GTC DC 2025 Healthcare Special Address

Category: Healthcare Special Address · Year: 2025 · ▶ Watch

Switch language → 中文

Segments (15)

  • 00:00 · Introduction
    • Kimberly Powell welcomes attendees to GTC DC, comparing the event to a family reunion and a wedding for the healthcare ecosystem.
  • 02:18 · NVIDIA’s Full Stack Computing in Healthcare
    • An overview of NVIDIA’s role as an infrastructure company providing full-stack computing, from chips to domain-specific platforms for healthcare.
  • 04:38 · Accelerating Healthcare Breakthroughs
    • Highlighting the history of accelerated computing in healthcare, including image reconstruction and record-breaking DNA sequencing.
  • 07:28 · The Dawn of AI for Science
    • Discussing the shift from pattern matching to generative design in biology, driven by models like AlphaFold.
  • 09:58 · Announcing OpenFold3 NIM
    • Introduction of the OpenFold3 NIM for enterprise-scale computational protein structure prediction.
  • 11:13 · NVIDIA Open Models and Clara
    • Showcasing NVIDIA’s commitment to open-source models, including Nemotron, Cosmos, GROOT, and the Clara suite for biomolecular design.
  • 14:25 · Eli Lilly AI Factory Partnership
    • Announcement of a partnership with Eli Lilly to build BioPharma’s most powerful AI factory using NVIDIA Blackwell.
  • 20:08 · CZI Partnership for Virtual Cell Models
    • NVIDIA and the Chan Zuckerberg Initiative join forces to accelerate virtual cell model development.
  • 21:59 · Healthcare Setting the Pace for AI Adoption
    • Data showing healthcare leading enterprise AI adoption to address critical supply and demand gaps in patient care.
  • 25:44 · Agentic Systems and High-Performance Inference
    • Exploring how agentic AI is automating workflows and the need for high-performance inference platforms like Blackwell.
  • 29:24 · Verily Partnership for Precision Health
    • Verily integrates NVIDIA AI tools into its Workbench to accelerate precision health research.
  • 32:00 · Physical AI and Healthcare Robotics
    • Introduction to Physical AI for healthcare, highlighting the scale of the opportunity across hospitals and medical devices.
  • 34:25 · Isaac for Healthcare and Moxi 2.0
    • Showcasing the IGX Thor platform and Isaac for Healthcare, featuring Diligent Robotics’ Moxi 2.0 and a ‘Hello World’ robot arm.
  • 36:00 · Johnson & Johnson MedTech Robotics
    • Johnson & Johnson MedTech uses NVIDIA Omniverse and Isaac to build and simulate next-generation surgical robots.
  • 38:35 · Clara Open Models for Multimodal Medical AI
    • Highlighting Clara open models for medical imaging, the success of MONAI, and integrations with partners like Kitware and Ansys.

Product Announcements (5)

  • [09:58] OpenFold3 NIM
    • A computational protein structure prediction microservice.
    • specs: Optimized for extreme performance, deployable at enterprise scale.
    • availability: Available now on build.nvidia.com
  • [12:45] NVIDIA Clara Open Models
    • A suite of biomolecular design models including La-Proteina, CodonFM, ReaSyn, and GenMol.
    • specs: La-Proteina can design all-atom 3D protein structures up to 800 amino acids.
    • availability: Available on Hugging Face
  • [17:15] Eli Lilly AI Factory
    • BioPharma’s most powerful AI factory built in partnership with Eli Lilly.
    • specs: Powered by over 1,000 NVIDIA Blackwell GPUs.
    • availability: Coming online in the next couple of months
  • [32:40] NVIDIA IGX Thor
    • An enterprise-ready platform for physical AI and robotics.
    • specs: Real-time processing capable of handling sensor inputs in tens of milliseconds, supports vision language models.
    • availability: Not explicitly stated
  • [34:25] Isaac for Healthcare
    • A ‘Hello World’ platform for healthcare robotics development.
    • specs: Includes simulation (Omniverse), data collection, policy training (GROOT), and deployment (Holoscan).
    • availability: Available now

Specific Numbers (9)

Timestamp Metric Value Context
06:20 Guinness World Record 1 Awarded for the fastest DNA sequencing technique using NVIDIA Parabricks.
13:10 Amino Acids 800 The size of proteins that the La-Proteina model can create, compared to previous models limited to 400.
17:28 GPUs 1,000+ Number of NVIDIA Blackwell GPUs powering the new Eli Lilly AI factory.
22:05 AI Adoption Rate 27% Share of healthcare businesses deploying AI applications, leading the U.S. economy according to Menlo Ventures.
26:50 Conversations Per Year 50M+ Scale of clinical conversations processed by Abridge.
26:50 Throughput Multiplier 5x Speech recognition throughput improvement achieved by Abridge.
32:00 Hospitals 160K Global scale of hospitals representing the physical AI opportunity.
35:45 Cost $200 Cost of the 3D-printed robot arm used in the ‘Hello World’ robotics demonstration.
39:20 Downloads 7M Number of downloads for the MONAI medical imaging framework.

Benchmark Claims (2)

  • [26:50] Token Throughput per GPU: Higher performance curve
    • vs: GB200 NVL72 vs H200 NVL8
    • gain: Significant increase in token throughput for agentic AI inference workloads.
  • [26:50] Speech Recognition Throughput: 5x
    • vs: Previous generation or baseline
    • gain: 5x improvement in processing clinical conversations for Abridge.

