Key Takeaways
- Generative AI will transform pharmaceutical research and development by strengthening innovation, offering cost efficiencies, and speeding the traditionally ponderous research-to-market cycle.
- Large pharmaceuticals companies' significant investments in AI could establish enduring competitive advantages over smaller rivals.
- The benefits of AI will likely take years to filter through to treatments (and pharmaceutical company revenues) and remain dependent on further investment, regulatory evolution, and effective change management.
Large pharmaceutical companies are investing heavily in AI, driven by the promise of augmented research and development (R&D) capabilities that could transform the scope, cost, and speed of producing new treatments, including for currently intransigent illnesses.
Nascent benefits are already emerging. Yet the scope of AI-driven gains and their timing, remains subject to further investment, effective integration of the technology, and the adaption of regulatory regimes and laws. Companies (and regions) that succeed in incorporating AI promise to fundamentally change both the industry and health outcomes and in doing so could create a competitive advantage that will prove enduring and material to creditworthiness.
The application of AI in biotechnology and pharmaceutical research is notably supported by big pharmaceutical companies. S&P Global Ratings' study of 15 of the largest pharmaceutical companies found that they spend an average of about 18% of revenues on R&D: ranging from (an average of) about 27% at Eli Lilly & Co., to 12% for Merck KGaA, (source: Company reports for 2023, Evaluate Pharma). Companies included in our research are: AbbVie Inc., Amgen Inc., AstraZeneca PLC, Bristol-Myers Squibb Co., Eli Lilly, Gilead Sciences Inc., GSK PLC, Johnson & Johnson, Merck, Novartis AG, Novo Nordisk A/S, Pfizer Inc., Roche Holding AG, Sanofi, Takeda Pharmaceutical Co. Ltd.). The size of the budgets currently dedicated to AI is often unclear, but it is likely relatively small, uneven in size, and growing.
All larger pharmaceutical companies have proprietary AI-platforms used in research, or use third-party solutions that are integrated into their R&D. At the same time, big pharmaceutical companies are increasingly partnering with smaller, AI-driven pharmaceutical companies to accelerate drug discovery processes (see table 1).
Table 1
Large Pharmaceutical Companies Are Taking An AI Lead
Big pharmaceutical companies were early adopters of AI, lured by machine learning's potential to reduce the cost and time it takes to develop new medicines. Bringing a new drug to market typically takes 10 to 15 years, according to industry group Pharmaceutical Research and Manufacturers of America (PhRMA) and comes at an average cost of $1.3 billion and a median cost of $985 million, according to a study by the London School of Economics.
The industry's earliest AI projects date back about a decade and have culminated in the licensing of machine learning modules used in drug discovery to identify potentially useful compounds and explore potential applications. Meanwhile, AI's application in data processing has enabled clinical trial augmentation through greater use of existing data sets (like electronic health records, patient demographics, and previous trial results), resulting in improvements in the selection of sample populations for trials.
Successes and advancements in AI have led to wider adoption, integration, and new applications of the technology in drug development. Notably, pharmaceutical companies' are increasingly using generative AI to facilitate the creation of new molecular structures and drug compounds (see "Artificial Intelligence Powering Synthetic Biology: The Fundamentals," June 25, 2024). AI is also facilitating analysis of real-world data for existing treatments, improving companies' potential to file for additional indications. And traditional pharmaceutical companies now often collaborate with specialist AI-driven pharma-tech companies to build healthcare platforms.
Machine Learning's Multifaceted Role In Pharmaceutical R&D
Machine learnings is a subset of AI that focuses on the development of statistical algorithms that can process vast amounts of data to provide predictive analytics (see "Machine Learning: The Fundamentals," Nov. 29, 2023).
Big pharmaceuticals companies have notably integrated machine learning into R&D, where its ability to predict the biological activity of compounds, and analyze chemical structures and properties, helps to focus research on promising candidates. The technology has also been used to identify unstable and critical drug interactions, enhancing patient safety, and by marketing divisions to optimize pricing strategies (see table 2).
