AI’s most resilient players rely on slow science and validation as Big Tech rewrites the rules

In AI’s gold rush, the most enduring value is the kind that takes the longest to prove. As Microsoft absorbs agentic tools into its platform, access-layer companies are collapsing. But in drug discovery, slow validation is a shield. Value isn’t claimed; it’s proven in the lab.

Alex Zhavoronkov is the Founder and CEO of Insilico Medicine, a clinical-stage biotech company using AI for drug discovery. Unlike startups chasing short-term traction, Insilico bets on value that can only be validated through years of lab testing, and that patience is proving to be a survival advantage.

Validation as insulation: "Experimental validation is more important than speed, reach or access to compute," Zhavoronkov says. "In drug discovery AI, many companies focused on chemistry and deep biology are reasonably safe." That slow, rigorous feedback loop insulates them from the platform shifts that can destroy faster-moving peers.

Death of the middleman: "Startups that were operating in purely agentic space, creating access apps, or providing other middleman services between AI and the ultimate customer just got destroyed at Build 2025," says Zhavoronkov. For those companies, value was tied to access and interface, functions that Microsoft has now made "seamlessly available and integrated" into its ecosystem. The result: "massive value destruction."

A joke with teeth: "Some people jokingly say that if Google Calico and Google DeepMind had a baby who went on to became the MMA champ in drug discovery, it would be Insilico Medicine," he says. Behind the humor is a hard truth: success in this space demands the foundational science of Calico, the AI horsepower of DeepMind, and the stamina to survive the long haul. That kind of resilience can’t be out-integrated.