Vijay Pande Reflects on Cautious Investment Strategies Following His Tenure Managing $4 Billion at a16z

Vijay Pande was primarily known in academic circles before gaining prominence in the investment sector. This dynamic changed dramatically about a decade ago when Marc Andreessen and Ben Horowitz, who had previously avoided healthcare and life sciences, chose to invest in this domain and appointed Pande to lead their efforts. At the time, he was a chemistry professor at Stanford, widely recognized for establishing Folding@home, a distributed computing project that transformed countless home computers into a supercomputer for disease research. Over the next ten years, he grew a16z’s investment practice to manage nearly $4 billion.

Unexpectedly, in June of last year, Pande made the decision to leave all of this behind in order to start a much smaller venture. His new firm, VZVC, co-founded with experienced investor Zach Werner, aims at making a limited number of concentrated investments each year rather than spreading resources across many projects, operates without associates, and extensively utilizes AI for its daily functions.

To explore Pande’s notable transition, we discussed his decision this week to focus on a select few concentrated investments rather than diffusing resources in today’s market. We also examined one of the fascinating challenges in AI-driven biotech: unlike text, biological data cannot be sourced from the internet, which leads companies to develop isolated datasets. What are the repercussions for AI’s potential advancements in medicine, and who will benefit from these innovations?

This conversation has been edited for length and clarity. You can also listen to the full discussion (below).

You’ve indicated that biology is evolving from a “science of discovery” to a domain that can be engineered. What does this involve?

Historically, drug development involved an element of chance. The significant shift now is that AI and machine learning empower computers to process complex data to identify drug targets for specific diseases, facilitate drug creation, and even assist in clinical trials—the most expensive stage of the process.

I had the impression that clinical trials were becoming cheaper due to the use of synthetic data, thereby requiring fewer participants.

That’s definitely an optimistic outlook. While the costs and duration to initiate clinical trials have decreased, particularly with AI’s assistance, conducting a trial can still cost hundreds of millions of dollars, explaining the high prices of new drugs. The likelihood of a drug progressing successfully from the initial trial through the third trial is only 20%. When 80% fail and each trial is expensive, the amortized cost escalates quickly. Most failures are not due to biologists’ errors, but because these drugs are tested on animal models, such as mice, which often fail to predict human outcomes accurately. Although AI models are not perfect, they tend to outperform animal models, and reaching that benchmark opens up exciting opportunities.

[The next question is]: Is the drug the right one for me?

You’re referring to personalized medicine…

The precise term is precision medicine. When patients consult a physician for complex conditions, doctors often have to make educated guesses because of insufficient information. They prescribe a medication, and if it fails, they try another one, and so forth. This is especially common in cancer treatment and various other areas. It would be immensely beneficial if the initial prescription were the correct one. Currently, blood test values are compared to general population averages, but they ought to be individualized to assess what’s unusual for each patient. We are starting to improve our understanding of what therapies are most effective for each individual.

Do you believe this progress has been gradual, or has it accelerated in recent times?

I think it is the result of multiple converging factors. For a long time, precision medicine was mainly centered on genomics. However, a human genome acts as the initial plan for a “house” that changes over time. Today, various other measurable elements, such as proteomics, offer more relevant insights into diseases and the current health status. Furthermore, advancements in robotics and automation are closely linked to AI, resulting in natural synergies.

In the last decade, we have witnessed consistent advancements in applying AI to biology and chemistry. The biological focus is on treating diseases, while the chemistry focus is on developing drugs targeting specific proteins. There has indeed been remarkable progress over this ten-year period.

You mentioned that biology is one of the few sectors where data cannot simply be harvested from the internet. What implications does this have for the field’s growth?

It implies that there is no common dataset for organizations to collaboratively train models. This lack creates distinct challenges from an AI standpoint.

Doesn’t that illustrate a recurring issue in medicine, where physicians frequently operate in competitive silos?

You’ve touched on a critical issue. When someone has a specific cancer that spans oncology and endocrinology, specialists in these fields often do not collaborate effectively. The exciting element of AI is its ability to synthesize knowledge across disciplines, simulating the cooperation of multiple top-tier physicians working together.

Is there sufficient data sharing for this vision to become a reality? While I understand the inclination of founders and investors to safeguard their discoveries…

I see a notable shift towards creating extensive biological information atlases, which are often built as foundational models. As these grow in prevalence, I expect a similar trend to what we’ve witnessed with open-source large language models outpacing corporate alternatives: open-source foundational models in biology will have a significant impact.

You have associations with Genesis Therapeutics, which spun out of your Stanford lab, and Insitro, the drug-discovery firm established by Daphne Koller, your former colleague. Additionally, you’re incubating a venture with a founder you’ve known for 20 years. What qualities do you look for in founders, and which sectors are you concentrating on?

I’m primarily focused on two areas: AI for healthcare delivery, which I explored deeply at a16z, and AI for clinical trials.

Trust is fundamental for me when evaluating founders. I seek partners with integrity, those who keep their commitments. I aim to build a relationship that spans 5 to 10 years, ideally nurturing them into their next business venture. I favor working with individuals who perceive success as a shared achievement rather than a competitive contest.

What successes and challenges have you encountered in your investment journey so far?

When I initially began discussing AI, machine learning, and their roles in medicine over a decade ago, the general response was skepticism, with many expressing doubt about its feasibility or benefits. Seeing the shift in perceptions has been fulfilling.

It took time for me to realize that no matter how appealing advanced technologies might be, market strategy remains vital. I often counsel my founders, particularly those with scientific or product backgrounds, to apply their creativity and intelligence toward market strategies since navigating the go-to-market landscape can be as challenging, or even more so, than the technology itself.

Could you describe how you are structuring this new firm differently from your role at a16z?

We’re taking a notably different approach now… VZ is referred to by the initial letters of my name, Vijay, and my co-founder, Zach Werner—he is the “Z.” We purposely maintain a small team… on the investment side, it consists solely of the two of us. Although we initially intended to hire associates, our developed systems made that redundant.

How do you define “concentrated” in your investment philosophy?

We do not pursue 30 investments per year… we’re aiming for approximately five concentrated investments. Adding a new company to our portfolio is similar to adding a child to a family—it’s a substantial commitment on our part.

Given this structure, who are you up against for deals?

Interestingly, with this model, we often do not compete for the most competitive rounds—investors usually welcome us aboard. This differs from the usual rat race for Series A or B investments. Frequently, investors appreciate our involvement thanks to our hands-on approach and the distinct value Zach and I offer. I have great respect for individuals like Antonio Gracias at Valor for his enduring contributions, and I am inspired by what Thrive has accomplished with a more concentrated portfolio as well. Naturally, my experiences at a16z continue to shape my thinking, but these other cases have expanded my perspective.

What elements of AI and biotech do you believe are currently overhyped?

The reality is that AI can indeed reveal insights beyond human capabilities. The challenge arises when claims are made that AI will solve all problems. The hesitation isn’t because of doubts about AI, but rather concerns surrounding data quality. Large language models thrive because they have vast datasets to learn from. When required data is unavailable, AI cannot simply fill that void.