Table 2
| Pharmaceutical applications of machine learning | ||
|---|---|---|
| Research and development | ||
| Virtual screening: identification of drug candidates from virtual chemical libraries | ||
| Drug targeting: analysis of biological data associated with a disease to identify promising molecules/ drugs for treatments | ||
| Biomarker research: identification of biomarkers that can be used to diagnose diseases, predict treatment responses, and improve monitoring of patient outcomes | ||
| Drug repurposing: predicting existing drugs suitability for new applications | ||
| Clinical trials: optimization of patient recruitment and stratification for clinical trials | ||
| Safety and efficacy | ||
| Drugs interactions prediction: identification of potentially harmful drug interactions | ||
| Drug combination therapy: identification of synergies between drugs that could improve treatment of complex diseases, e.g., in oncology | ||
| Drug safety: monitoring safety of drugs on the market and reports of adverse effects | ||
| Operational and commercial | ||
| Drug pricing and market access: assessment of new drugs' pricing strategies , market access, and reimbursement landscapes | ||
| Drug manufacturing: optimization of the drug manufacturing process for efficiency and quality control | ||
| Source: S&P Global Ratings. | ||
Early adoption of AI was characterized by big pharmaceutical companies' deployment of machine learning systems developed by technology companies. Pfizer, for example, in 2016 announced a collaboration with IBM Watson, a computer system, to develop drug discovery capabilities in the fields of immunity and oncology. That combination of big pharma and big tech has expanded, with similar partnerships underpinning much of the sector's application of AI (see table 3).
Table 3
| Notable pharmaceutical and technology partnerships | ||||||||
|---|---|---|---|---|---|---|---|---|
| Pharmaceutical | Technology | Year | Partnership | |||||
| Pfizer | IBM Watson | 2016 | Application of machine learning system Watson Health (a commercial cloud-based AI tool) to drug discovery and other research capabilities. | |||||
| Merck/Astra Zeneca | Amazon Web Services (AWS) | 2017 | Development of a cloud-based drug discovery platform. | |||||
| Novartis | Microsoft | 2018 | Generative chemistry pipeline using supervised ML in genomic sequencing, protein science, and structural biology to provide preliminary links and possible combination for testing. | |||||
| Merck KGaA | AWS (Amazon Neptune--graph database) | 2021 | Creation of the ‘Change Assessment Knowledge Engine’ (CAKE) to assist in regulatory impact reporting. CAKE reduced change assessment duration by up to 90% and total manual effort per assessment by 30%–70%. | |||||
| Roche (Genentech) | Nvidia | 2023 | Protein structures prediction and screening of small molecules. | |||||
| ML--Machine learning. Sources: Company announcements, S&P Global Ratings. | ||||||||
AI, Biotech, And Pharma: New Operating Models And Competition
Collaboration between big pharmaceutical companies, biotechnology companies (biotechs), and technology providers has given rise to new breed of AI-driven biotechnology (biotech) entities that have the capability to accelerate the drug discovery and design process.
Biotechs traditionally focused on the discovery and development of new medicines, not least because they often lacked the scale and relationships to engage in manufacturing and manage stakeholders (including regulators). To overcome those limitations, the companies often partnered with pharmaceutical companies, which would manufacture and commercialize medicines.
That model has evolved with the emergence of a newer breed of biotechs that have placed AI at the heart of their R&D platform. These AI-biotechs are often part-owned by big pharmaceutical groups, whose equity underpins funding in exchange for a share of R&D successes, potentially through a share of resulting revenues, but also often through options to commercialize new treatments.
The relationship between AI-biotechs and big pharma won't necessarily be symbiotic. The formers' significant potential for innovation means they could yet emerge as competitors to pharmaceutical incumbents (and big pharma's R&D functions). For example, biotechnology company Insilico Medicine, in June 2023, announced a first, when a fully AI-generated drug (for the treatment of a rare lung disease) entered Phase 2 clinical trials in the U.S. and China.
Elsewhere, AI-biotechs are accelerating pharmaceutical research and development. For example, U.K.-based Exscientia has created an advanced drug candidate (called ELUCIDATE) for the treatment of solid tumors, while Dutch company Cradle is using AI to accelerate protein sequencing. And AI-biotech companies (see table 4) are also increasingly leveraging their technology to provide specialist services to established pharmaceutical companies.
Independent AI-driven companies' smaller size will mean they often lack the investment firepower needed to efficiently progress treatments through approval and launch. We expect that will result in partnerships with big pharmaceutical companies, contract research organizations (CROs), and contract development and manufacturing organizations (CDMOs).
Table 4
| Notable AI-biotech companies | ||||||||
|---|---|---|---|---|---|---|---|---|
| Selected biotech | AI platform | Expertise | Number of programs (as of September 2024) | |||||
| Atomwise | AtomNet | Prediction of small molecules for drug discovery using target (disease) structure analysis. | 4 | |||||
| BenevolentAI | BenAI Engine | Drug discovery, with ability to provide prediction rationales. | 5 | |||||
| Insilico Medicine | PHARMA.AI | Target identification, ability to generate novel molecule data, clinical trial outcome predictions. | 31 | |||||
| Exscientia | CentaurAI | Target selection, experiment design, enhancement of clinical assessment. | 4 | |||||
| Source: Company records, S&P Global Ratings | ||||||||
Generative AI Offers New Pathways To Accelerate R&D
Generative AI's ability to create useful and novel outputs from complex training data has enlarged the scope of AI's applications in the pharmaceuticals sector. Notably, the technology opens the door to the identification of new treatments from training sets that incorporate diverse elements such as molecules, proteins, and enzymes (much like LLMs generate meaningful output from text). This capability is already being deployed by pharmaceutical and biotechnology companies to generate new compounds, explore new possibilities for existing medicines, and accelerate the development process.
