Major Breakthrough Announced in AI Industry

On May 19, 2025, at COMPUTEX Taipei, NVIDIA founder and CEO Jensen Huang delivered a keynote titled "AI Next," sharing the company’s latest advancements and strategic roadmap in artificial intelligence, accelerated computing, and robotics.

image.png

Below are the key highlights from the presentation:

1:Next-Generation Hardware Launches

  • 1.1:Grace Blackwell GB300 System
  • Jensen Huang announced that the Grace Blackwell system, based on the Blackwell architecture, has entered full production, with an upgraded GB300 version slated for release this quarter.
  • The GB300 features enhanced Blackwell chips, delivering 1.5x faster inference performance, 1.5x higher HBM memory capacity, and double the networking bandwidth, transforming standard computers into supercomputers.
  • The next-gen supercomputing system integrates 72 Blackwell processors (144 GPU dies), achieving 130 TB/s bandwidth via NVLink Spine technology, with a total of 1.3 trillion transistors.
  • 1.2:DGX Spark Personal AI Supercomputer Now in Production
  • The DGX Spark, first unveiled at CES, is now in full production and expected to hit the market in the coming weeks.
  • Designed for developers and enterprises, it delivers 1 PetaFLOP of AI compute and supports seamless cloud scalability. Huang remarked, "Everyone can own a DGX Spark by Christmas."
  • 1.3:RTX 5060 GPU & Laptops
  • Huang showcased an MSI laptop powered by the RTX 5060 GPU, confirming its May release. The GPU targets gamers and creators with significantly upgraded performance.

image.png

2:AI and Robotics Convergence

  • 2.1:Robotics: A Trillion-Dollar Opportunity
  • Huang emphasized robotics as the next multi-trillion-dollar industry. NVIDIA’s Isaac Groot platform (powered by Jetson Thor processors) is advancing autonomous driving and human-robot collaboration, including a Mercedes-Benz fleet set to achieve end-to-end autonomy this year.
  • 2.2:Newton Physics AI Engine Goes Open-Source
  • Co-developed with DeepMind and Disney Research, the Newton physics engine will be open-sourced in July.
  • GPU-accelerated and highly differentiable, Newton integrates with the Isaac Simulator to enhance robotic training with real-world physics.
  • 2.3:GR00T: Foundational Model for Humanoid Robots
  • NVIDIA introduced Isaac GR00T N1.5, an open-source foundational model improving environmental adaptability and task execution.
  • The GR00T-Dreams technology generates synthetic training scenarios to address data scarcity in physical AI.

3:Software Ecosystem & Open-Source Strategy

  • 3.1:CUDA-X Libraries for Accelerated Computing
  • The CUDA-X stack now supports applications from 5G/6G networks to Earth-2 climate modeling.
  • Huang highlighted that accelerated computing offers 100x performance gains at 3x power efficiency and 1.5x cost, lowering barriers to enterprise AI adoption.
  • 3.2:NVIDIA NIM Inference Microservices
  • The NIM (Inference Microservices) streamlines generative AI deployment with optimized containers. Models like Meta Llama 3-8B achieve 3x faster inference on NIM.
  • Over 28 million developers can access NIM via platforms like Hugging Face.

4:Global Collaborations & Industry Enablement

  • 4.1:AI Factories & Supercomputers
  • NVIDIA is building its first mega AI supercomputer in Taiwan, alongside a new office, "NVIDIA Constellation."
  • A partnership with TSMC on COOS-L (CoWoS-on-Substrate with Lasers) enables large-scale chip manufacturing.
  • 4.2:NVLink Fusion Open Ecosystem
  • The NVLink Fusion technology allows partners (e.g., Fujitsu, Qualcomm) to build semi-custom AI infrastructure for large-scale model training/inference.
  • 4.3:Industry Case Studies
  • Foxconn leverages Omniverse digital twins to optimize robotic factory layouts.
  • BYD and Siemens adopt the Isaac platform to enhance industrial automation.

Huang concluded: "A decade from now, AI will be as ubiquitous as electricity and the internet." NVIDIA continues to drive innovation—from chips to software—with annual roadmap updates (e.g., the Rubin platform), leading the physical AI and robotics revolution.