Summary of "NVIDIA GTC Washington, D.C. Keynote with CEO Jensen Huang"

Summary of NVIDIA GTC Washington, D.C. Keynote with CEO Jensen Huang

Technological Concepts and Product Features

1. Accelerated Computing and GPU Innovation

NVIDIA invented a new computing model called accelerated computing, overcoming the limits of Moore’s Law and Dennard scaling.
The GPU, paired with the CUDA programming model, enables massive parallelism beyond traditional CPUs.
CUDA has evolved over 30 years with backward compatibility and a rich ecosystem of libraries (CUDA-X), enabling diverse applications from computational lithography (cuLitho) to quantum computing (cuQuantum).
Accelerated computing is now at an inflection point, powering AI, scientific computing, graphics, and more.

2. AI as a New Industrial Revolution

AI is described not just as chatbots but a fundamental reinvention of the computing stack.
AI uses tokenization to represent language, images, video, proteins, chemicals, and more, enabling learning and reasoning across many domains.
AI transforms software from tools (like Excel, Word) into autonomous workers that use tools, augmenting productivity across industries.
The concept of an AI factory is introduced: specialized data centers optimized exclusively for AI token generation at scale and cost efficiency.
AI growth is driven by two exponentials: smarter models needing more compute, and more users demanding more compute.
Extreme co-design (chip, system, software, models) is essential to continue exponential performance and cost improvements.
New NVIDIA technologies include Grace CPU, Blackwell GPU, NVLink 72 interconnect, and Spectrum Ethernet for AI scale-out.

3. NVIDIA Arc and 6G Wireless Technology

Partnership with Nokia to develop NVIDIA Arc: a new product line combining Grace CPU, Blackwell GPU, and ConnectX networking for 6G wireless infrastructure.
Arc enables a software-defined, programmable wireless network with AI-enhanced Radio Access Network (AI for RAN) to improve spectral efficiency and reduce energy consumption.
AI on RAN enables edge cloud computing at base stations, creating a new AI-powered wireless cloud infrastructure.
This partnership aims to restore American leadership in telecommunications technology.

4. Quantum Computing and NVQLink

Quantum computing requires error correction and close integration with classical supercomputers.
NVIDIA announced NVQLink, a high-speed interconnect architecture linking quantum processors (QPUs) with NVIDIA GPUs for error correction, calibration, and hybrid quantum-classical simulations.
CUDA-Q, an extension of CUDA, supports quantum GPU computing and orchestration of quantum devices.
NVQLink is scalable to future quantum systems with tens or hundreds of thousands of qubits.
Broad industry and DOE lab support for NVQLink and quantum GPU computing was announced.

5. AI Supercomputers and AI Factories

Introduction of Vera Rubin (Ruben), a third-generation NVLink 72 rack-scale AI supercomputer with 72 GPUs interconnected via NVLink switches.
The Blackwell GPU generation delivers 10x performance improvements and 10x lower cost per AI token compared to previous generations.
NVIDIA is manufacturing Blackwell GPUs and AI systems in the U.S., reindustrializing American semiconductor and AI hardware production.
AI factories are designed and optimized digitally using NVIDIA Omniverse DSX, a digital twin platform that simulates and manages AI infrastructure, power, cooling, and operations.
DSX enables faster AI factory build times and operational optimization, partnering with companies like Siemens, Schneider Electric, and Vertiv.

6. Open Source AI Models and Ecosystem

NVIDIA leads in open-source AI model contributions with 23 models across language, physical AI, robotics, and biology.
Open source is critical for startups and innovation; NVIDIA supports a rich ecosystem of GPU clouds and AI startups.
NVIDIA collaborates with major cloud providers (AWS, Google Cloud, Microsoft Azure, Oracle) to integrate GPUs, libraries, and models.
Enterprise partnerships include CrowdStrike (cybersecurity AI), Palantir (data analytics acceleration), SAP (enterprise AI), Synopsis, and Cadence (EDA tools with AI acceleration).

7. Physical AI and Robotics

Physical AI requires three types of computers: training (Grace Blackwell), simulation/digital twin (Omniverse computer), and robotic operation (Thor Jetson robotics computer).
Digital twins simulate factories, robots, and workflows for design, optimization, and training.
NVIDIA partners with Foxconn, Siemens, Caterpillar, Figure, Agility, Johnson & Johnson, and Disney Research for robotic factories, warehouse automation, surgical robots, humanoid robots, and interactive AI agents.
Demonstrated advanced robot simulation with Disney’s “Blue” robot in Omniverse, emphasizing real-time physics-based learning.

8. Autonomous Vehicles and NVIDIA Drive Hyperion

NVIDIA Drive Hyperion is a standardized AI platform for autonomous vehicles, enabling safety through comprehensive sensor suites (cameras, radar, lidar).
Hyperion is adopted by automakers like Lucid, Mercedes-Benz, Stellantis, and supports many AV developers (Wayve, Waabi, Aurora, Momenta).
Partnership with Uber to integrate Hyperion-powered robo taxis into a global network.
The robo taxi market is poised for significant growth, augmenting tens of millions of taxis worldwide.

Key Announcements and Guides

Launch of NVIDIA Arc for 6G wireless infrastructure with Nokia partnership.
Introduction of NVQLink for quantum-classical computing integration.
Reveal of Vera Rubin AI supercomputer and Blackwell GPU generation with 10x performance and cost efficiency.
Deployment of Omniverse DSX digital twin platform for AI factory design and operation.
Expansion of open-source AI models and ecosystem support.
New enterprise AI partnerships with CrowdStrike, Palantir, SAP, Synopsis, and Cadence.
Updates on physical AI and robotics computing platforms and partnerships.
NVIDIA Drive Hyperion platform adoption and Uber partnership for robo taxis.

Main Speakers and Sources

Jensen Huang, NVIDIA Founder and CEO — primary speaker and presenter of keynote.
Mentioned partners and collaborators include:
- Nokia (6G wireless)
- DOE Labs (quantum computing)
- Foxconn, Siemens, Caterpillar, Figure, Agility, Johnson & Johnson, Disney Research (robotics and manufacturing)
- Cloud providers: AWS, Google Cloud, Microsoft Azure, Oracle
- Enterprise partners: CrowdStrike, Palantir, SAP, Synopsis, Cadence
- Uber (autonomous vehicle network)

This keynote highlights NVIDIA’s comprehensive strategy spanning hardware innovation, AI software and models, quantum computing, telecommunications, robotics, and manufacturing — all underpinned by a vision of AI-powered industrial transformation and American reindustrialization.