Generative AI's potential to unlock efficiencies and economies across the whole drug development process could be considerable. The technology could reduce the cost of clinical trials by up to 50%, while also accelerating the process, according to McKinsey & Co., a management consultant (see "Generative AI in the pharmaceutical Industry: Moving from hype to reality," McKinsey & Co., published Jan. 9, 2024).
Three key opportunities of generative AI in pharma:
Discovery and development of novel therapies
Generative AI is reducing the time and cost of R&D, including through faster identification of new molecules and compounds, improved efficacy and safety forecasting (including through digital testing), by suggesting repurposing of existing drugs, by optimizing trial design, and by personalizing medicines using patient data. Examples of such initiatives include Merck KGaA's Aiddison AI-platform, Astra Zeneca's Reinvent, and the deployment of the Melloddy project, which is backed by 10 pharmaceutical companies.
AI's contribution to accelerated drug discovery could create sustainable competitive advantage for early adopters. Given the likelihood that those benefits will accrue to larger-scale actors (with larger AI budgets) the resulting benefits could also prompt consolidation.
Operational efficiency and process optimization
Generative AI can assist in streamlining manufacturing processes (including through improvements in robotics) and regulatory compliance (by monitoring drug reactions, maintaining data integrity, and assisting with regulatory submissions). This could reduce labor costs, particularly for manual tasks, and enable further investment in AI expertise. Examples include, Novartis's AE Brain, which scours texts for evidence of unexpected drug outcomes, and Sanofi's Plai, which aggregates company data to improve and accelerate decision making, improve supply chain management, and to design better research projects.
Facilitating engagement with patients and healthcare professionals
AI-powered tools can provide personalized information to patients and predictive analysis and recordkeeping, which should enable health professionals to offer better and more rapid advice and treatment. Novartis's AI Nurse, for example, is a disease management tool that facilitates health professional engagement with heart failure patients' health indicators and anticipates disease progression.
The Challenges And Risks Of AI In Pharma
AI's potential to drive positive change in the pharmaceuticals sector is clear, but the technology will also introduce challenges and threats. Some of those issues are inherent to AI generally, while others will be particular to the pharmaceutical industry:
Overabundant data and variability
Drug discovery data sets can involve millions of compounds, which traditional machine learning tools can struggle to analyze. Moreover, a compound's molecule can be represented in different ways, its toxicity and side effects can depend on dosage and a human's biological systems (which are assessed by using individual patients' clinical information), while the behavior of a compound can change in patients over time.
This complexity demands models that incorporate the relevant issues and are capable of scenario analysis. Deep-learning models, which typically depend on stochastic analysis with optimization of key parameters, are sensitive to changes to initial assumptions and can offer improved parameter variability. This facilitates their assessment of large numbers of scenarios and adds sensitivity to analysis that can mitigate risks and improve efficiencies--though results remain dependent on input (data) quality, sufficient size, and computational power.
Ethical considerations
Pharmaceutical research's reliance on, often sensitive, patient data raises concerns around the handling of information and its sharing with third parties involved in drug discovery. This could be mitigated through the use of differential privacy systems, which make personal information sources unidentifiable and thus anonymizes data sets while retaining the characteristics that make them useful.
The lack of transparency that is inherent to many AI algorithms' inputs and computations could also raise ethical concerns surrounding diversity and the potential for undesirable effects on trials and clinical studies. Alleviating those worries will require robust AI governance, investment in cyber security systems, communication, and ongoing compliance with evolving regulations and best practice.
The risk of data theft by bad actors is evidenced by past, high-profile cyber attacks including:
- The NotPetya cyber attack on Merck in 2017, which affected the groups manufacturing and supply chain and ultimately resulted in a $1.4 billion settlement with insurers that initially denied coverage to the company.
- A breach of the European Medicines Agency (EMA) in 2020 that enabled cyber criminals to access documents relating to the COVID-19 vaccine developed by Pfizer and BioNtech. The intrusion was potentially motivated by efforts to undermine trust in the vaccine.
