8.9 billion raised, zero drugs approved: the bottleneck of pharmaceutical AI is not chemistry

An editorial argues that AI platforms were built for speed, not compliance: the delay is paid for at the time of the regulatory file.

In an op-ed published on August 8, 2026, Moe Alsumidaie, editor-in-chief of The Clinical Trial Vanguard, compares two figures: the $8.9 billion raised in 2024 by machine learning-assisted drug discovery companies, spread over 264 financing deals, including $5.6 billion earmarked for biotechnology, and the number of FDA approvals obtained to date by an AI-based drug: none.

The author does not dispute the preclinical gains. He cites the case of Insilico Medicine, whose candidate for idiopathic pulmonary fibrosis was reportedly conducted to the end of preclinical in eighteen months for $150,000. His argument is about what happens next: the benefits of AI fade as soon as human trials are entered, and the real obstacle lies less in chemistry than in regulatory preparation — documentation and data architecture required for FDA filing.

The criticism is structural: most platforms were designed for speed, not for compliance, leaving it to sponsors to rebuild a posteriori a regulatory infrastructure that should have been put in place upstream. Finally, the author warns that the first approval, probably obtained under orphan drug status in a rare disease, will say nothing about the ability of the method to scale up on large therapeutic areas and ordinary regulatory pathways.