After developing the Folding@home project at Stanford, Vijay Pande grew a16z’s healthcare and life sciences investment business and managed a practice worth approximately $4 billion. Pande, who left his position in June, founded VZVC with Zach Werner. The new firm makes approximately five focused investments per year instead of 30; its investment team has only two partners, and it uses AI agents in its day-to-day operations.
According to Pande, AI can help identify disease targets, design drugs and support clinical trials in drug development. However, the likelihood of a drug successfully progressing from the first trial to the end of the third trial is only 20%; clinical trials can cost hundreds of millions of dollars. Pande says AI could produce better results than animal models and advance personalized medicine.
Pande said that the inability to easily collect biological data from the internet limits advances in AI, and that companies mostly create their own closed datasets. In contrast, he said the proliferation of biological knowledge atlases and open-source foundation models could have a broad impact on the field. Pande emphasized that AI cannot solve everything on its own; the main limiting factor is often the lack of data.
Why it matters
VZVC’s structure, built around a small team and a limited number of concentrated investments, shows that specialization in healthcare and life sciences is being pursued alongside operational efficiency. The fundamental problem behind this approach is that, because of the high costs and low probability of success in drug development, technical progress does not directly translate into clinical outcomes. The use of AI in disease targets, drug design, and research processes therefore stands out as not merely a software issue, but also an issue of access to biological data. The fact that data remains closed within companies limits the transfer of the tools developed to different research efforts. The proliferation of open biological atlases and foundation models could change this fragmented data structure; however, their impact on clinical success remains an open question.