- Incidents in 2022-2023, that affected India-based Sun Pharma, Swiss-based Novartis, Germany-based Evotec, UK-based AstraZeneca, and Japan's Eisai.
Workforce management
Increased adoption of generative AI will demand labor force upskilling, particularly within R&D. We expect strong demand for AI experts, including for data engineers who will be integrated into R&D functions to ensure data and model veracity, while professionals whose knowledge spans medicine and computer science could become critical to operations. Management and boards will also require new skills in order to provide appropriate oversight and risk management.
AI's Competitive Advantages Could Have Rating Implications
Generative AI's implementation in the pharmaceuticals sector remains at an early phase and we believe it is unlikely to have a meaningful influence over the next 12-24 months. Yet the investments being made today should offer successful early adopters competitive advantages, notably including in the scope, speed, cost, and success of R&D programs (see chart 1).
Chart 1
The advantages stemming from AI could prove lasting and have the potential to affect ratings. For example, improvements in the identification of molecules for development and pre-clinical research should free time and resources for other profitable or promising projects.
We don't expect AI will necessarily result in more blockbuster drugs--those with over $1 billion of sales in a year. But it should speed treatments to market, with benefits for patients and the market shares of companies. It is notable that AI has particularly been applied to both oncology and neurology R&D, reflecting the pressing need to address patients' often critical health issues and the typically longer-times required to develop therapies and obtain health agencies' approvals.
We think that the deployment of AI could lead to some polarization/specialization in research, but effects on the breadth of medical advancements should be mitigated by companies' differing strategies. And AI should facilitate greater expertise in certain areas, particularly if resulting R&D cost savings facilitate the emergence of specialty pharmaceutical firms with a focus on rare diseases that have modest potential markets.
Beyond R&D, we expect AI will also increasingly be applied to manufacturing, and in particular production lines, where it should improve efficiency and enable resources (currently dedicated to machinery operation) to be redirected toward network oversight and maintenance. That should lead to improved quality-monitoring on production lines, decreased manufacturing downtime, and fewer supply disruptions.
Why The AI Effect Will Be Difficult To Discern
The global pharmaceutical-AI market was valued at $1 billion in 2022, according to consulting firm Boston Consulting Group, but is expected to expand to almost $22 billion by 2027, equating to a compound annual growth rate of about 85%. Despite that increase, the extent to which investment in AI will directly and clearly improve drug discovery, and ultimately contribute to profitability of individual pharmaceutical companies, will be difficult to accurately discern.
That is partly because the resultant gains are likely to be incremental. Drug discovery, with or without AI, will remain a complex and time-consuming practice characterized by experimentation, false starts, and failures. We expect algorithms that effectively and efficiently identify promising molecule combinations will deliver only a slightly lower failure rate, though even a small improvement in speed and efficiency will have an important impact for patients and companies.
AI's ability to expedite drug development will also be hard to measure. The passage of a treatment through discovery, pre-clinical trials, to adoption and marketing currently takes an average of about 10-15 years. Much of that process will not be compressed by AI. Furthermore, if AI enables pharmaceutical companies to pursue treatments for more complex diseases and pathologies, then it may create demand for more resources and extend development timelines.
Pharmaceutical companies also have good reason to seek to obscure the contribution of AI to their drug development. Accurate information could provide competitors with insight into the success or failure of AI programs, enabling them to direct their own spending more efficiently.
What To Watch For Over The Next Five Years
We expect AI's application in pharma will take time to bear fruits, not least due to the industries' long development lead times, heavy regulation, and labor-intensiveness. Furthermore, AI's successful implementation at individual companies is not assured and will be the result of investment, a willingness to adopt new processes, and their ability to manage change.
We expect the greatest benefits from AI projects will accrue to companies that share several traits, including:
Endurance. AI partnerships and projects are increasing in number but remain predominantly early stage and characterized by the exploration of the possibilities of generative AI. They will require the development and curation of large data sets and time to improve their precision.
Scalability. AI investment tends to be narrowly focused and limited in its applications. Generative AI will need to address a wider range of diseases and a greater number of activities in order to significantly contribute to health outcomes and create meaningful value for adopters.
Integration. Generative AI will have to be combined with existing proprietary systems to optimally and efficiently drive improvements in research and manufacturing. Companies that prove adept at that integration will reap the greatest competitive advantages.
Expertise. While we believe that generative AI will improve R&D efficiency, it will not replace fundamental research lead by scientists. Maximum synergies will thus rely on the combination of the technology and specialized human resources that prove capable of working together.
