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

We have spent decades treating drug discovery like a lottery. You throw a thousand darts at a board, hope one hits a protein correctly, and then spend a decade praying it doesn't kill anyone in Phase III trials. It is a slow, bloated, and incredibly inefficient way to build anything. Vijay Pande, who spent years running the massive biotech practice at Andreessen Horowitz, seems to have reached his limit with that old-school model.

Pande recently left the $4 billion behemoth to start VZVC, a much smaller, AI-native firm. His thesis is simple: biology is finally transitioning from a discovery-based science to an engineering discipline. For builders, this is the most important shift in the sector since the invention of CRISPR. It means we are moving away from the era of 'lucky breaks' and into the era of 'predictable builds.'

The End of the Billion-Dollar Guessing Game

In the traditional biotech world, the scale of capital often masked the lack of efficiency. When you have billions to deploy, you can afford to be wrong 95% of the time. But Pande’s move to a smaller fund suggests that the game is changing. He isn't planning to spray and pray with thirty bets a year. Instead, he is looking for founders who treat biological systems like software stacks.

For a long time, the bottleneck in medicine wasn't just data; it was the nature of the data itself. Biology is messy, non-linear, and notoriously difficult to model. However, AI is starting to bridge that gap. We are seeing a shift where researchers can simulate how a molecule behaves before they ever touch a pipette. This doesn't just save money; it changes the entire risk profile of a startup.

Why Data Moats are Overrated

One of the most interesting takeaways from Pande’s new direction is his skepticism toward walled-off data. In the early days of AI, everyone thought the winner would be whoever owned the biggest proprietary dataset. If you had the most patient records or the most unique chemical library, you had a moat. Pande is arguing the opposite: open, shared datasets are the real catalysts for transformation.

For builders, this is a breath of fresh air. If you are starting a company today, you don't necessarily need to spend five years and fifty million dollars just to acquire a foundational dataset. The power is shifting from data ownership to model execution. The companies that win won't be the ones sitting on a mountain of private files, but the ones who can most effectively use open-source breakthroughs to solve specific engineering hurdles in the lab.

The Brutal Reality of Clinical Trials

Despite all the hype around AI, Pande remains grounded about the 'wet lab' reality. You can have the most sophisticated neural network in the world, but eventually, you have to put your product into a human being. Clinical trials remain brutally expensive and heavily regulated. AI hasn't disrupted the FDA yet, and it won't for a long time.

This is where many founders get stuck. They build a great model, they show promise in vitro, and then they hit the brick wall of clinical reality. Pande’s lean approach at VZVC suggests he is looking for companies that understand how to navigate this gap without burning through hundreds of millions in unnecessary overhead. The goal is to get to a 'yes' or 'no' faster, not just to spend more money.

The Founder Perspective: Build for Precision, Not Scale

If you are building in the AI-biotech space right now, the signal from Pande is clear: stop trying to be a 'platform' that does everything and start being a solution that fixes something specific. The era of the general-purpose biotech giant is being challenged by smaller, hyper-efficient teams that use AI to bypass the traditional discovery phase.

We are looking for founders who understand that the 'bio' part of the equation is now a data problem. If you can't explain your drug candidate in terms of a predictable engineering outcome, you are still playing the lottery. And as Pande has signaled by moving away from the $4 billion fund model, the smart money is getting tired of paying for tickets to that lottery.

The transition from discovery to engineering is the most significant pivot in the history of medicine. It turns a game of chance into a game of skill.

The takeaway for the crypto and AI crowd is that the 'lean startup' methodology is finally coming for the most expensive industry on earth. You don't need a massive headcount or a sprawling campus to change how we treat disease. You need a better model, a clear engineering path, and the discipline to stay small until the science proves you right.

What This Means for the Future

Pande’s shift away from high-volume betting toward concentrated, AI-native investments is a microcosm of the broader tech landscape. We are seeing a move away from 'growth at all costs' and toward 'efficiency at all costs.' In biotech, that efficiency is driven by machine learning.

We should expect to see a new wave of startups that look more like software companies than traditional labs. These teams will be smaller, their milestones will be more technical, and their path to market will be driven by data accuracy rather than just sheer capital. For those of us watching the intersection of AI and real-world utility, this is exactly where the most exciting work is happening.

The era of betting big and hoping for a miracle is closing. The era of betting small on precise engineering is just beginning. As a builder, you should be asking yourself: are you still trying to discover something, or are you ready to start building?

Read the original at TechCrunch AI →