Over the past 20 years, the average cost of bringing a medicinal product to market has increased approximately threefold, reaching $2.5 billion, while the probability of success for an individual molecule at the late stages of clinical trials, according to Morgan Stanley estimates, has remained at just 8–12%.
Against this backdrop, the traditional R&D model based on sequential experiments and intuitive hypotheses is increasingly proving insufficient, prompting the pharmaceutical industry to turn to artificial intelligence and digital technologies. Over the past five years, venture capital investment in AI-driven drug discovery projects has totaled about $14 billion, including $6.7 billion in 2024 after a period of stagnation. The number of strategic partnerships between major pharmaceutical companies and AI/IT projects has exceeded 900, compared with 105 at the end of 2021.
Among the most significant deals, the publication highlights, in particular, the acquisition by Switzerland’s Roche of the U.S.-based oncology real-world data platform Flatiron Health for $1.9 billion in 2018.
The partnership between Sanofi and the UK-based Exscientia, concluded in early 2022, became one of the most high-profile AI deals in the history of the pharmaceutical industry, primarily due to the agreement’s potential value of more than $5 billion with an upfront payment of $100 million. Exscientia offers what is especially attractive to big pharma—acceleration of the early stages of R&D. Using machine learning algorithms, the company automates small-molecule design, shortening the path from hypothesis to candidate molecule.
The partnership between Novartis and the UK-based Relation Therapeutics, concluded in February, appears more modest, but strategically it is more ambitious. Novartis has focused on the earliest and least formalized stage—understanding disease biology. This is not about optimizing known targets, but about systematically addressing the question of why a disease arises. For Novartis, this is a way to increase the probability of success across its entire development portfolio, since errors at the target selection stage are precisely what lead to failures at later stages, regardless of the quality of the molecule.
Source: GMP News, December 23, 2025.
