GTC 2025 Spring Healthcare Special Address
Category: Healthcare Special Address · Year: 2025 · ▶ Watch
Speakers: Kimberly Powell, Vice President of Healthcare and Life Sciences, NVIDIA
Segments (13)
- 00:00 · Introduction
- A video montage showcasing NVIDIA’s impact on healthcare over the years.
- 02:08 · Welcome and GTC Overview
- Kimberly Powell welcomes attendees and highlights the scale of healthcare presence at GTC.
- 03:56 · Accelerated Computing in Healthcare
- A review of how accelerated computing has transformed medical imaging, genomics, and drug discovery.
- 05:56 · Decoding Biology and Evo 2
- Discussion on understanding biology with AI and the announcement of the Evo 2 genomic foundation model.
- 10:53 · Announcing DGX Spark
- Introduction of the DGX Spark, a benchside AI supercomputer for researchers.
- 12:28 · Every Wet Lab Needs a Dry Lab
- Exploring the integration of digital AI labs with physical wet labs using the BioNeMo platform.
- 16:05 · AI Drug Discovery Ecosystem
- Highlighting partners using AI to accelerate drug discovery and improve R&D productivity.
- 17:39 · AI Agents in Healthcare
- Introduction of reasoning models and the NVIDIA AI Agent Platform for healthcare applications.
- 19:55 · Synchron BCI Patient Story
- A video demonstrating how a patient uses an AI-enabled Brain-Computer Interface to interact with his environment.
- 22:50 · Physical AI and Robotics
- Announcing Isaac for Healthcare to build physical AI for medical devices and robotics.
- 24:12 · MONAI Multimodal
- Updates to the MONAI platform, introducing multimodal and agent frameworks for medical imaging.
- 26:48 · Holoscan 3.0 and Ecosystem
- Launch of Holoscan 3.0 for real-time sensor processing and highlights of the growing ecosystem.
- 28:44 · Conclusion
- Summary of the AI revolution in healthcare across drug discovery, digital health, and medical devices.
Product Announcements (8)
- [07:48] Evo 2 Genomic Foundation Model
- The world’s largest biology foundation model.
- specs: Trained on 9 trillion nucleotides, 1 million context length.
- availability: Available now
- [10:53] NVIDIA DGX Spark
- A benchside AI supercomputer.
- specs: 1 petaflop of FP4 performance, fits on a desk.
- availability: Shipping in July
- [14:10] NVIDIA BioNeMo Platform Updates
- Platform for drug discovery.
- specs: New NIMs and Blueprints for generative chemistry and virtual screening.
- availability: Available now
- [19:50] NVIDIA AgentIQ
- Toolkit to connect agents, data, and tools.
- specs: Uses reasoning models to orchestrate complex workflows.
- availability: Available now
- [20:25] Epic with NVIDIA NIM on Microsoft Azure
- Integration of NVIDIA AI into Epic’s EHR.
- specs: Accelerates AI deployment for healthcare providers.
- availability: Available now
- [23:30] Isaac for Healthcare
- Platform to build physical AI for medical devices.
- specs: Includes training, simulation (Omniverse), and deployment (Holoscan).
- availability: Early access
- [24:45] MONAI Multimodal and Agent Framework
- Updates to the medical imaging AI platform.
- specs: Supports DICOM, text, EHR, video, and omics data with reasoning agents.
- availability: Available now
- [26:50] NVIDIA Holoscan 3.0
- Multimodal real-time AI sensor processing platform.
- specs: Dynamic flow control for runtime decision making.
- availability: Available now
Specific Numbers (9)
| Timestamp | Metric | Value | Context |
|---|---|---|---|
| 02:58 | Healthcare companies at GTC | 700+ | Number of healthcare companies participating in the conference. |
| 07:58 | Evo 2 training data | 9 trillion | Number of nucleotides the Evo 2 model was trained on. |
| 08:54 | Evo 2 context length | 1 million | The context window size for the Evo 2 model. |
| 11:28 | DGX Spark performance | 1 petaflop | FP4 compute performance of the DGX Spark. |
| 15:51 | R&D industry spend | $300 billion | Annual spending on R&D and drug discovery. |
| 16:26 | Drug discovery timeline | 13 months | Time taken by Insilico Medicine to go from target to candidate. |
| 19:23 | Synthetically accessible compounds | 1.2 trillion | Number of compounds explored by Cadence’s Orion platform. |
| 24:15 | MONAI downloads | 4.5 million | Total downloads of the MONAI platform. |
| 30:40 | Epic patient records | 325 million | Number of patient records managed by Epic worldwide. |
Benchmark Claims (1)
- [08:55] Context Length: 1 million
- vs: Evo 1 (100k)
- gain: 10x increase in context length.
