"We're not doing 30 bets a year": Vijay Pande on betting small after running $4 billion at a16z

Vijay Pande ran one of the most influential biotech investment practices on the planet — roughly $4 billion in assets under management at Andreessen Horowitz. Then he walked away from it to start something dramatically smaller. That decision, and the reasoning behind it, tells you something important about where AI and biology are actually headed — and what it means for founders and developers building at that intersection right now.

"We're not doing 30 bets a year": Vijay Pande on betting small after running $4 billion at a16z is more than a profile of one investor's career pivot. It's a signal about how serious technical minds are repositioning around AI — not by spraying capital broadly, but by going deep on a smaller number of genuinely hard problems. For the Asia tech ecosystem, where AI infrastructure is maturing fast and biotech is just beginning to attract serious engineering talent, that signal is worth reading carefully.

What Happened

Vijay Pande left a16z's biotech practice — which had grown to roughly $4 billion — to launch VZVC, a much smaller, AI-native venture firm. The move is notable not just because of the scale difference, but because of the philosophy driving it.

According to the TechCrunch interview, Pande believes biology is undergoing a fundamental shift — from a "discovery" science to an "engineering" one. That framing matters. Discovery science is probabilistic and portfolio-driven: you make many bets because most will fail and a few will pay off enormously. Engineering science is different. When you can design, simulate, and iterate on biological systems the way software engineers iterate on code, the failure modes change. You need fewer bets, but each one requires deeper conviction and closer collaboration.

Pande is also candid about what hasn't changed: clinical trials remain brutally expensive. AI can accelerate drug discovery, compress timelines in the lab, and surface candidates that human researchers would miss — but getting a therapy through Phase II and Phase III trials still costs hundreds of millions of dollars and years of calendar time. The bottleneck has shifted upstream, into the lab and the data, but the downstream constraint hasn't moved.

His other strong conviction is about data architecture. Pande argues that open, shared datasets — not walled-off proprietary ones — are what will actually let AI transform medicine. The companies and institutions that hoard biological data may win short-term competitive advantages, but the models trained on siloed data will plateau. The breakthroughs will come from the researchers and companies willing to contribute to and draw from shared data commons. That's a structurally different bet than most pharma incumbents are currently making.

VZVC, as Pande describes it, is built around this thesis: fewer investments, more involvement, and a genuine AI-native approach to evaluating and building biotech companies from the ground up.

Why It Matters for Asia

Asia's biotech and healthtech sectors are at an inflection point that mirrors, with roughly a five-year lag, where the US was when Pande first started making serious AI-biotech bets at a16z. That lag is compressing fast.

Countries like South Korea, Singapore, Japan, and India have world-class genomics research institutions, large patient populations with distinct genetic profiles that are underrepresented in Western datasets, and increasingly sophisticated AI engineering talent. What they've historically lacked is the venture infrastructure to connect deep science with aggressive product development. That gap is closing.

Pande's argument about open, shared datasets has particular resonance in Asia. Several governments in the region — Singapore's Precision Health Research initiative, Japan's biobank programs, India's genomics grid — have made explicit policy bets on open data infrastructure. If Pande is right that shared data commons will outcompete walled gardens, Asian institutions that have invested in open infrastructure may find themselves with a structural advantage over the next decade, not a disadvantage.

The shift from discovery science to engineering science also plays to Asia's strengths. The region produces enormous numbers of engineers — people who think in systems, iteration loops, and measurable outputs. Biology-as-engineering is a framing that maps naturally onto how engineering-heavy cultures approach problem-solving. The challenge has always been translating that engineering mindset into biological domains where the feedback loops are slow and expensive. AI is compressing those feedback loops.

For founders in Southeast Asia and Northeast Asia building at the AI-biology intersection, Pande's fund size choice is also instructive. VZVC being smaller and more focused means the next generation of serious AI-biotech companies may not need — or want — the megafund treatment. Smaller, more technically engaged investors who can actually evaluate the science may become the preferred partners for founders who are themselves deeply technical.

The Asia tech ecosystem has watched the US AI wave from a position of catching up. In AI-native biotech, the starting lines are closer together than they've ever been.

What This Means for Developers

If you're a developer working anywhere near AI and biology — or thinking about where to point your next few years of technical energy — Pande's thesis has concrete implications.

Biology is becoming an engineering domain. That means the tools, workflows, and mental models that software engineers use are becoming directly applicable to biological research. Protein structure prediction, gene circuit design, drug-target interaction modeling — these are increasingly software problems. If you're a strong ML engineer who hasn't looked seriously at computational biology, the entry cost is lower than it's ever been, and the upside is enormous.

Data infrastructure is the leverage point. Pande's conviction about open datasets is a data engineering argument as much as it is a scientific one. The teams that will win in AI-driven drug discovery are the ones that can ingest, clean, harmonize, and query biological datasets at scale — across modalities (genomics, proteomics, clinical records, imaging) and across institutions. That's a hard engineering problem, and it's one where developers with strong data pipeline experience have a genuine edge.

Fewer, deeper bets apply to careers too. The spray-and-pray approach to learning new AI frameworks — picking up a new library every month, shipping toy demos, staying permanently at the surface — doesn't compound. Pande's conviction-based approach maps onto how the best technical careers are built: find a genuinely hard problem at the intersection of AI and a domain with real-world stakes, go deep, and stay there long enough to develop judgment that can't be replicated by someone who just read the same blog posts you did.

For developers building on platforms like MonstarX — where AI-native development is the default, not an afterthought — the Pande thesis is a useful lens. The most valuable thing you can build right now isn't another wrapper around a foundation model. It's infrastructure, tooling, or applications that make the hard parts of AI-biology development faster and more reliable: data connectors, evaluation pipelines, experiment tracking, regulatory documentation generation.

The engineering surface area in AI-native biotech is vast, and most of it is still underdeveloped. Developers who move into that space now, with strong AI engineering fundamentals and a willingness to learn enough biology to ask the right questions, are positioning themselves ahead of a wave that is still building.

Key Takeaways

• Vijay Pande left a16z's ~$4 billion biotech practice to start VZVC, a smaller, AI-native fund built around fewer, higher-conviction investments — not the broad portfolio approach that defines megafund biotech investing.
• Biology is shifting from discovery science to engineering science. AI is enabling researchers to design and iterate on biological systems with increasing precision, changing the fundamental risk profile of biotech investment and development.
• Clinical trials remain the immovable bottleneck. AI compresses upstream timelines — lab work, candidate identification, target validation — but the cost and duration of human trials hasn't changed. Any honest AI-biotech thesis has to account for that.
• Open, shared datasets will outcompete proprietary silos. Pande's conviction is that the AI models trained on broad, open biological data will surpass those trained on walled-off corporate datasets. This has direct implications for how biotech companies should think about data strategy.
• Asia's structural position is stronger than it looks. Government investment in open genomics infrastructure, large and genetically diverse patient populations, and deep engineering talent pipelines put several Asian countries in a better position for AI-native biotech than the conventional narrative suggests.
• For developers, the lesson is depth over breadth. The most durable technical careers — and companies — in AI will be built by people who go deep on a genuinely hard problem, not by people who stay permanently at the surface of every new framework and model release.

The most interesting thing about Pande's move isn't the fund size — it's the underlying bet that the era of throwing capital at many uncertain experiments is giving way to something more precise. When one of the most experienced AI-biotech investors alive decides that conviction matters more than coverage, that's worth taking seriously, whether you're allocating capital, allocating engineering time, or deciding what to build next.