NVIDIA AI Podcast — GTC Live Washington, D.C.
Episode 4: AI for Science
Date: November 11, 2025
Host: NVIDIA (Moderated by Brad)
Guests:
- George Church (Chief Scientist, Lela Sciences)
- Matt Kinzella (CEO, Inflection)
- Mark Tessier-Levine (Co-founder/Chairman/CEO, Zyra Therapeutics)
- Anirudh Devgan (President/CEO, Cadence)
Episode Overview
This special GTC edition of the NVIDIA AI Podcast focuses on the transformative impact of AI and accelerated computing on scientific discovery. The panel brings together leaders from quantum computing, biotech, and chip design to discuss how technological advances—from GPUs to quantum computers—are accelerating breakthroughs across physics, chemistry, biology, and drug discovery. The conversation explores the interplay between computation and scientific progress, the rise of hybrid (AI + quantum + classical) approaches, commercialization strategies, and the realities versus the hype cycles surrounding these innovations.
Key Discussion Points & Insights
1. AI and Quantum Computing: Defining Terms and Opportunities
-
Quantum Fundamentals:
- Matt Kinzella kicked off with a primer on quantum ("the world of the very small, the atomic and the subatomic," 02:13), highlighting the necessity for quantum and classical computing to work together, especially as quantum GPU computing could address complex problems.
- Jensen Huang (in comic relief cameo) underscored the synergy:
"Quantum and classical computing really needs to work together so that we could bring in the usefulness of quantum computing... the two industries really need to work together as one." (03:56)
-
Real-World Quantum Commercialization:
- Inflection's commercialization strategy mirrors NVIDIA's staged approach—finding practical domains like timekeeping and RF sensors with real, current quantum advantage before full-scale quantum computing arrives. (07:00)
2. Accelerating Discovery: AI in Chip Design, Physical Systems, and Drug Design
- The Three-Layer Cake of Innovation (Anirudh Devgan, 08:52):
- AI for automation and productivity
- Ground truth (deep science & physics)
- Accelerated compute (CPU+GPU platforms)
- Quote:
"The more the modeling is accurate, more simulation happens, more optimization with AI can make drugs happen... in drug design... it is still very, very early days." (Anirudh Devgan, 17:29)
3. Drug Discovery & Molecular Science: Breakthroughs and Bottlenecks
- Inflection Point in Drug Discovery:
- Mark Tessier-Levine explained the dramatic challenges: 13-year timelines and high failure rates for drug development.
"AI... should be able to transform this from an artisanal endeavor into an engineering discipline... with much higher success rates and shorter timelines." (11:49)
- Mark Tessier-Levine explained the dramatic challenges: 13-year timelines and high failure rates for drug development.
- Recent Advances & Examples:
- AlphaFold and related breakthroughs in protein structure prediction cited as real, recent accelerators. (13:10)
- George Church shared practical examples:
"20 million fold reduction in cost of sequencing... we're really actively using not just the dichotomy between screening and analysis and prediction, but putting them together."
"At Dyno and Manifold we've made proteins that target the nervous system 100 times better..." (15:00) - Moderator/Brad linked theoretical potential to concrete results, such as rapid clinical outcomes for rare diseases (baby KJ gene therapy: 7 months from diagnosis to cure). (15:00)
4. Simulation and Data: Bridging Physical and Digital Experimentation
-
The Gradient from Physical to Digital (Anirudh Devgan, 17:05):
- Chip design: ~99% now done in silico (on computers)
- System design (cars, planes): ~20% digital, 80% physical
- Drug/molecular design: Only a few percent digital—huge room to grow
-
AI's Role in Biology:
- Mark Tessier-Levine:
"AI is to biology what math is to physics. For biology, we don’t have equations... AI can see patterns where we can't with the human eye." (18:29)
- Mark Tessier-Levine:
5. International Competition & Scale
- Moderator Brad raised the question of why China outpaces the US in raw number of drugs in the pipeline, despite the US's massive data center and compute resources.
