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

The article profiles Vijay Pande’s shift from a16z’s ~$4B biotech practice to founding VZVC, an AI-native biotech venture. It argues that biology is moving from discovery to engineering and highlights that clinical trials remain extremely costly. Pande’s thesis is that open, shared datasets (vs. walled-off data) are the key enabler for AI-driven advances in medicine.

Analysis

The investable takeaway is less about “AI in biotech” as a theme and more about who owns the bottleneck inputs. If biology becomes more engineering-like, the durable economics should accrue to companies that generate, curate, and operationalize high-quality experimental data, plus the lab workflow vendors that convert more iteration into revenue. That is structurally better for diversified tools/automation names than for pure-play discovery platforms, where shared datasets can rapidly commoditize the model layer and compress the moat.

The market is probably still underpricing how long the clinical-validation step remains the binding constraint. AI can shorten target selection and candidate design over the next 6-18 months, but trial timelines, enrollment, and regulatory review still dominate cash burn and valuation realization over 2-4 years. If open datasets actually improve hit rates, the first-order winner may be CROs and trial-enabling software rather than the companies selling “AI drug discovery” narratives; the second-order loser is any biotech whose equity story depends on proprietary data being hard to replicate.

Contrarian view: consensus likely overstates the speed of value migration into AI-native therapeutics and understates the value of boring infrastructure. The more data becomes shared, the less differentiated a single-model advantage is, and the more important distribution, assay depth, and patient linkage become. Falsifier: if we start seeing materially faster Phase II readouts or a step-down in preclinical failure rates, the discovery-platform thesis deserves a higher multiple; absent that, this is a story to own through picks-and-shovels, not through the hype cohort.

Key Decisions for Investors

• Prefer a 3-6 month long TMO / DHR basket as the cleaner expression of rising experimental throughput and outsourced validation demand; add on weakness if biotech funding improves.
• Fade pure-play AI-discovery enthusiasm with a tactical short RXRX or SDGR against long TMO; thesis breaks if either name shows repeatable partner revenue or a material de-risking clinical readout within 2 quarters.
• Avoid chasing a broad XBI rally as a proxy for AI-biotech; the article is more likely to support infrastructure multiples than the basket’s highest-beta discovery names.
• Set an alert for any evidence that shared-dataset initiatives are reducing preclinical failure rates by a visible margin; that would be the first signal to revisit the moat assumptions in discovery software and platform biotechs.
• If forced to own the theme, express it via life-science tools and trial-enabling software rather than venture-style private exposures, since the cash conversion path is much clearer and the drawdown risk is lower.