Vijay Pande was once better known in academia than investing. That changed when he grew a16z’s bio practice to nearly $4 billion. In June last year, he left a big platform to launch a smaller firm, VZVC, with Zach Werner. The new model makes a few concentrated bets each year, has no associates, and leans on AI for daily operations. What does that mean for biotech, data, and founders? And how does it change precision medicine and clinical trials?
Why did Vijay Pande move from a big platform to a small firm?
He made a deliberate pivot. Pande left a large practice to build a compact firm, VZVC, with Zach Werner. It focuses on a handful of yearly bets, runs without associates, and relies on AI “agents” in workflows. The pace, depth of engagement, and founder partnership model are different by design.
They are not doing dozens of deals per year. Think about roughly five. Adding a company is not like “adding a Facebook friend,” but more like deciding “are we ready for another child?” That metaphor signals accountability and depth. The team wants to engage deeply and for the long haul, not spread thin.
In this setup, they rarely chase “hot rounds.” Pande says people typically “make room” for them. The reason is what he and Werner can do hands-on. Founders want investors who truly engage and help, not just another logo on the cap table.
The name is symbolic, too. “VZ” stands for Vijay and Zach. The team is intentionally small on the investing side. An initial plan to hire associates was shelved because in-house AI agents absorbed much of the routine. That is part of the thesis — a light structure and high speed.
“We’re intentionally really quite small… on the investment side, it’s really just the two of us.”
Is AI really turning biology from discovery to engineering?
Yes, and the shift is already visible. AI and machine learning wrap their understanding around very complex systems. They help choose the right targets for specific diseases, design and make the drugs, and now support clinical trials, which are the most expensive part of the process.
The myth of “cheap clinical trials,” however, remains aspirational. Synthetic data is more a hope than a current reality. Time and cost to reach trials are shrinking, especially with AI. Yet a trial can still cost hundreds of millions of dollars. That explains expensive drugs and rapid capital amortization.
The probability of successfully moving a drug from phase one to the end of phase three is just 20%. Eight out of ten typically fail. Often the issue is not that biologists erred. It’s that many experiments start from animal models that poorly predict humans. Once AI models cross that “animal-model bar,” things get exciting.
Then comes another question: “Is this drug the right drug for me?” That’s the realm of precision medicine. Doctors often have to guess and try multiple options in sequence. We would all be better off if the first choice worked. AI is helping move care toward that match.
“The AI model is not going to be perfect, but it’s going to be way better than any animal model.”
What changes when you look beyond genomics to broader biomodalities?
Precision medicine is expanding beyond genomics. Your genome is a day-one blueprint. But your “house” changes, and today proteomics and other layers often matter more. They better reflect current physiology and disease context. That opens a path to choices that fit your profile, not population averages.
Routine labs usually compare you to population norms. A better question is: “Is this unusual for you?” That’s how individual diagnosis and therapy take shape. Alignment with your healthy baseline matters more than matching an average. AI helps spot these shifts and link them to targets.
Meanwhile, automation and robotic measurements pair naturally with AI. The two evolve hand in hand. Over the past decade, AI for biology and AI for chemistry have steadily advanced: the former asks “how do we treat this disease?”, the latter “what drug targets this protein?”
Progress over the last ten years has been significant. It is not a single breakthrough but a convergence of technologies and methods. That combination moves drug discovery and development from intuition and luck to a more engineered discipline.
“We would all be much better off if the first drug was the right one.”
How do you advance if biology data can’t be scraped from the internet?
This is a core puzzle for AI in biology. Unlike text, biological data is not sitting open online. Nearly every company builds its own walled-off dataset. You cannot just train “the same thing” on a shared corpus or easily distill knowledge across models.
That reinforces medicine’s classic “silo” problem. Oncology and endocrinology may view the same case from different angles and fail to sync. Here, AI can, in principle, be “a specialist in everything.” It can combine signals no single human would see and deliver a “best-team huddle” at each step of care.
For that vision, a growing trend is to build “atlases” of biological information — often implemented as foundation models. As they become common, we may see what happened with open-source LLMs: open-source foundation models in biology driving very broad impact beyond corporate walls.
This does not end data competition. But it reframes the playbook: shared “foundations” plus private “extensions.” That way the field can preserve innovation and speed knowledge transfer across teams and specialties within medicine.
“It’s a place where you don’t have any of this data that people can just all train the same thing.”
What does VZVC back: founders, areas, and market lessons?
The focus spans two areas: AI for healthcare delivery and AI for clinical trials. Pande is involved with Genesis Therapeutics, which came out of his Stanford lab, and Insitro, launched by Daphne Koller. He is also incubating a company with a founder he has known for twenty years. It reflects a “few, trusted, long-term” approach.
The most important founder traits are trust and integrity. People who do what they say. Relationships that last 5–10 years and ideally across multiple companies. And a mindset of “how do we win together,” rather than simply “beat others.” This aligns with concentrated bets and deep collaboration.
Competition for deals does not look like chasing hot rounds. Often, “people make room” for VZVC because of how hands-on they are. Pande cites inspirations like Antonio Gracias at Valor, Thrive, and, of course, a16z as part of his own DNA. These are playbooks for concentration and durable partnerships.
There are clear lessons, too. Years ago, many resisted blending AI, ML, and medicine. That resistance is largely gone now. Yet techno-romance is still a trap. The coolest tech does not remove the hardest part — go-to-market. Pande urges technical founders to apply the same creativity to distribution.
“Go-to-market is at least as hard as the technology side. Often harder.”
What is overhyped in AI and biotech right now?
AI can indeed find insights beyond human reach. Hype creeps in when people claim “AI will cure everything.” The hesitation here is not doubt about AI. It is doubt about data. LLMs work because there is abundant training data. When data simply is not there, AI cannot magically solve the problem.
Hence the grounded lens on progress. The field needs advances in data, measurement, experimentation, and open “atlases.” Then models can improve target selection, drug design, and precise therapy matching for individuals. Engineering reduces the lottery in drug development.
This also matches the firm’s style. Few decisions, each with full immersion. In-house AI agents accelerate operations. Internal discipline preserves focus. The result is a portfolio where every company is a true partnership, not a statistical ticket in a long queue.
So, back to the opening question — why a small firm? Because it fits the current AI-biotech reality. Data is scarce. Experiments are expensive. Concentration gives each bet a real chance at impact, not just “being present” in the theme.
“When the data is just simply not there, then AI can’t magically solve that problem.”
Based on the original material.
