Vijay Pande, who had managed a 16z nearly $4 billion bio-medical investment, has now shifted its focus to a smaller fund. According to him, the new institution VZVC makes only a few concentrated investments each year, no longer pursues large-scale projects and relies more on AI tools for daily operations.
Pande, formerly Professor of Chemistry at Stanford University, has dominated the distributional computing project Folding@home. After joining a16z, he led the institution into the medical and life sciences investments and expanded its operations for over a decade. Now that he has left the big platform, he and Zach Werner, long-term investors, have set up new funds, with a clear contraction in strategy and a greater emphasis on high-confidence projects.
The investment tempo turned to concentration
According to Pande, the new fund would not make dozens of transactions a year, but would only select a small number of entries. He attributed this change to the current market environment and to the adjustment of the way individuals judge, i.e., to concentrate time and resources on a few long-term projects rather than to decentralize.
He mentioned that the new team did not have a large number of assistants or analysts in the traditional sense and that part of the day-to-day work was done by AI. This also reflects the fact that some early investment institutions are trying to operate with a lighter organizational structure.
Attention AI Medical Services and Clinical Trials
In the direction of investment, Pande is currently looking at two types of opportunities: AI-driven medical services and AI applications in clinical trials. In his view, drug development had relied heavily on experience and error, and AI and machine learning were helping researchers to identify drug targets earlier, design candidate drugs and improve clinical trial processes.
He also pointed out, however, that clinical trials were not quickly made low because of AI. While the time to enter the clinical phase is being shortened, a trial may cost hundreds of millions of dollars, while the proportion of drugs from phase one to phase three and ultimately successful remains low. According to him, many of the failures were not due to operational errors by researchers, but rather to the limited ability of animal models to predict the human body.
According to Pande, as long as the AI model is more effective than the traditional animal model in predicting human reactions, there will be more significant efficiency gains in the industry. This is one of the reasons why he continues to bet in that direction.
Biological data remains an industry bottleneck
Unlike Internet textual data, data in the field of biomedicine are difficult to capture and harmonize. According to Pande, this has led many companies to establish their own closed data systems, and data barriers have become a real constraint on AI biomedical development.
At the same time, he indicated that a larger range of bio-information “species” and base models were emerging in the industry. If such open models continue to develop, the future, like the large open source model, may complement enterprise self-building systems and expand AI's coverage in the field of medicine.
In the choice of the founders, Pande said she valued long-term cooperation and credibility. He is also currently incubating new projects and continues to focus on his familiar AI pharmaceutical and medical technology entrepreneurs.