Vijay Pande Bets Small With VZVC: How AI Is Turning Biology From Discovery Into Engineering

TL;DR

• Former a16z Bio + Health lead Vijay Pande left his $4 billion practice to launch the lean, AI-native firm VZVC, shifting from 30+ bets a year to a concentrated portfolio of just a handful of deeply-supported companies.
• Pande argues biology is finally becoming an engineering discipline, where AI can design rather than just discover drugs, but says clinical trials remain the industry's biggest cost and time bottleneck.
• He is calling for open, shared datasets as the critical infrastructure for AI-driven medicine, warning that without them the field will stall despite advances in models.

Why Walking Away From $4 Billion Was The Strategy

For nearly a decade, Vijay Pande was the face of bio at Andreessen Horowitz. As the founding general partner of a16z Bio + Health, he built one of the largest dedicated biotech practices in venture, overseeing more than $4 billion in assets and a famously high-velocity strategy of backing dozens of companies per year.

So when he quietly stepped away to start VZVC, a tiny, intentionally lean firm, the move raised eyebrows across both Silicon Valley and biotech. In a recent interview detailing his decision, Pande said the scale itself had become the problem.

At a16z, the model was breadth: make 30 bets a year, cover the waterfront of bio, health, and AI, and let the power law do its work. At VZVC, Pande is doing the opposite. The new firm plans to make only a few investments per year and work far more closely with founders from inception through scale.

The logic, he explained, is that AI-native bio companies don't need the same kind of spray-and-pray venture support. They need deep, technical partnership. VZVC is designed to be AI-native from the ground up — not a traditional bio fund bolting on AI, but a firm where computation, data engineering, and model development are core to how companies are built and evaluated.

From Discovery to Engineering: Biology's Platform Shift

Pande's central thesis for VZVC is one he has championed for years, first as a Stanford professor who pioneered distributed computing for protein folding with Folding@home: biology is transitioning from a science of discovery to a discipline of engineering.

For a century, drug development has been largely empirical — screen thousands of molecules, hope one works, and iterate slowly. Pande argues AI is finally flipping that model. With generative models, large-scale omics data, and physics-based simulations, researchers can now design biology with intent: engineer proteins, edit pathways, and predict behavior before ever touching a wet lab.

That shift changes what a venture firm needs to be. Instead of funding a portfolio of shots on goal, VZVC wants to back the small number of teams building true engineering platforms — companies where the loop between dry lab and wet lab is tight, automated, and increasingly driven by AI. In that world, Pande said, being lean is an advantage. Fewer companies means more time to help solve the hard technical problems that actually determine success.

The Real Bottleneck Isn't The Lab. It's The Trial

If AI is making the early stages of drug discovery faster and cheaper, Pande is quick to point out where it isn't: clinical trials.

Even as AI compresses years of preclinical work into months, the cost and complexity of human trials have remained stubbornly high. Pande called clinical development the single biggest bottleneck to translating AI advances into real medicines — a system that still takes 7 to 10 years and can cost more than a billion dollars per approved drug.

AI can help optimize trial design, identify better patient populations, and predict toxicity earlier, but it can't yet bypass the fundamental need for rigorous human data. Until the trial process itself is re-engineered — with better biomarkers, decentralized infrastructure, and more predictive models — Pande warned that faster discovery will just create a traffic jam at the clinic door.

That view is shaping VZVC's investment focus. The firm is particularly interested in startups tackling the trial bottleneck directly, whether through AI-powered clinical operations, new ways to generate real-world evidence, or platforms that can fail bad candidates faster and cheaper before they reach Phase 2 and 3.

No Data Commons, No AI Revolution

The other piece of infrastructure Pande says is missing is data. While AI models in biology have exploded in capability, he argues they are still starved of the high-quality, standardized, and open datasets needed to truly transform medicine.

Unlike fields like software or even large language models, where the internet provided a vast commons of training data, biology's most valuable data remains siloed inside pharma companies, academic labs, and hospitals. Much of it is unstructured, inconsistent, or simply not shared.

Pande has become a vocal advocate for open, shared datasets as a public good for AI-driven biology. He pointed to examples like the Protein Data Bank and UK Biobank as proof that when the field rallies around a common data resource, progress accelerates dramatically. Without more efforts like that — especially for longitudinal patient data, perturbation datasets, and clinical outcomes — he said even the best models will hit a ceiling.

For VZVC, that philosophy is more than talk. Pande said the firm will encourage and support its portfolio companies to contribute to and build on shared data commons where possible, betting that openness will create larger markets for everyone rather than eroding competitive advantage.

The Lean Future of Bio VC

VZVC's launch reflects a broader recalibration happening in biotech venture. After the boom-and-bust cycle of 2020-2023, when mega-funds and rapid-fire investing became the norm, many investors are now questioning whether that scale serves science.

Pande's bet is that the next generation of transformative bio companies will look more like elite AI labs than traditional biotechs — small teams of deeply technical founders iterating quickly with models and automation. Backing them, he argues, requires a different kind of firm: smaller, more concentrated, and far more technical.

He's not done with biology. He's just done doing it at volume.

AndroGuider Team

Articles written by the AndroGuider team. We try to make them thorough and informational while being easy to read.