Vijay Pande left Andreessen Horowitz's $4 billion biotech fund last year to launch VZVC, a smaller, AI-focused venture vehicle that operates on a fundamentally different thesis than the mega-fund model. The shift reflects both a personal philosophy and a broader recalibration in how biotech investing works in the AI era.
At a16z, Pande oversaw one of the most active biotech practices in venture capital, deploying capital across dozens of companies annually. VZVC takes the opposite approach. "We're not doing 30 bets a year," Pande said, signaling a move toward concentrated conviction investing rather than the portfolio-spray strategy that defined his previous role.
The fund's focus targets the intersection of AI and biology at a moment when the field fundamentally shifts in character. Pande sees biology transitioning from pure discovery science—where researchers hunt for novel molecules or mechanisms—to applied engineering. That distinction matters enormously. Discovery requires exploration, serendipity, and tolerance for failure. Engineering requires precision, repeatability, and systems-level thinking. AI excels at the latter.
This reframing explains why Pande left despite a prominent perch at one of the world's largest VCs. The inflection point he identifies in biology is real. Machine learning has already proven utility in drug target identification, molecular screening, and protein structure prediction. But the field remains hamstrung by two structural problems that no amount of AI can fully solve: clinical trials remain brutally expensive, and datasets remain fragmented and proprietary.
On trials, the economics have not budged meaningfully. Bringing a new drug to market still costs $2.5 billion to $3 billion on average, with Phase 3 trials consuming the bulk of that spend. AI can optimize trial design and patient recruitment, but it cannot compress the clinical evidence requirements that regulators demand. That remains a bottleneck no startup solves alone.
The dataset problem cuts deeper. Pande argues that the companies and institutions sitting on proprietary biological data—research hospitals, pharma giants, diagnostic labs—have structured incentives to keep it walled off. Competitive advantage flows from proprietary information. Yet AI systems trained on isolated datasets fail to generalize. Open, shared datasets, by contrast, unlock network effects. Models trained on diverse, representative data outperform siloed ones. The incentive structure needs to flip.
This perspective shapes VZVC's investment strategy. The fund will back companies that either solve concrete engineering problems in drug discovery and development, or that build platforms to surface and standardize biological data at scale. Neither category comprises the flashy AI-biology startups that dominated 2023 coverage. Both require patience and conviction.
Pande's move also reflects maturation in venture itself. The mega-fund model, which a16z perfected, works for scale and market dominance but can obscure conviction. Smaller, focused funds led by experienced operators can move faster on specific theses. Pande has domain expertise spanning computational biology, venture returns, and regulatory science. VZVC represents a return to thesis-driven investing in biotech at a moment when the thesis itself—engineering biology with AI—has shifted.
The question ahead: whether a smaller fund can attract top talent and reserve capital to sustain concentrated bets over the 10-plus year timescales biotech requires. Pande has the track record and reputation to answer that. The market will soon test whether the thesis holds.