- Nvidia CEO Jensen Huang announced Nvidia Arc, a software-defined programmable computer integrating gray CPU, Blackwell GPU, and ConnectX Melanox Connect X networking.
- Nvidia Arc will enhance 6G and AI capabilities in millions of Nokia base stations worldwide, improving spectral efficiency through AI and reinforcement learning.
- Nvidia’s CUDA X library introduces a fundamentally different programming model for wireless communication, requiring new algorithms and application rewrites for GPU compatibility.
Quantum Computing Innovations
- Nvidia’s MVQLink technology seamlessly connects quantum processors with NVIDIA GPUs, facilitating quantum error correction by transferring terabytes of data thousands of times per second.
- The CUDA Q platform allows researchers to run quantum GPU applications, enabling hybrid simulations of quantum and classical systems using MVQLink.
AI and Supercomputing
- Nvidia’s Grace Blackwell AI supercomputer features 72 GPUs and MVLink 72 with 130 terabytes per second bandwidth, nearly the global internet’s peak traffic.
- The AI factory represents a fundamentally different computing stack, replacing traditional CPUs and Windows with a new architecture.
- Nvidia’s Grace Blackwell is 100% liquid cooled and completely cableless, designed for a gigawatt data center housing 9,000 racks.
Partnerships and Open Source
- Partnering with CrowdStrike, Nvidia is developing a cybersecurity AI system for real-time threat detection.
- Nvidia collaborates with Palantir to create the fastest enterprise stack globally.
- Nvidia is committed to open-source models, with 23 models leading across domains like language and robotics, integrated into platforms like AWS and Google Cloud.
Automotive and Robotics
- The Hyperion architecture empowers car manufacturers to build robo taxi-ready vehicles with comprehensive safety features, integrated into brands like Lucid and Mercedes-Benz.
- Nvidia’s co-design approach leverages the simultaneous design of chips and systems for exponential performance benefits.