AI slashes drug discovery to 9 months, Insilico CEO says China labs cut years

Insilico Medicine is hitting drug candidate milestones in as little as nine months—a timeline that makes the industry’s standard four-and-a-half years look glacial. CEO Alex Zhavoronkov told Reuters the Hong Kong-listed company now routinely moves from target discovery to preclinical candidate in about 13 months. The fastest program got there in nine.

The secret isn’t just the generative AI modeling done in Montreal and Abu Dhabi. It’s the Shanghai lab work. Zhavoronkov pointed to China’s research infrastructure and regulatory speed as the real accelerant. While Western companies grind through years of early-stage work, Insilico’s teams in China handle automated biological sampling and compound screening, synthesizing and testing between 60 and 200 molecules per program. “We now compete with Chinese pharmaceutical companies on timelines, and with traditional biotechnology companies in the West on novelty,” he said.

That hybrid model has produced 31 preclinical candidates since 2021. Thirteen have secured investigational new drug clearances. A Phase III trial for Rentosertib, an oral drug for idiopathic pulmonary fibrosis, is now registered in China. The 320-patient study marks a critical test of whether AI-discovered molecules can survive the gauntlet of late-stage clinical evidence. A proposed alliance with Taiwan’s Bora Pharmaceuticals could exceed $2.5 billion if fully implemented.

Yet the China advantage comes with strings. More than 90% of Insilico’s revenue flows from Western pharma companies, including Eli Lilly and Takeda. Zhavoronkov said China’s national insurance system pays too little for novel drugs, and geopolitical concerns mean most of its software isn’t sold there. The company is betting it can keep one foot in China’s fast-track research ecosystem while the other stays firmly planted in Western licensing deals. Whether that balancing act holds may determine if these compressed timelines translate into an enduring business.