Overview of Can America Win The AI Biotech Race Against China? (A16Z podcast)
This episode features Lada Nuzhna (founder, General Control) and Elliot Hershberg (partner, Amplify) in a wide-ranging conversation about the paradox of biotech in 2025: the underlying science and tools are dramatically better, yet the economics, market structure, and incentives are deteriorating. They diagnose why U.S. biotech is losing advantageous axes (cost, speed) to China, what AI realistically can and cannot fix, how aging and GLP-1s changed the game, and what policy or industry moves could re-start the next wave of iconic biotech companies.
Key takeaways
- Science vs economics paradox: biology and discovery tools have improved massively, but drug development costs and timelines have continued to rise (often cited: ~$2B per approved drug).
- Structural and incentive problems—industry consolidation, clinical trial execution, and regulatory practices—are at least as important as pure regulatory policy.
- China competes successfully on speed and cost (investigator-initiated trials, implied approvals, parallelized reviews), causing companies to run first-in-human trials outside the U.S.
- AI will be ubiquitous in biotech within 5 years, but its biggest early wins will be in discovery and enabling new modalities; predicting human efficacy (phase II failures) remains the hardest problem.
- The next generational winners will likely be either (a) new modalities that change what’s possible biologically or (b) foundational platforms/infrastructure companies that become indispensable to biotech workflows.
- Aging is a high-value target but faces payer and approval incentives that don’t favor preventative or long-horizon interventions; GLP-1s proved both commercial demand and that big consumer-facing biotech is possible.
- Concrete policy/industry fixes suggested: lower per-patient trial costs, enable investigator-initiated trials for advanced modalities in the U.S., and create incentive mechanisms (e.g., “orphan-like” designation) for age-related/chronic diseases.
The current state of biotech (market vs science)
- Market indicators: one-fifth of public biotechs trading at or below cash, months without IPOs, many EV-negative public companies — sentiment and capital are constrained.
- Science indicators: major advances—zero-shot antibody design, virtual cell projects, gene editing, mRNA platforms—meaningful translational potential exists at early stages.
- The industry is split: early-stage science is exciting; later-stage economics (clinical development, commercialization) remain painful and costly.
Structural and regulatory barriers
- Regulatory history tightened over time (safety + efficacy requirements after incidents like thalidomide), making approvals more burdensome.
- Two-pronged execution lag:
- The rule-of-law/regulatory framework (what you can legally do).
- Industry entrenchment (e.g., ~dozen large clinical research organizations (CROs) that resist rapid operational change).
- Cultural norms: a tacit acceptance that trials are expensive and slow; companies routinely take trials overseas (Australia, China) for speed and cost advantages.
- Suggested priorities: regulatory modernization that reduces low-value friction, adoption incentives for CRO modernization, and metrics such as "cost per patient in trial" as a regulator-considered KPI.
China’s role and the geopolitical angle
- China has moved from a low-value manufacturing role to a leading role in trials and novel modality work by deregulating and streamlining review processes (implied approvals, parallel review, investigator-initiated trials).
- Speed + lower costs = geographic arbitrage; Chinese teams can beat US teams to clinic even with identical mechanisms.
- Consequences:
- Secrecy and IP strategy may matter more (shorter commercial half-life of inventions).
- US must either match speed/cost or double down on inventing entirely new modalities that can’t be easily copied.
- Long-term, the U.S. need not be hollowed out if it focuses on “zero-to-one” invention and regulatory innovation to keep early human testing viable domestically.
The role and limits of AI in drug discovery
- Consensus: AI will be ubiquitous in biotech within five years. It’s a critical new experimental tool analogous to the role of software.
- Where AI helps now:
- Faster preclinical discovery (molecule design, antibody engineering).
- Enabling new product types and raising target product profiles (TPPs).
- Infrastructure/platform plays (sequencing + modeling pipelines).
- Limits:
- Most development spend and failures are in clinical stages—especially phase II (efficacy). AI models trained on preclinical data may not yet reliably predict human efficacy.
- Biggest gains come from integrating large human datasets into models and generating novel modalities that were previously inexpressible.
- Practical framing used in the podcast: tackle three “horsemen” of Eroom’s Law — regulation/structural cost, biological understanding (predicting efficacy), and expression/engineering of modalities.
Aging, GLP-1s, and big indications
- GLP-1s (e.g., semaglutide) have demonstrated that a consumer-facing, high-demand therapeutic can change market dynamics and investor appetite.
- Aging as an indication:
- Scientific progress exists (mouse/monkey lifespan extensions), but we lack validated human aging biomarkers (surrogate endpoints) to expedite approvals.
- Payer incentives (e.g., Medicare structure) discourage preventative aging therapeutics—companies face uncertain reimbursement.
- Policy/incentive ideas:
- Create incentives akin to orphan-drug status but targeting chronic/age-related diseases (to offset high development cost and long timelines).
- Pursue stepwise approval strategy: start with safe, small-effect interventions, then move to higher-risk modalities as evidence and regulation evolve.
“Magic wand” recommendations (what they’d change)
- Lada/Elliot suggestions:
- Make per-patient trial cost a KPI (example: $10k/patient historically → now often ~$500k).
- Enable investigator-initiated trials for cell/gene therapies in the U.S. (as in Australia/China).
- Incentivize aging and chronic-disease R&D with targeted policy tools (e.g., orphan-for-common concept).
- Encourage regulatory and CRO modernization so U.S. remains the place where first-in-human and pivotal trials are run.
- Protect incentives that keep invention in the U.S. (balance openness and needed secrecy/IP strategies).
Where the next iconic biotechs will come from
- Two major value-creation paths:
- New modalities — technologies that change what drugs can be made (e.g., recombinant DNA, mRNA, human monoclonal antibodies historically).
- Infrastructure/platforms — foundational companies (sequencing, design tools, generative platforms) that become industry utilities.
- The winners will likely combine modalities with generative/design platforms and delivery/targeting breakthroughs (e.g., non-liver targeted LNPs, epigenetic editing, better targeting to brain/kidney/heart).
Actionable recommendations (by stakeholder)
- Founders:
- Focus on modality/inventive bets that can’t be commoditized by speed/cost arbitrage.
- Think carefully about secrecy/IP, and plan clinical paths (consider non-US trials but design for global/regulatory strategy).
- Build platform-first products when possible (platform = product).
- Investors:
- Back novel modalities and platform infrastructure—both can generate outsized returns.
- Evaluate companies on their path to lower clinical development friction and realistic regulatory timelines.
- Policymakers/regulators:
- Modernize trial standards, reduce low-value friction, and consider incentive programs for chronic/aging disease R&D.
- Consider mechanisms to keep early human testing competitive in the U.S.
- Industry/CROs:
- Adopt technologies that reduce per-patient costs and accelerate execution (digital trials, parallelized workflows).
Notable quotes
- “Everyone will be using AI in biotech industry five years from now.”
- “There is no law of physics that requires it to be $500,000 [per patient] in terms of complexity and cost to dose a patient in a trial.”
- “China is an engineering state and America is a lawyer state—but America is also an inventor state. We have to invent our way out.”
- “The drug is part information product, part diagnostic” — on platform/personalized products like mRNA cancer vaccines.
Final thought
The path forward is not a single lever—solving biotech’s crisis requires simultaneous progress in invention (new modalities), operational modernization (trial execution and CROs), smarter regulatory incentives, and selective use of AI (especially where it helps enable novel medicines). Investing in modality breakthroughs and platform infrastructure, while pushing regulatory reforms that lower practical barriers to first-in-human testing, are the clearest routes for the U.S. to stay competitive in the AI–biotech race against China.