We're Not Doing Bets: Pande's AI Biotech Pivot

Former **a16z** partner **Vijay Pande**, who previously managed an impressive **$4 billion** in the firm's healthcare and life sciences practice, made an unexpected move in June last year. He co-founded a new firm, **VZVC**, with longtime investor **Zach Werner**, pivoting sharply to a highly concentrated investment strategy focused on AI-driven biotech, as Pande explained in a recent conversation. This new venture dramatically departs from his previous model, emphasizing a select handful of annual investments rather than dozens, operating without traditional associates, and heavily leveraging AI for its day-to-day operations.

Pande's Distinguished Journey from Academia to Venture Capital

By Decode Today News

“We’re not doing 30 bets a year”: Vijay Pande on betting small after running $4 billion at a16z AI

Pande's career trajectory is notable, initially gaining recognition in academic circles before his abrupt entry into venture capital over a decade ago. As a **Stanford chemistry professor**, he was celebrated for creating **Folding@home**, a pioneering distributed-computing project that harnessed millions of home PCs to form a supercomputer dedicated to disease research. In a significant shift, **Marc Andreessen** and **Ben Horowitz** of **a16z**, after five years of explicitly avoiding healthcare and life sciences, decided to bet on the sector and entrusted Pande with leading the charge. Over the next decade, he scaled **a16z's** commitment into a practice overseeing nearly **$4 billion**, cementing his reputation as a formidable investor in the space.

Understanding the Mechanics of “We’re Not Doing Bets”

The operating philosophy at **VZVC** represents a fundamental departure from conventional venture capital models. Pande articulated a strategic shift from engaging in "30 bets per year" to focusing on "probably five, not a lot of investments – very concentrated." This approach, according to Pande, reflects a profound commitment to each portfolio company, likening the process of adding a new company to "wanting to have another child" rather than merely "adding a Facebook friend." Key characteristics defining **VZVC's** distinct operational and investment model include:

• Concentrated Portfolio: A deliberate choice to make only a handful of highly focused investments annually, enabling deeper engagement and support for each venture.
• Lean Structure: The firm operates with a remarkably small team, primarily Pande and Werner themselves on the investment side.
• AI-Driven Operations: Heavy reliance on AI agents for daily tasks, which, surprisingly, eliminated the initial intent to hire associates, enhancing operational efficiency.
• Long-Term Partnership: A strong emphasis on building enduring relationships with founders, extending ideally beyond a single company's lifecycle.

This model prioritizes hands-on involvement and strategic depth over sheer volume, reflecting a belief in the power of focused resources to drive success.

The Transformative Power of AI in Biology and Drug Development

Pande posits that biology is undergoing a paradigm shift, transitioning from a "science of discovery" to a discipline that is increasingly "engineerable." He highlights how AI and machine learning are pivotal in this transformation, enabling computers to understand intricate biological complexities. This advanced comprehension allows for more precise identification of drug targets for specific diseases, facilitates the design of novel therapeutics, and even aids in the critical, most expensive phase of clinical trials. The integration of AI promises to streamline drug development, moving away from serendipitous discoveries towards a more deliberate, design-driven process.

Revolutionizing Clinical Trials and Precision Medicine

Clinical trials remain the most capital-intensive phase of drug development, often costing **hundreds of millions of dollars** to run, which significantly contributes to high drug prices. The success rate for a drug progressing from its first trial to the end of the third is a mere **20%**. Pande explains that failures often stem not from biological missteps, but from the limitations of animal models (like mice), which are not sufficiently predictive of human responses. AI models, while not perfect, offer a substantially improved predictive capability over animal models, marking a critical advancement in **cost efficiency** and **investment yield** for biopharmaceutical companies. Beyond trial efficiency, AI is also driving **precision medicine**, a concept focused on tailoring medical treatments to individual patients. Traditionally, doctors must often guess diagnoses and prescribe drugs through trial and error, as seen in conditions like cancer. AI's ability to analyze individual biological data, moving beyond population averages, allows for personalized understanding, ensuring "the first drug was the right one" for the individual. This shift is further fueled by advances in measuring various biological markers like proteomics, which provide more relevant, real-time insights into a body's current state compared to a static genomic blueprint. Moreover, automation in robotic measurements works hand-in-hand with AI, accelerating both biological and chemical research over the past decade.

