AI has been a buzzword in the pharmaceutical industry for almost a decade, with countless headlines promising to revolutionise how drugs are discovered, developed, and deployed. The hype has typically outpaced reality; however, with a growing number of AI-generated drug candidates entering clinical trials, we find ourselves at a crucial potential turning point for the field. As the first AI-designed molecules begin clinical trials, the industry is watching closely to see if AI can deliver novel, successful therapies or if it will remain a supporting tool in an already complex research process.
The evolving role of AI in research and development
AI is not a single technology, but a broad set of computational tools that can support nearly every stage of drug development. In early discovery, AI can help identify novel targets, generate new chemical compounds, predict how drugs will interact with proteins, and optimise lead compounds (promising compounds that show potential as an effective drug) for desirable properties. In later stages, it is increasingly used to select patients for clinical trials, repurpose existing drugs, and predict adverse effects.
Go deeper with GlobalData
Early attention focused on the ability of generative AI models, particularly deep learning and reinforcement learning frameworks, to create candidate compounds faster than any human chemist, with a theoretical ability to explore vast chemical spaces more efficiently than traditional methods.
AI-designed candidates used to remain confined to in silico (computational) experiments or the preclinical stage. That is now changing.
AI steps into the clinic
In the past two years, several AI-designed drugs have advanced into human trials, providing the first real-world test of these technologies in the clinic.
Insilico Medicine, a Hong Kong- and New York-based AI biotech company, gained attention in 2023 when it announced that its drug candidate INS018_055, developed using its proprietary Pharma.AI platform, had entered Phase II trials (testing effectiveness and safety). The compound was designed from scratch using generative models trained on structural data, an early confirmation that AI can do more than just screen libraries.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataOther companies such as BenevolentAI, Recursion, Schrödinger, and Relay Therapeutics are advancing candidates identified, optimised, or prioritised using AI tools, though not all are strictly ‘AI-generated’. Importantly, these companies are not just discovering molecules but also positioning themselves as drug developers and strategic collaborators.
Pharma’s buy-in: a wave of strategic partnerships
Big pharma has increasingly embraced AI, using partnerships, joint ventures, and acquisitions to reduce risk in target selection and accelerate early-stage drug development. A prime example is this deal between Sanofi and Exscientia, worth up to $5.2bn, focused on AI-designed small molecules across oncology and immunology.
Furthermore, AstraZeneca has been collaborating with BenevolentAI to identify new drug targets in chronic kidney disease and fibrosis. Pfizer, Bayer, Merck, and Roche have all partnered with AI-native biotechs or built their own internal capabilities, often focused on areas such as rare diseases, central nervous system disorders, or target deconvolution.
These collaborations reflect a broader shift in mindset: AI is increasingly seen not as a competitor to traditional research and development (R&D), but as a strategic enabler, using human expertise to improve accuracy and compress timelines in high-risk areas.
Gain in efficiency, or a new bottleneck?
The core value proposition of AI in drug discovery is speed and efficiency. Drug development typically costs over $2bn and can take 10-15 years per new drug. AI promises to shorten the time from target identification to candidate nomination by rapidly generating hit compounds, improving target-disease linkage accuracy, and selecting better patient populations for trials.
Insilico, for example, claimed that INS018_055 progressed from target discovery to Investigational New Drug filing in under 30 months, significantly faster than industry averages. However, real-world validation of these efficiency claims is still pending, and it remains unclear whether AI shortens timelines in later-stage development, where most costs and failures still occur.
Moreover, regulatory pathways for AI-designed drugs are still evolving. While the molecules themselves follow industry standards, questions remain around intellectual property ownership, algorithm transparency, and validation of model-generated hypotheses.
Hype, valuations, and investor reality checks
The excitement around AI in drug discovery has driven enormous investor interest, bringing volatility along with it. Many AI-first biotech companies raised substantial capital through initial public offerings or special-purpose acquisition companies (SPACs) deals during the 2021-22 biotech boom. Yet several have seen share prices decline sharply as investor enthusiasm met the realities of long development timelines, modest clinical progress, and uncertain monetisation strategies.
BenevolentAI, for example, went public via a €1.5bn ($1.8bn) SPAC in 2022 but lost over 70% of its value by mid-2024. Recursion, despite its ambitious data-driven drug discovery platform and partnerships with Bayer and Roche, has also faced pressure from shareholders seeking faster returns.
Despite this, the field is still maturing. Investors and pharma partners are shifting from broad platforms to more focused evaluations based on data-driven productivity, clinical progress, and pipeline value.
What would success look like?
The industry is now approaching a key milestone: the first regulatory approval of an AI-designed drug. If INS018_055 delivers positive Phase II data, it would mark a transformative moment, validating AI not just as a tool but as a source of new and unique therapies.
Looking further ahead, AI could diversify drug development by enabling smaller companies to design high-quality molecules without vast lab infrastructure. It may also unlock progress in rare diseases, neglected indications, or emerging pathogens, where traditional business models struggle to justify investment.
Beyond small molecules, AI is also being explored in biologics design, protein engineering, messenger ribonucleic acid optimisation, and clinical trial design, suggesting its impact could extend across the entire pharmaceutical value chain.
From proof of concept to clinical proof
AI is no longer just a futuristic concept in pharma; it is producing real assets, forming strategic alliances, and slowly earning its place in the clinical pipeline. However, expectations remain high, and the burden of proof now rests on human trials, not machine models.
As the first wave of AI-designed drugs enters mid-stage development, the industry will be watching not just for approvals, but for evidence that AI can improve outcomes, accelerate timelines, and reduce costs in a field where failure is still the norm. If successful, the next decade may not just belong to AI, it may be designed by it.