Customer Stories (7)

  • [06:20] Broad Clinical Labs, Roche, Boston Children’s Hospital
    • Utilized NVIDIA Parabricks for genomic sequencing analysis.
    • outcome: Broke the Guinness World Record for the fastest DNA sequencing technique, improving diagnostic speed for NICU patients.
  • [14:25] Eli Lilly
    • Partnered with NVIDIA to build a massive AI factory using Blackwell GPUs and established Lilly Tune Labs.
    • outcome: Accelerating drug discovery by turning proprietary data into biomedical intelligence and fostering startup innovation.
  • [20:08] Chan Zuckerberg Initiative (CZI)
    • Joined forces with NVIDIA to use Clara and MONAI for their Virtual Cells Platform.
    • outcome: Accelerating the development of virtual cell models to better understand biology and disease.
  • [26:50] Abridge
    • Deployed NVIDIA inference platforms for their clinical documentation AI.
    • outcome: Achieved 5x speech recognition throughput, processing over 50 million conversations per year to reduce clinician burnout.
  • [29:24] Verily
    • Integrated NVIDIA AI tools, including NeMo and Parabricks, into the Verily Workbench.
    • outcome: Accelerating precision health research by providing AI-ready data and tools to thousands of researchers.
  • [33:40] Diligent Robotics
    • Adopted the NVIDIA IGX Thor platform for their Moxi 2.0 hospital robot.
    • outcome: Enabled advanced spatial awareness and autonomous navigation to assist healthcare staff with logistical tasks.
  • [36:00] Johnson & Johnson MedTech
    • Utilized NVIDIA Omniverse and Isaac to create digital twins of operating rooms and simulate surgical procedures.
    • outcome: Accelerating the development of next-generation surgical robots through simulation-first training.

Key Technologies (10)

  • CUDA: NVIDIA’s parallel computing platform and programming model that accelerates computational workloads.
  • AlphaFold: An AI system that predicts the 3D structure of proteins from their amino acid sequences.
  • NIM (NVIDIA Inference Microservice): Optimized, enterprise-ready containers for deploying AI models efficiently.
  • Nemotron: An open model family designed for building agentic AI systems.
  • Clara: NVIDIA’s suite of biomedical AI models and frameworks for drug discovery and medical imaging.
  • Blackwell Architecture: NVIDIA’s next-generation GPU architecture designed to power massive AI factories and high-performance inference.
  • IGX Thor: An enterprise-ready computing platform designed for real-time physical AI and medical robotics.
  • Isaac: NVIDIA’s platform for robotics simulation, training, and deployment.
  • Omniverse: A computing platform for developing 3D workflows and physically accurate digital twins.
  • MONAI: The Medical Open Network for AI, an open-source framework standard for medical imaging research and development.

Demos Shown (1)

  • [34:25] A demonstration of the ‘Hello World’ for Healthcare Robotics, featuring a $200 3D-printed robot arm (SO-ARM) performing a tool-picking task after being trained in simulation.
    • Yes

Predictions / Commitments (3)

  • [15:05, A few short years] It is not going to surprise me in a few short years that every pharma company is going to be announcing their billion dollar supercomputer.
  • [22:25, Current and ongoing] The demand for healthcare services is just far outstripping our supply… there is now technology that can enable incredibly new workflows.
  • [32:25, Future] At some point, we actually are going to want a digital twin of each one of us.

Companies Mentioned (8)

Schrödinger · DeepMind · MIT and Genesis Therapeutics · OpenFold Consortium · Menlo Ventures · Epic · Kitware · Ansys

Notable Quotes (4)

What math is to physics, AI is to biology. — Kimberly Powell @ 07:35

It is not going to surprise me in a few short years that every pharma company is going to be announcing their billion dollar supercomputer. — Kimberly Powell @ 15:05

Software used to be tools that humans had to use… Today software is software that can use tools on behalf of a human. — Kimberly Powell @ 25:50

The first cut isn’t made with a scalpel. It’s made in code. — Video Narrator (J&J) @ 36:03

Key Topics

AI in Healthcare · Drug Discovery · Computational Biology · Generative AI · Agentic AI · Digital Health · Medical Robotics · Physical AI · Digital Twins · Open Source Models · AI Factories · Medical Imaging · Genomics

Takeaways

  • NVIDIA provides a comprehensive full-stack computing platform tailored for healthcare, spanning from hardware infrastructure to domain-specific AI models like Clara and MONAI.
  • The transition from pattern matching to generative design in biology is accelerating drug discovery, supported by new tools like the OpenFold3 NIM.
  • Pharmaceutical leaders like Eli Lilly are building massive ‘AI Factories’ using NVIDIA Blackwell GPUs to process proprietary data into biomedical intelligence.
  • Healthcare is leading the U.S. economy in enterprise AI adoption, driven by the urgent need to close the care gap and automate workflows using Agentic AI.
  • Physical AI and robotics are transforming medical devices and surgical procedures, utilizing platforms like IGX Thor and Omniverse for simulation-first development.
  • NVIDIA is heavily investing in open-source models and platforms to foster a collaborative ecosystem across research institutions, startups, and enterprise healthcare companies.