a chronic lung disease with few effective treatments. The drug, Rentosertib (INS018_055), was developed by Insilico Medicine using a platform that identified a novel target and generated the molecule in silico before it ever entered a laboratory. Its progression to late-stage testing is not an isolated event: it arrives in the same year the EU AI Act’s high-risk provisions take effect, the AI drug discovery market is projected to reach $8–10 billion, and the FDA is finalising guidance on AI in drug development. For the pharmaceutical industry, 2026 is the year AI stops being an optional experiment and becomes a structural part of R&D decision-making.

The stakes are unusually high. AI is projected to compress early drug discovery timelines by 30–40%, reducing the time to a preclinical candidate from three to four years down to 13–18 months. Across the industry, generative AI could deliver $60–110 billion in annual value, according to market projections. Yet those gains are concentrated in the earliest stages of research. AI does not shorten clinical trials or regulatory review, and most AI tools still fall outside formal regulatory scope. For executives and investors, 2026 is therefore a make-or-break year: positive Phase III results could establish AI as a credible engine for first-in-class therapies, while failures would likely accelerate consolidation and further discipline in a market already wary of hype.

A pivotal trial and an early efficacy signal

Rentosertib’s path to Phase III rests on a Phase IIa trial (NCT05938920) that produced a notable lung-function signal. After 12 weeks, patients receiving 60 mg once daily showed a mean forced vital capacity (FVC) improvement of +98.4 mL, compared with a –20.3 mL decline in the placebo group. FVC is a standard measure of lung function in IPF, where decline typically indicates disease progression. The result, while from a small early-stage study, suggests the AI-designed molecule may be modifying the disease rather than merely managing symptoms.

Insilico Medicine, led by CEO Alex Zhavoronkov and Co-CEO Feng Ren, used its Pharma.AI platform to discover the drug. PandaOmics identified the biological target, and Chemistry42 generated the molecule. The company reported $106 million in revenue in the first half of 2026, a figure that underscores the commercial viability of AI-native drug discovery even before a pivotal trial reads out. The Phase III trial that began in July 2026 will determine whether the early efficacy signal holds in a larger, longer study.

Consolidation, capital and the end of hype

The market’s structure changed before 2026. In November 2024, Recursion Pharmaceuticals and Exscientia completed a merger that created a combined company with about 800 employees and a vertically integrated AI platform. Recursion’s biological mapping capabilities were combined with Exscientia’s molecular design expertise, with Recursion CEO Chris Gibson leading the merged entity. The deal reflected a broader shift: AI drug discovery is no longer a collection of point solutions but an integrated capability that spans target identification, molecular design and preclinical validation.

The financial picture is mixed. The AI drug discovery market is expected to grow from $5–7 billion in 2025 to $8–10 billion in 2026. But investor discipline has tightened. According to industry analyses, the ratio between announced “biobucks” deal values and actual upfront payments is roughly 50:1, indicating that much of the headline value in AI-pharma partnerships is contingent and uncertain. Critics also point out that AI-discovered compounds have so far shown similar clinical progression rates to traditional drugs. That means the technology’s ability to pick better targets or design better molecules has not yet translated into a clearly higher probability of late-stage success.

Regulation arrives in August

On 2 August 2026, the high-risk provisions of the EU AI Act take effect. Some AI systems used in drug development could be classified as high-risk, triggering new compliance obligations around data governance, transparency, human oversight and risk management. The FDA is also finalising its own guidance on AI in drug development in 2026. For global companies, this creates a fragmented regulatory landscape: a tool that is not regulated in one jurisdiction may require formal conformity assessment in another.

The ambiguity is significant. Most AI tools in drug discovery currently fall outside regulatory scope because they do not make final decisions about patient safety or efficacy. But as AI moves deeper into clinical trial design, patient selection and digital twin simulation, the line between research tool and regulated medical device becomes harder to draw. Dr. Raminderpal Singh, writing in Drug Target Review, forecasts 2026 as a validation year, while Dr. Gen Li of Phesi notes that digital twins are moving from pilot projects into operational use in clinical trials. Both trends push AI closer to regulatory attention.

What AI can and cannot yet do

The clearest measurable impact of AI in 2026 is in early discovery. Computational prediction is now integrated early in workflows, with in silico exploration preceding wet-lab validation. This is where the 30–40% timeline compression is most credible: reducing preclinical candidate development from three to four years to 13–18 months changes the economics of R&D, even if later stages remain unchanged.

What AI cannot do is shorten the human testing that regulators require. Clinical trials still take years, and regulatory review still takes months. A molecule designed by AI must prove itself in the same Phase I, II and III gauntlet as any other drug. Rentosertib’s Phase III trial will not read out in 2026; its outcome will shape the narrative in subsequent years. For founders and R&D leaders, the practical implication is to focus on platform robustness and regulatory readiness rather than claiming end-to-end acceleration that the evidence does not yet support.

By the end of 2026, the industry will have a much clearer picture of whether AI can consistently deliver disease-modifying therapies, not just faster hits. The Rentosertib Phase III trial, the EU AI Act’s high-risk regime, and the post-merger integration of Recursion and Exscientia are all tests of a single proposition: that AI can improve the probability and economics of drug development. If the signals hold, AI will become a default component of discovery pipelines. If they do not, the market will consolidate further, and the next wave of investment will demand harder evidence of clinical differentiation. Either way, the era of AI as an optional add-on in drug discovery is over.