- George Church responded:* Quality vs. quantity is key, and streamlined regulatory approaches (investigator-initiated trials) contribute, but integration of computation is the US's strongest advantage.* (19:55)
6. Investment, Public Expectations & Bubbles
- Matt Kinzella addressed the "quantum hype cycle" directly:
"Never give stock tips to friends... unlocking the power of quantum mechanics and turning that into products will result in orders of magnitude improvement... not 50%, not 100%, we're talking 10, 10,000, 1 million x improvement in performance." (21:50)
- Real quantum advantage today is in specific applications (e.g., timekeeping, sensors); for quantum computing, the key is building up logical qubits. "In 2023, we saw the first logical qubits... 100 logical qubits: material science advantage; 1,000: possibly drug discovery." (23:00)
7. Hybrid Compute Future: CPUs, GPUs, Quantum, and More
- Layered Compute Ecosystem:
- Mark Tessier-Levine: "We're big consumers of GPUs... For now, we're... with the technology that's being made available by Nvidia. In the future, quantum will help, especially for molecular dynamics simulations." (25:00)
- Anirudh Devgan: The future is hybrid; CPUs, GPUs, FPGAs, custom silicon, quantum—all have a role. (25:40)
- Matt Kinzella: "QPU will start to slowly layer into the data center and just expand what we can do with compute... it will result in more CPUs and more GPUs being deployed and sold..." (26:17)
8. Risks, Scaling, and the Multi-Phase AI Trajectory
-
Moderator sought concerns about market bubbles and risks from overbuilding infrastructure.
- Anirudh Devgan outlined the three “horizons”:
- Infrastructure AI (happening now—data centers, LLMs)
- Physical AI (coming next—autonomous vehicles, robotics)
- Sciences AI (long-term—drug/material discovery)
- Each horizon reinforces demand from the one below; AI’s impact is still just beginning. (27:55)
- Anirudh Devgan outlined the three “horizons”:
-
Mark Tessier-Levine cautioned on expectations:
"We're not going to get down that whole 13-year process [for drugs] down to two years. Can we cut it in half?... Let's be ambitious but realistic." (29:22)
-
George Church noted progress is tangible and energy efficiency (biological systems) can inspire future hybrid designs (31:00).
Notable Quotes & Memorable Moments
-
"When we say quantum, we're talking about the world of the very small...taking advantage of those very strange quantum mechanical properties."
—Matt Kinzella (02:13) -
"AI...should be able to transform this [drug discovery] from an artisanal endeavor into an engineering discipline with much higher success rates and shorter timelines."
—Mark Tessier-Levine (11:49) -
"AI is to biology what math is to physics."
—Mark Tessier-Levine, citing Eric Schmidt (18:29) -
"The more the modeling is accurate, more simulation happens, more optimization with AI can make drugs happen."
—Anirudh Devgan (17:29) -
"Quantum and classical computing really need to work together... the two industries really need to work together as one."
—Jensen Huang (impersonated in comic relief segment, 03:56) -
"The future will be hybrid... CPUs, GPUs, FPGA, custom silicon, quantum—all have a role."
—Anirudh Devgan (25:40) -
"We're not going to get down that whole 13-year process down to two years. Let's be ambitious but realistic."
—Mark Tessier-Levine (29:22) -
"If you talk petaflops per watt, biological intelligence is 12 orders of magnitude better [than current AI]."
—George Church (30:47)
Important Timestamps
- 00:52–03:41: Setting the scene—AI’s role in accelerating science, introduction to panel and quantum computing.
- 06:12–08:52: The importance of leadership (Jensen and NVIDIA) in deep tech, commercialization strategies.
- 11:49–15:00: Are we at an inflection point in drug and molecular discovery? Breakthroughs with AI, AlphaFold, gene therapy examples.
- 17:05–19:24: Shift from lab to in silico; digital simulation in chip/system/drug design.
- 21:50–24:10: Quantum bubbles, logical qubits, and actual quantum advantage.
- 25:00–26:24: Layering CPUs, GPUs, and QPUs in data center; hybrid and future compute paths.
- 27:55–33:01: AI's three horizons, risk management, setting expectations for impact and investment returns.
Tone & Atmosphere
The tone remains optimistic but grounded—panelists blend excitement about the wave of breakthroughs in AI and quantum with clear-eyed assessments of bottlenecks, scaling challenges, and the slow-but-steady nature of foundational scientific shifts. Camaraderie, technical depth, and humor (including a cameo comic relief "busboy"/Jensen appearance) keep the discussion engaging and relatable.
Summary Takeaway
The episode highlights that the intersection of AI, quantum, and accelerated compute is catalyzing a new scientific era. While the breakthroughs are tangible and accelerating—particularly in genomics, molecular biology, and materials science—panelists stress the need for hybrid approaches, robust infrastructure, realistic timelines, and vigilance against hype-driven bubbles. The transition from artisanal, empirical methods to engineering- and AI-driven science is well underway, promising a future where discovery scale, speed, and impact are fundamentally transformed.