Customer Stories (7)
- [16:15] Insilico Medicine
- Used AI for drug discovery.
- outcome: Went from target to candidate in 13 months and brought 10 assets to the clinic.
- [17:39] BIOVIA
- Built a generative therapeutics design platform with BioNeMo.
- outcome: Integrated generative chemistry NIMs into their workflow for high-quality synthesizable compounds.
- [18:29] Sapio Sciences
- Built a no-code informatics platform with BioNeMo.
- outcome: Enabled scientists to use AI microservices without writing code.
- [19:02] Cadence
- Built the Orion molecular design platform with BioNeMo.
- outcome: Allowed researchers to efficiently explore 1.2 trillion synthetically accessible compounds.
- [19:55] Synchron
- Developed an implantable Brain-Computer Interface.
- outcome: Enabled a patient with ALS to control smart home devices and communicate using Apple Vision Pro and NVIDIA technology.
- [25:03] Hippocratic AI
- Created a marketplace of healthcare AI agents.
- outcome: Deployed agents for over 400 use cases across 25 specialties.
- [27:00] GE HealthCare
- Partnered with NVIDIA to build physical AI capabilities.
- outcome: Developing autonomous ultrasound and X-ray systems to improve access and early detection.
Key Technologies (7)
- Accelerated Computing: Speeds up complex calculations for imaging, genomics, and simulations.
- Generative AI / Foundation Models: Understands and generates biological sequences (DNA, RNA, proteins).
- BioNeMo: Platform for generative AI in drug discovery.
- AI Agents: Autonomous systems that can reason, plan, and execute tasks.
- Physical AI (Isaac): AI that understands and interacts with the physical world via robotics.
- MONAI: Medical Open Network for AI, used for medical imaging.
- Holoscan: Real-time AI sensor processing platform for medical devices.
Demos Shown (2)
- [19:55] Video of a patient using the Synchron BCI to control an Apple Vision Pro and smart home devices.
- Yes
- [27:16] UI walkthrough of a Biomedical Research Agent Blueprint performing literature search, hypothesis generation, and virtual screening.
- Yes
Predictions / Commitments (3)
- [12:28, Current/Near future] Every wet lab needs a dry lab.
- [21:48, Current/Near future] Reasoning is an incredibly new frontier of large language models.
- [28:46, Current] This is the most exciting time in healthcare in at least my 20 years.
Companies Mentioned (5)
Arc Institute · ASUS, Lenovo, Dell, HPE · Epic, Microsoft Azure · Apple · GE HealthCare
Notable Quotes (2)
Every wet lab needs a dry lab. — Kimberly Powell @ 12:28
This is the most exciting time in healthcare in at least my 20 years. — Kimberly Powell @ 28:46
Key Topics
Accelerated Computing · Generative AI · Genomics · Drug Discovery · BioNeMo · AI Agents · Reasoning Models · Brain-Computer Interfaces · Physical AI · Medical Robotics · Medical Imaging · MONAI · Holoscan · Digital Health
Takeaways
- NVIDIA is driving a massive transformation in healthcare through accelerated computing and AI.
- The Evo 2 model represents a major leap in understanding biology, with a 1 million context length.
- DGX Spark brings petaflop-scale AI supercomputing directly to the researcher’s desk.
- The integration of ‘dry labs’ (AI) with ‘wet labs’ (physical) is essential for modern drug discovery.
- AI agents with reasoning capabilities are automating complex workflows in research and clinical settings.
- Physical AI, powered by platforms like Isaac and Holoscan, is enabling autonomous medical devices and robotics.
- NVIDIA’s ecosystem of partners, from pharma to medical device manufacturers, is rapidly adopting these technologies.