Editor's NoteArtificial intelligence (AI) is fundamentally upending the pharmaceutical and biotechnology markets. NVIDIA's decision to team up with global big pharma Eli Lilly and invest 1 billion dollars (approximately 1.499 trillion won) in AI-driven drug discovery is emblematic of a paradigm shift that is rewriting the history of the pharmaceutical and biotechnology industries. The territory of AI, once limited to candidate compound identification, is now expanding throughout the entire drug development cycle—including sophisticated clinical trial design, optimal patient cohort selection, and real-time manufacturing process control—fundamentally reorganizing the economics of the industry itself. The time required to derive candidate compounds has been reduced from an average of 3 to 4 years to as little as one year. With 3D understanding of target proteins and enhanced toxicity prediction accuracy, AI has increased the success rate of Phase 1 clinical trials from the typical 40–65% range up to 80–90%. The efficiency of AI in reducing early-stage failure risks—where astronomical sunk costs are common—offers the only viable breakthrough for companies striving to cross the so-called "valley of death" in drug development. Through the special feature series "K-Bio on the AI Test Bench," The Asia Business Daily examines the rapidly evolving landscape of global drug development in 2026, now defined by a "race for speed," and explores the path to survival and growth for Korea's biotech industry.

The expectation that artificial intelligence (AI) will transform the drug development timeline is no longer an exaggeration. AI's efficiency in candidate compound discovery, protein structure prediction, and drug design phases has already been demonstrated. However, feedback from the field tells a slightly different story. While research using AI to explore candidate compounds is expanding rapidly, this momentum has yet to fully carry over into the industrialization stage.


The prevailing assessment is that domestic AI-driven drug development remains largely focused on the "front end." AI utilization has proliferated during literature exploration, target validation, candidate structure generation, and initial filtering stages, but it is still uncommon for this process to extend seamlessly into cell and animal testing, clinical strategy, and regulatory strategy as an integrated development loop.

[K-Bio on the AI Test Bench] ④Behind the Ambitious Blueprint, Bottlenecks on the Ground View original image

While domestic AI drug development still lingers at target validation and initial filtering, major global AI drug discovery companies have already advanced 32 pipelines into clinical trials. An industry insider said on March 18, "AI certainly plays a decisive role up to the stage of generating hundreds of thousands of structures and selecting the most promising candidates. However, when it comes to experimental validation or development strategy, we inevitably end up relying on human and organizational capabilities."


The industry points to a shortage of AI specialists as a key reason for this phenomenon. AI-driven drug development requires talent who can simultaneously understand biology and AI—a fusion skill set—but many believe these two domains remain separate on the ground. Instead of receiving systematic training in AI utilization, many bio researchers still resort to online lectures or public documents for self-study. A pharmaceutical industry official commented, "AI engineers often lack knowledge of biology, and bio researchers find it difficult to master AI deeply. Ultimately, the biggest problem is the shortage of personnel who understand the languages of both fields."


Research infrastructure is also cited as a major bottleneck. AI-driven drug development requires repeated learning and simulation of protein, genomic, and structural data, making large-scale computational resources indispensable. This is why global companies tie their AI competitiveness directly to capital and computing infrastructure. Riding this trend, the global AI drug development market is expanding rapidly with an average annual growth rate approaching 30%. In contrast, the domestic research environment is criticized for lacking the organizational structure and systems to operate graphics processing units (GPUs) over the long term, beyond mere procurement.


Kyuri Kim, Professor of Biological Sciences at KAIST, said, "The number of GPUs is important, but having the infrastructure to manage them reliably and utilize them for research is even more critical."


The structure for utilizing data is also identified as a challenge. The performance of AI for drug development ultimately depends on both the quality and precision of the data used for training. Bio research cannot be advanced using only textual information. Multimodal data, integrating genomic information, protein sequences, tertiary structures, and clinical records, is essential.


Tae-hyung Kim, CEO of Bionexus, stated, "Biology is so complex that text data alone is insufficient. Only by training on genetic, protein, structural, and clinical data together can meaningful results be achieved." As a result, there is a growing awareness that, to overcome limited resources and data, collaboration between large bio companies with capital and production infrastructure and AI biotech firms with technological capabilities is becoming a prerequisite for survival in Korea.


Regulation is also cited as a variable determining the pace of industrialization. Although AI is expanding its role in various areas such as candidate compound design and clinical trial design, regulatory standards for evaluating these applications remain at an early stage. The Korea Biotechnology Industry Organization pointed out in last year's policy report, "Industrialization Strategy for AI-Based Drug Development," that a "Good Machine Learning Practice (GMLP)" guideline should be established for the development, validation, and operational management of machine learning models used in AI drug development.



GMLP, similar to GMP for pharmaceutical manufacturing quality control and GCP for clinical trial management, refers to a quality management system that ensures AI algorithms operate safely and effectively in medical and drug development settings. If an AI algorithm is trained on biased data or contains critical errors when selecting target compounds, this can immediately escalate into a life-threatening safety issue for patients. This is why regulatory authorities must have clear standards to evaluate the reliability of both the data and the AI models themselves when reviewing AI-driven drug approvals.


This content was produced with the assistance of AI translation services.

© The Asia Business Daily. All rights reserved. Unauthorized AI training and use prohibited.

Today’s Briefing