Biotech investor Elliot Hershberg on AI's growing role in drug discovery and the century of biology thesis

Summary

Elliot Hershberg, a computational biologist turned venture investor now at Amplify Partners, sees the AI-biology convergence moving past the hype trough and into a phase of genuine, measurable productivity gains. Amplify recently closed $900 million in new capital, including a dedicated $200 million bio-focused vehicle that Hershberg is helping build. The firm has landed in the top 5% of venture returns across three of its four funds, a track record largely built without fanfare.

On the state of AI in drug discovery, Hershberg frames the last decade through a Gartner hype cycle lens. Early enthusiasm around companies like Recursion gave way to disillusionment when one-shotting cancer cures proved elusive, but foundational breakthroughs have since validated the thesis. The Nobel Prize awarded to DeepMind's AlphaFold team and David Baker at the University of Washington is his marker that AI is now making meaningful, peer-recognized contributions to hard biological problems.

DNA sequencing costs are falling faster than Moore's Law, creating a data tailwind that compounds with advances in molecular machine learning and virtual cell modeling. Hershberg argues that stacking these curves is beginning to produce impressive results at speed.

Picks-and-Shovels vs. Drug Products

Amplify is backing both categories. On the drug side, Hershberg points to Centivax founder Jake as an example of a computational immunologist building medicines that simply could not exist without these tools, including a universal flu vaccine with broader platform potential. On the infrastructure side, he cites Schrödinger, a molecular dynamics company with roughly 25 years of history, as a cautionary parable about how long pharma sales cycles can be, but notes the environment has structurally shifted.

Post-AlphaFold and post-ChatGPT, pharma CEOs are now fielding pressure from shareholders and their own scientists to adopt AI tooling. Hershberg describes this as a qualitatively different demand signal, one where executives see their own families using large language models and conclude adoption is no longer optional.

Biotech as an Asset Class

Hershberg acknowledges the broader critique that biotech has historically been a poor index-level investment, with dramatic underperformance outside a handful of breakout names like Genentech. He recommends Bruce Booth at Atlas Ventures as a source of more nuanced return data that challenges the gloomiest narratives. His view is that engineering-driven biology, where drug development becomes more predictable and scalable, is the regime change that could make the next five years a genuine inflection point for venture bio returns. The core investment logic is simple: if AI inference can produce a billion-dollar drug product, the asymmetry is compelling.