Jensen Huang delivered a speech at Computex 2025, predicting that within ten years, all industries will be disrupted by generative and simulative AI, and the key to productivity will shift to discussing Token generation, which will become the new landscape of human work and lifestyle.
(Previous Context: Jensen Huang: NVIDIA Taiwan Headquarters to be Located in Beishi Technology Park! )
(Background Information: Huawei Claims to Surpass Nvidia, Ascend 910D Reportedly "Outperforming H100" with Samples Potentially Shipping in May )
Nvidia's founder and CEO Jensen Huang, during his Computex 2025 keynote speech, once again outlined the AI-driven future of the global tech industry, claiming that Nvidia has transformed from a technology platform to a more crucial entity. He positioned the company as a global reform promoter and AI infrastructure advocate, introducing AI's latest evolution from perception and generation to agency and physical intelligence, and predicting the imminent arrival of a new era of AI empowerment.
[The rest of the translation follows the same professional and accurate approach, maintaining the original structure and meaning while translating to English.]New Products and Technological Breakthroughs
The most anticipated focus of this presentation was undoubtedly Nvidia's new generation of Blackwell architecture and its flagship product Grace Blackwell superchip (such as GB200/GB300). Huang Renxun pointed out that Blackwell achieves unprecedented computing scale and efficiency through revolutionary MVLink high-speed interconnect technology and advanced liquid cooling systems, and will become the core engine driving the next generation of AI factories. To address different market needs, Huang Renxun also launched a series of innovative products:
- DGX Spark and New DGX Station: Desktop and personal AI supercomputers designed for AI developers, researchers, and professionals, making cutting-edge AI computing power readily accessible.
- Nvidia AI Data Platform and AIQ Query System: Providing high-performance, AI-driven storage and semantic search solutions for enterprise unstructured data.
- Robotics Technology Platform: A brand new Isaac Groot platform, Jetson Thor robot processor, and Nvidia Cosmos and Groot Dreams models for large-scale synthetic data generation, aimed at accelerating the development and deployment of various types of robots.
- MVLink Fusion Open Platform: Allowing partners to integrate their own ASIC or CPU into Nvidia's MVLink ecosystem to jointly create customized, high-performance AI infrastructure.
Robotics and Digital Twins: Deep Collaboration with Taiwanese Manufacturers
Finally, Huang Renxun also demonstrated robotic language capabilities and digital twin cases with eight Taiwanese manufacturers. Robots can now recognize human language and make corresponding actions, indicating that robots will become fully functional production tools in the future. The slides claimed that future AI language-based robot systems could replace labor in thousands of factories.
The presentation also showcased Omniverse's production optimization in eight Taiwanese factories, including Foxconn,
Wistron, Pegatron, Quanta, Wiwynn, Delta, Gigabyte, and TSMC, which have simulated factory pipelines, thermal control, and reduced production scrap rates in various physical simulation scenarios. When each digital twin system is connected to real machines, a real-time interactive dashboard will appear, capable of monitoring and identifying potential causes of yield reduction on the production line. Therefore, all production will be optimized based on Omniverse and digital twin systems:
GeForce brought AI to this world, and artificial intelligence has come back. Now undergoing transformation, AI will return and change everything you have experienced. I am glad to witness this with you.