Embracing those factors, and thus harnessing the potential of AI, is a journey that is just starting for the pharmaceutical sector. But the potential for improvement to treatments, processes, and patients' lives is already evident, and suggests that AI will have a significant and lasting role at the heart of the pharmaceutical industry.
Related Research
AI In Healthcare: A Path to Long-Term Immunity, June 25, 2024.
Artificial Intelligence Powering Synthetic Biology: The Fundamentals, June 25, 2024
Machine Learning: The Fundamentals, Nov. 29, 2023
Other Research
Generative AI in the pharmaceutical Industry: Moving from hype to reality, McKinsey & Co., published Jan. 9, 2024
Writer: Paul Whitfield
This report does not constitute a rating action.
| Primary Credit Analyst: | Ihsane Mesrar, Paris (33) 1-4075-2591; ihsane.mesrar@spglobal.com |
| Secondary Contacts: | Guillaume Benoit, Paris + 33 14 420 6686; Guillaume.Benoit@spglobal.com |
| Nikolay Popov, Dublin + 353 (0)1 568 0607; nikolay.popov@spglobal.com |
No content (including ratings, credit-related analyses and data, valuations, model, software, or other application or output therefrom) or any part thereof (Content) may be modified, reverse engineered, reproduced, or distributed in any form by any means, or stored in a database or retrieval system, without the prior written permission of Standard & Poor’s Financial Services LLC or its affiliates (collectively, S&P). The Content shall not be used for any unlawful or unauthorized purposes. S&P and any third-party providers, as well as their directors, officers, shareholders, employees, or agents (collectively S&P Parties) do not guarantee the accuracy, completeness, timeliness, or availability of the Content. S&P Parties are not responsible for any errors or omissions (negligent or otherwise), regardless of the cause, for the results obtained from the use of the Content, or for the security or maintenance of any data input by the user. The Content is provided on an “as is” basis. S&P PARTIES DISCLAIM ANY AND ALL EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, ANY WARRANTIES OF MERCHANTABILITY OR FITNESS FOR A PARTICULAR PURPOSE OR USE, FREEDOM FROM BUGS, SOFTWARE ERRORS OR DEFECTS, THAT THE CONTENT’S FUNCTIONING WILL BE UNINTERRUPTED, OR THAT THE CONTENT WILL OPERATE WITH ANY SOFTWARE OR HARDWARE CONFIGURATION. In no event shall S&P Parties be liable to any party for any direct, indirect, incidental, exemplary, compensatory, punitive, special or consequential damages, costs, expenses, legal fees, or losses (including, without limitation, lost income or lost profits and opportunity costs or losses caused by negligence) in connection with any use of the Content even if advised of the possibility of such damages.
Credit-related and other analyses, including ratings, and statements in the Content are statements of opinion as of the date they are expressed and not statements of fact. S&P’s opinions, analyses, and rating acknowledgment decisions (described below) are not recommendations to purchase, hold, or sell any securities or to make any investment decisions, and do not address the suitability of any security. S&P assumes no obligation to update the Content following publication in any form or format. The Content should not be relied on and is not a substitute for the skill, judgment, and experience of the user, its management, employees, advisors, and/or clients when making investment and other business decisions. S&P does not act as a fiduciary or an investment advisor except where registered as such. While S&P has obtained information from sources it believes to be reliable, S&P does not perform an audit and undertakes no duty of due diligence or independent verification of any information it receives. Rating-related publications may be published for a variety of reasons that are not necessarily dependent on action by rating committees, including, but not limited to, the publication of a periodic update on a credit rating and related analyses.
To the extent that regulatory authorities allow a rating agency to acknowledge in one jurisdiction a rating issued in another jurisdiction for certain regulatory purposes, S&P reserves the right to assign, withdraw, or suspend such acknowledgement at any time and in its sole discretion. S&P Parties disclaim any duty whatsoever arising out of the assignment, withdrawal, or suspension of an acknowledgment as well as any liability for any damage alleged to have been suffered on account thereof.
S&P keeps certain activities of its business units separate from each other in order to preserve the independence and objectivity of their respective activities. As a result, certain business units of S&P may have information that is not available to other S&P business units. S&P has established policies and procedures to maintain the confidentiality of certain nonpublic information received in connection with each analytical process.
S&P may receive compensation for its ratings and certain analyses, normally from issuers or underwriters of securities or from obligors. S&P reserves the right to disseminate its opinions and analyses. S&P's public ratings and analyses are made available on its Web sites, www.spglobal.com/ratings (free of charge), and www.ratingsdirect.com (subscription), and may be distributed through other means, including via S&P publications and third-party redistributors. Additional information about our ratings fees is available at www.spglobal.com/usratingsfees.