From managing $4 billion to placing just a handful of bets: Vijay Pande's new approach to AI in biotech
Former a16z investor Vijay Pande, who oversaw a nearly $4 billion portfolio, has launched a small firm called VZVC that makes only about five concentrated investments a year. In an interview, he discusses how AI is reshaping drug development and why biological data remains the field's core challenge.
More than a decade ago, Stanford chemistry professor Vijay Pande — known for creating Folding@home, the distributed-computing project that harnessed home PCs for disease research — joined venture firm a16z to lead its bets on healthcare and life sciences. Over time, he built that practice into a portfolio worth nearly $4 billion. Last June, Pande unexpectedly left a16z to co-found a much smaller firm, VZVC, alongside longtime investor Zach Werner.
A concentrated approach
Unlike traditional funds that close dozens of deals annually, VZVC plans to make only about five investments a year. The firm has no associates, relying heavily on AI tools for day-to-day operations instead. Pande compares adding a new company to the portfolio not to a quick decision but to deciding to grow one's family.
AI's role in drug development
Pande explains that AI and machine learning now help identify drug targets for specific diseases, assist in designing the drugs themselves, and even support clinical trials — the most expensive phase of the process. He notes that only 20% of drugs successfully complete all three clinical trial phases, and failures are typically not due to researcher error but because experiments rely on animal models that don't predict human outcomes well. AI models, he says, can meaningfully outperform animal models in this regard.
The data challenge
One key difference between AI in biotech and, say, language models is that biological data can't simply be scraped from the internet — each company ends up building its own walled-off dataset. Still, Pande sees a growing trend toward building broader biological information "atlases," or foundation models, which he expects could eventually follow a path similar to open-source language models.
Pande also stresses that when choosing founders to back, mutual trust and long-term thinking matter more to him than the technology itself — in his experience, a company's success often hinges less on the technology and more on successfully bringing it to market.
