Vijay Pande made a deliberate choice to walk away from scale. After running Andreessen Horowitz's $4 billion biotech practice, he launched VZVC last year as a lean, AI-focused fund that operates on a radically different thesis from the venture machine he left behind. The shift reflects a broader recalibration in how serious money thinks about biology.
"We're not doing 30 bets a year," Pande told TechCrunch AI, signaling that VZVC takes a concentrated approach over portfolio breadth. This isn't retreat. It's strategic focus.
The core bet Pande makes is that biology is transitioning from discovery science to engineering discipline. That distinction matters. Discovery science asks what exists in nature. Engineering science asks what you can build and optimize. This shift opens space for AI tools to work differently than they did before, because engineered systems are predictable in ways natural discovery is not. You can apply machine learning to optimization problems that have clear parameter spaces and measurable outputs. You cannot reliably apply it to finding unknown unknowns.
This reframing has direct implications for how biotech companies approach drug development and which startups merit capital. Companies that see AI as a tool for systematic optimization of known problems win. Companies that hope AI will replace human intuition in radical discovery tend to stumble.
Pande also circles back to a hard truth that rarely makes venture headlines: clinical trials remain brutally expensive. No amount of AI-driven computational biology eliminates the cost of testing molecules in humans. The AI component compresses the early research phase, screens compounds faster, and identifies candidate drugs with higher precision. But phase one, two, and three trials still consume billions and consume years. This pushes successful biotech startups toward problems where the trial population is smaller, endpoints are clear, and regulatory pathways are defined. Rare disease, precision oncology, and localized indication spaces attract capital more reliably than broad-population conditions.
The other insight Pande emphasizes centers on data architecture: open, shared datasets beat proprietary walled gardens for AI training in medicine. This contradicts the venture narrative of defensibility through IP moats. But it aligns with what actually happens in practice. Teams that build on top of shared, curated, transparent datasets iterate faster and generate better models because they inherit a massive training foundation and can focus engineering effort on domain-specific problems rather than basic data plumbing. Proprietary datasets create short-term competitive advantage but long-term technical debt.
VZVC's investment strategy reflects all three insights. Smaller fund size allows Pande to move quickly on concentrated bets where thesis alignment runs deep. Focus on companies that treat biology as an engineering problem, not a discovery lottery. Backing teams that contribute to and benefit from open data infrastructure rather than hoarding information creates network effects that compound over time.
This approach runs counter to the mega-fund playbook that dominated biotech venture for the past decade. Pande's departure from a16z and launch of VZVC signals that even inside the venture establishment, conviction is shifting. The future of AI in biology belongs to teams that sweat optimization and embrace openness, not teams that chase headline-grabbing moonshots or guard proprietary treasure.