AI Drug Development Boom Meets Sober Capital: A Veteran Biotech Investor’s Brake Signal

After artificial intelligence entered new drug R&D, its most compelling promise has always been to “find drugs faster.” But in the biotech industry, speed does not equal success; from targets, molecular design, and animal testing to human clinical trials, every checkpoint can stop a seemingly attractive candidate. According to The Wall Street Journal, Sofinnova Investments managing partner Jim Healy remains reserved about AI drug discovery. What gives his remarks weight is not that they reject AI, but that they remind the market: if algorithms cannot raise the probability of approval, speed itself may simply mean reaching uncertainty faster.

Sofinnova is a long-established biotech venture capital firm that was already investing in early biotech companies such as Genentech in the 1970s. Healy is not opposed to artificial intelligence; the report notes that AI is already being used in diagnostic and clinical-development settings within Sofinnova’s portfolio, such as helping interpret molecular diagnostic data, or identifying suitable participants in clinical trials and checking inclusion and exclusion criteria. What these uses have in common is that they are close to data interpretation and process management, where results are easier to test through accuracy, efficiency, or recruitment quality.

More controversial is AI-driven drug discovery itself. Companies in this category typically claim they can use models to predict protein structures, screen targets, and generate or optimize compounds, allowing research teams to obtain more candidate molecules in a shorter time. Healy’s concern is that producing more molecules or shortening early exploration time does not yet amount to improving clinical success rates; if candidate drugs still fail because of toxicity, insufficient efficacy, or incorrect patient stratification, investment returns and patient benefit may not necessarily improve.

This skepticism is not fear of new technology, but comes from cycles the biotech industry has seen repeatedly. Combinatorial chemistry and proteomics were also once expected to greatly improve the efficiency of new drug R&D. They did bring important scientific tools, but they did not simply rewrite the overall success rate of drug development. AI may be more powerful than those technologies, but publicly comparable long-term evidence remains limited for now; many AI-designed drugs are still in early clinical or preclinical stages, and there is still a time gap before they can prove they are “more often successfully approved.”

Background Context

Drug R&D in recent years has been driven by multiple tools at the same time: organ chips are attempting to make preclinical models closer to the human scale, molecular diagnostics and bioinformatics are helping with patient stratification, and generative models are being placed into imaging, text, and molecular-design workflows. These technologies do not replace one another; each is trying to fill a different gap in the R&D chain. For AI drug discovery, the most rigorous question remains: can the biology predicted by the model hold true in real patients?

The report also mentions that Healy focuses investment on clear biology, measurable patient benefit, and strong management teams; he also believes that sources of innovation are becoming increasingly globalized, and that China is not only providing follow-on products but is also beginning to become a source of more original assets. This means U.S. biotech companies must look more carefully at global competition and licensing opportunities when evaluating pipelines.

Therefore, this signal is not a verdict that the AI drug-development wave is receding, but a reaffirmation of investment discipline. If AI can produce verifiable results in patient selection, trial design, or molecular optimization, it may still become an important tool; but for serious new drug development, the final measure will not be how new the model is or how quickly compounds are generated, but whether more reliable evidence can bring effective drugs to patients.