Traditionally, developing a single drug takes between 10 and 15 years, at an estimated cost of billions of dollars, while the success rate of drug candidates reaching final approval does not exceed 10% of all candidates entering the trial phase.

This slow and costly equation is now undergoing a fundamental transformation with the entry of artificial intelligence into drug discovery and vaccine design, as computer models can now shorten stages that used to take years into weeks or just a few days.

How is AI accelerating drug discovery?

DeepMind's AlphaFold system is one of the most prominent breakthroughs in this field, as it managed to predict the three-dimensional structure of more than 200 million proteins, almost all known proteins to science. Since understanding the shape of the target protein is a fundamental step in designing a drug that binds to it precisely, this achievement shortened a stage that previously required months of complex laboratory experiments such as X-ray crystallography.

The company spun off from this technology, Isomorphic Labs (a subsidiary of Alphabet, Google's parent company), has leveraged this advance to develop the first drug fully designed through protein structure prediction that enters human trials, which specialized coverage described as a paradigm shift from theoretical prediction to actual therapeutic application.

First fully 'AI' drug

Insilico Medicine is a leading example of integrating AI into the entire discovery chain, from target identification to drug molecule design. The company developed its drug INS018-055 for idiopathic pulmonary fibrosis, moving from target discovery to the start of human clinical trials in less than 30 months, an unprecedented speed compared to the usual industry average.

According to the company, this drug is the first 'true AI drug' combining a therapeutic target and molecular discovery fully identified through generative models, and it is currently in phase II clinical trials.

The circle of leading companies in this field is expanding to include Schrödinger, whose physics-based simulation software serves 18 of the top 20 global pharmaceutical companies, and Recursion Pharmaceuticals, which entered a partnership with Swiss company Roche for $150 million upfront and up to $12 billion in future milestone payments, indicating the growing investment confidence in this sector.

Coronavirus vaccines: A model of record speed

The COVID-19 pandemic provided the clearest practical evidence of AI's ability to accelerate vaccine development. According to Amazon Web Services, Moderna completed the full genetic sequencing of the COVID-19 vaccine in just two days using machine learning, followed by readiness of the first clinical batch within only 25 days of sequencing, an unprecedented record time in the history of vaccine manufacturing.

For its part, BioNTech, in partnership with Pfizer, focused on the first part of the process, namely antigen design and prediction of immune response through computational methods, leveraging prior experience in personalized cancer vaccines based on neoantigens.

BioNTech later strengthened its capabilities by acquiring the AI-specialized company InstaDeep for $440 million, to integrate its technologies into its mRNA vaccine development platform.

Specialized sources estimate that the Moderna and BioNTech/Pfizer vaccines together saved between 15 and 20 million people worldwide, illustrating that the impact of AI-driven acceleration did not remain confined to laboratories but directly affected global public health.

Generative AI design of antigens

AI applications in vaccines go beyond responding to emergency epidemics, as specialized language models in protein sequences are used to explore new antigen designs that go beyond what exists in nature, opening the door to more precise vaccines and broader immune coverage.

Specialized platforms combine these models with computational immunology to design multi-epitope vaccines with higher immunological efficiency. These efforts also include cancer vaccines based on neoantigens, for which Moderna has shown promising clinical results in advanced stages of skin cancer treatment.

Challenges and scientific limitations

Despite these achievements, industry experts warn against overestimating AI's impact on the entire economic and temporal equation of drug development. The technology so far clearly shortens early discovery stages, but it has not yet proven its ability to shorten advanced clinical trial phases (Phase II and III), which remain governed by the same biological rules regarding safety and efficacy in humans.

Scientific reviews also indicate that some molecules discovered through AI record success rates similar to those discovered by traditional methods, meaning that the clearest benefit so far lies in speed and cost rather than in the final success rate.

On the regulatory front, the U.S. Food and Drug Administration is finalizing official guidance for the use of AI in drug development by 2026, requiring companies to submit detailed reliability assessment plans for high-risk models. Meanwhile, the European AI Act will come into force in August 2026 with provisions that may classify some AI applications in drug development as high-risk categories requiring additional oversight.

Today, artificial intelligence represents a truly transformative tool in the early stages of drug discovery and vaccine design, from predicting protein structures to designing antigens and accelerating genetic sequencing for mRNA vaccines.

But the full success of this technological revolution will remain contingent on the results of advanced clinical trials in the coming years, which will determine whether AI-designed drugs actually offer greater therapeutic benefit, or whether their most prominent impact remains confined to saving time and cost—an important achievement in itself, but it does not replace the rigorous scientific verification upon which the pharmaceutical industry is based.