Navigating the Unique Data Landscape of AI in Biotech

A significant challenge for AI in biology, Pande notes, is the scarcity of readily available, scrapable data compared to other domains like text processing. Unlike the vast internet repositories that fuel large language models (LLMs), biological data is often proprietary and contained within "walled-off datasets." This creates a fragmented landscape where companies cannot easily train on shared information or distill models across different datasets. Pande recognizes an echo of this problem in traditional medicine, where doctors often operate in competitive or territorial silos, making integrated care challenging. He finds **AI infrastructure** particularly intriguing in its potential to overcome these human silos. In principle, AI can act as a "specialist in everything," integrating diverse medical insights across oncology and endocrinology, for instance, to see patterns and connections that no single human physician or team could identify. This represents a substantial opportunity for enhanced **enterprise integration** in healthcare.

The Rise of Biological Foundation Models and Open-Source Impact

Despite the current data fragmentation, Pande observes a nascent trend towards building "atlases of biological information," which, from a technological standpoint, are typically foundation models. He anticipates that as these become more common, they could mirror the success of open-source LLMs that have demonstrated competitive performance against corporate counterparts. Open-source foundation models in biology hold the potential for broad impact by fostering greater data sharing and collaborative innovation across the field, even as founders and investors naturally seek to protect their specific findings. Such developments would be critical for advancing **compliance security** and **data integrity** across shared platforms.

Identifying High-Integrity Founders for Long-Term Partnerships

At **VZVC**, Pande's investment focus primarily lies in **AI for healthcare delivery** and **AI for clinical trials**—areas he extensively explored at **a16z**. When evaluating founders, his paramount criteria revolve around trust and integrity. He seeks individuals who "do what they say they're gonna do" and are committed to relationships spanning "5, 10 years plus into, ideally, their next company." Pande prioritizes working with founders who embrace a long-term perspective and are motivated by the question, "how do we win together?" rather than solely competing against others. He mentions his involvement with **Genesis Therapeutics**, which originated from his Stanford lab, and **Insitro**, a drug-discovery company founded by his former Stanford colleague **Daphne Koller**. He is also incubating a company with a founder he has known for two decades, underscoring his emphasis on established trust.

Insights from a Decade of Biotech Investing

Reflecting on his investment career, Pande notes that a decade ago, there was significant resistance to the idea of AI, machine learning, and technology playing a substantial role in medicine and biology. Witnessing the widespread acceptance of these concepts today has been "very fulfilling." His most significant learning, however, is the enduring importance of **go-to-market strategy**. Despite the allure of cutting-edge technologies, Pande stresses that "it really always comes back to go-to-market." He advises founders, particularly those from scientific or product backgrounds, to dedicate their brilliance and creativity to the go-to-market side, considering it "at least as hard or harder than the technology side."

VZVC's Differentiated Competitive Edge

The concentrated investment model of **VZVC** gives the firm a unique competitive position. Pande explains that they are typically "not trying to compete for a hot round." Instead, "people make room for us" because of the distinct value proposition offered by Pande and Werner's hands-on approach and expertise. This allows them to secure deals outside the highly competitive Series A or Series B rounds. Pande draws inspiration from investors like **Antonio Gracias** at **Valor**, known for the **SpaceX** deal, and **Thrive** for its concentrated portfolio strategy, while acknowledging the foundational influence of **a16z** in his investment DNA.

Demystifying AI Hype in Biotech: The Data Conundrum

While acknowledging that AI possesses an undeniable ability to uncover insights beyond human capacity, Pande cautions against the overhyped notion that "AI is going to cure all everything." His hesitancy does not stem from a doubt about AI's potential, but rather from a pragmatic assessment of **data availability**. He emphasizes that **market valuation** for AI often overlooks the crucial factor of data. Unlike LLMs, which thrive on an abundance of data, AI in biology faces a stark reality: when the foundational data simply isn't there, AI cannot magically solve the problem. This highlights the critical need for continued effort in biological data generation and aggregation to fully realize AI's promise in medicine.

More coverage from Decode Today

• More AI news from Decode Today
• More World news from Decode Today
• More Technology news from Decode Today
• More Business news from Decode Today
• More Automotive news from Decode Today